19 files changed, 198 insertions, 77 deletions
diff --git a/Eigen/src/Core/DenseStorage.h b/Eigen/src/Core/DenseStorage.h
index 82201d96a..7958feeb9 100644
--- a/Eigen/src/Core/DenseStorage.h
+++ b/Eigen/src/Core/DenseStorage.h
@@ -13,9 +13,9 @@
 #define EIGEN_MATRIXSTORAGE_H
 
 #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN
-  #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN EIGEN_DENSE_STORAGE_CTOR_PLUGIN;
+  #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X) X; EIGEN_DENSE_STORAGE_CTOR_PLUGIN;
 #else
-  #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+  #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X)
 #endif
 
 namespace Eigen {
@@ -184,12 +184,16 @@ template<typename T, int Size, int _Rows, int _Cols, int _Options> class DenseSt
 {
     internal::plain_array<T,Size,_Options> m_data;
   public:
-    EIGEN_DEVICE_FUNC DenseStorage() {}
+    EIGEN_DEVICE_FUNC DenseStorage() {
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size)
+    }
     EIGEN_DEVICE_FUNC
     explicit DenseStorage(internal::constructor_without_unaligned_array_assert)
       : m_data(internal::constructor_without_unaligned_array_assert()) {}
     EIGEN_DEVICE_FUNC 
-    DenseStorage(const DenseStorage& other) : m_data(other.m_data) {}
+    DenseStorage(const DenseStorage& other) : m_data(other.m_data) {
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size)
+    }
     EIGEN_DEVICE_FUNC 
     DenseStorage& operator=(const DenseStorage& other)
     { 
@@ -197,7 +201,7 @@ template<typename T, int Size, int _Rows, int _Cols, int _Options> class DenseSt
       return *this; 
     }
     EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) {
-      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       eigen_internal_assert(size==rows*cols && rows==_Rows && cols==_Cols);
       EIGEN_UNUSED_VARIABLE(size);
       EIGEN_UNUSED_VARIABLE(rows);
@@ -343,7 +347,7 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
     EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols)
       : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(rows), m_cols(cols)
     {
-      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       eigen_internal_assert(size==rows*cols && rows>=0 && cols >=0);
     }
     EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other)
@@ -351,6 +355,7 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
       , m_rows(other.m_rows)
       , m_cols(other.m_cols)
     {
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*m_cols)
       internal::smart_copy(other.m_data, other.m_data+other.m_rows*other.m_cols, m_data);
     }
     EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
@@ -403,7 +408,7 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
           m_data = internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size);
         else
           m_data = 0;
-        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       }
       m_rows = rows;
       m_cols = cols;
@@ -422,7 +427,7 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
     explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {}
     EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_cols(cols)
     {
-      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       eigen_internal_assert(size==rows*cols && rows==_Rows && cols >=0);
       EIGEN_UNUSED_VARIABLE(rows);
     }
@@ -430,6 +435,7 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
       : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(_Rows*other.m_cols))
       , m_cols(other.m_cols)
     {
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_cols*_Rows)
       internal::smart_copy(other.m_data, other.m_data+_Rows*m_cols, m_data);
     }
     EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
@@ -477,7 +483,7 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
           m_data = internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size);
         else
           m_data = 0;
-        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       }
       m_cols = cols;
     }
@@ -495,7 +501,7 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
     explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {}
     EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(rows)
     {
-      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       eigen_internal_assert(size==rows*cols && rows>=0 && cols == _Cols);
       EIGEN_UNUSED_VARIABLE(cols);
     }
@@ -503,6 +509,7 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
       : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(other.m_rows*_Cols))
       , m_rows(other.m_rows)
     {
+      EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*_Cols)
       internal::smart_copy(other.m_data, other.m_data+other.m_rows*_Cols, m_data);
     }
     EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
@@ -550,7 +557,7 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
           m_data = internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size);
         else
           m_data = 0;
-        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       }
       m_rows = rows;
     }
diff --git a/Eigen/src/Core/PlainObjectBase.h b/Eigen/src/Core/PlainObjectBase.h
index 639fb92bf..77f4f6066 100644
--- a/Eigen/src/Core/PlainObjectBase.h
+++ b/Eigen/src/Core/PlainObjectBase.h
@@ -812,6 +812,13 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       this->_set_noalias(other);
     }
 
+    // Initialize an arbitrary matrix from an object convertible to the Derived type.
+    template<typename T>
+    EIGEN_DEVICE_FUNC
+    EIGEN_STRONG_INLINE void _init1(const Derived& other){
+      this->_set_noalias(other);
+    }
+
     // Initialize an arbitrary matrix from a generic Eigen expression
     template<typename T, typename OtherDerived>
     EIGEN_DEVICE_FUNC
@@ -834,7 +841,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       this->derived() = r;
     }
     
-    // For fixed -size arrays:
+    // For fixed-size Array<Scalar,...>
     template<typename T>
     EIGEN_DEVICE_FUNC
     EIGEN_STRONG_INLINE void _init1(const Scalar& val0,
@@ -846,6 +853,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       Base::setConstant(val0);
     }
     
+    // For fixed-size Array<Index,...>
     template<typename T>
     EIGEN_DEVICE_FUNC
     EIGEN_STRONG_INLINE void _init1(const Index& val0,
diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h
index d392bf3ff..84a56bdcc 100644
--- a/Eigen/src/Core/arch/NEON/PacketMath.h
+++ b/Eigen/src/Core/arch/NEON/PacketMath.h
@@ -46,7 +46,7 @@ typedef uint32x4_t  Packet4ui;
   const Packet4f p4f_##NAME = pset1<Packet4f>(X)
 
 #define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \
-  const Packet4f p4f_##NAME = vreinterpretq_f32_u32(pset1<int>(X))
+  const Packet4f p4f_##NAME = vreinterpretq_f32_u32(pset1<int32_t>(X))
 
 #define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \
   const Packet4i p4i_##NAME = pset1<Packet4i>(X)
@@ -83,7 +83,7 @@ template<> struct packet_traits<float>  : default_packet_traits
     HasSqrt = 0
   };
 };
-template<> struct packet_traits<int>    : default_packet_traits
+template<> struct packet_traits<int32_t>    : default_packet_traits
 {
   typedef Packet4i type;
   typedef Packet4i half; // Packet2i intrinsics not implemented yet
@@ -105,11 +105,11 @@ EIGEN_STRONG_INLINE void        vst1q_f32(float* to, float32x4_t from) { ::vst1q
 EIGEN_STRONG_INLINE void        vst1_f32 (float* to, float32x2_t from) { ::vst1_f32 ((float32_t*)to,from); }
 #endif
 
-template<> struct unpacket_traits<Packet4f> { typedef float  type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; };
-template<> struct unpacket_traits<Packet4i> { typedef int    type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; };
+template<> struct unpacket_traits<Packet4f> { typedef float   type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; };
+template<> struct unpacket_traits<Packet4i> { typedef int32_t type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; };
 
 template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float&  from) { return vdupq_n_f32(from); }
-template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int&    from)   { return vdupq_n_s32(from); }
+template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int32_t&    from)   { return vdupq_n_s32(from); }
 
 template<> EIGEN_STRONG_INLINE Packet4f plset<Packet4f>(const float& a)
 {
@@ -117,7 +117,7 @@ template<> EIGEN_STRONG_INLINE Packet4f plset<Packet4f>(const float& a)
   Packet4f countdown = vld1q_f32(f);
   return vaddq_f32(pset1<Packet4f>(a), countdown);
 }
-template<> EIGEN_STRONG_INLINE Packet4i plset<Packet4i>(const int& a)
+template<> EIGEN_STRONG_INLINE Packet4i plset<Packet4i>(const int32_t& a)
 {
   const int32_t i[] = {0, 1, 2, 3};
   Packet4i countdown = vld1q_s32(i);
@@ -240,20 +240,20 @@ template<> EIGEN_STRONG_INLINE Packet4f pandnot<Packet4f>(const Packet4f& a, con
 }
 template<> EIGEN_STRONG_INLINE Packet4i pandnot<Packet4i>(const Packet4i& a, const Packet4i& b) { return vbicq_s32(a,b); }
 
-template<> EIGEN_STRONG_INLINE Packet4f pload<Packet4f>(const float* from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_f32(from); }
-template<> EIGEN_STRONG_INLINE Packet4i pload<Packet4i>(const int*   from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_s32(from); }
+template<> EIGEN_STRONG_INLINE Packet4f pload<Packet4f>(const float*    from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_f32(from); }
+template<> EIGEN_STRONG_INLINE Packet4i pload<Packet4i>(const int32_t*  from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_s32(from); }
 
-template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f32(from); }
-template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int* from)   { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_s32(from); }
+template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float*   from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f32(from); }
+template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int32_t* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_s32(from); }
 
-template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float*   from)
+template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float* from)
 {
   float32x2_t lo, hi;
   lo = vld1_dup_f32(from);
   hi = vld1_dup_f32(from+1);
   return vcombine_f32(lo, hi);
 }
-template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int*     from)
+template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int32_t* from)
 {
   int32x2_t lo, hi;
   lo = vld1_dup_s32(from);
@@ -261,11 +261,11 @@ template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int*     from)
   return vcombine_s32(lo, hi);
 }
 
-template<> EIGEN_STRONG_INLINE void pstore<float>(float*   to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f32(to, from); }
-template<> EIGEN_STRONG_INLINE void pstore<int>(int*       to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_s32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstore<float>  (float*    to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstore<int32_t>(int32_t*  to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_s32(to, from); }
 
-template<> EIGEN_STRONG_INLINE void pstoreu<float>(float*  to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); }
-template<> EIGEN_STRONG_INLINE void pstoreu<int>(int*      to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstoreu<float>  (float*   to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); }
+template<> EIGEN_STRONG_INLINE void pstoreu<int32_t>(int32_t* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); }
 
 template<> EIGEN_DEVICE_FUNC inline Packet4f pgather<float, Packet4f>(const float* from, Index stride)
 {
@@ -276,7 +276,7 @@ template<> EIGEN_DEVICE_FUNC inline Packet4f pgather<float, Packet4f>(const floa
   res = vsetq_lane_f32(from[3*stride], res, 3);
   return res;
 }
-template<> EIGEN_DEVICE_FUNC inline Packet4i pgather<int, Packet4i>(const int* from, Index stride)
+template<> EIGEN_DEVICE_FUNC inline Packet4i pgather<int32_t, Packet4i>(const int32_t* from, Index stride)
 {
   Packet4i res = pset1<Packet4i>(0);
   res = vsetq_lane_s32(from[0*stride], res, 0);
@@ -293,7 +293,7 @@ template<> EIGEN_DEVICE_FUNC inline void pscatter<float, Packet4f>(float* to, co
   to[stride*2] = vgetq_lane_f32(from, 2);
   to[stride*3] = vgetq_lane_f32(from, 3);
 }
-template<> EIGEN_DEVICE_FUNC inline void pscatter<int, Packet4i>(int* to, const Packet4i& from, Index stride)
+template<> EIGEN_DEVICE_FUNC inline void pscatter<int32_t, Packet4i>(int32_t* to, const Packet4i& from, Index stride)
 {
   to[stride*0] = vgetq_lane_s32(from, 0);
   to[stride*1] = vgetq_lane_s32(from, 1);
@@ -301,12 +301,12 @@ template<> EIGEN_DEVICE_FUNC inline void pscatter<int, Packet4i>(int* to, const
   to[stride*3] = vgetq_lane_s32(from, 3);
 }
 
-template<> EIGEN_STRONG_INLINE void prefetch<float>(const float* addr) { EIGEN_ARM_PREFETCH(addr); }
-template<> EIGEN_STRONG_INLINE void prefetch<int>(const int*     addr) { EIGEN_ARM_PREFETCH(addr); }
+template<> EIGEN_STRONG_INLINE void prefetch<float>  (const float*    addr) { EIGEN_ARM_PREFETCH(addr); }
+template<> EIGEN_STRONG_INLINE void prefetch<int32_t>(const int32_t*  addr) { EIGEN_ARM_PREFETCH(addr); }
 
 // FIXME only store the 2 first elements ?
-template<> EIGEN_STRONG_INLINE float  pfirst<Packet4f>(const Packet4f& a) { float EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; }
-template<> EIGEN_STRONG_INLINE int    pfirst<Packet4i>(const Packet4i& a) { int   EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; }
+template<> EIGEN_STRONG_INLINE float   pfirst<Packet4f>(const Packet4f& a) { float   EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; }
+template<> EIGEN_STRONG_INLINE int32_t pfirst<Packet4i>(const Packet4i& a) { int32_t EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; }
 
 template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) {
   float32x2_t a_lo, a_hi;
@@ -361,7 +361,7 @@ template<> EIGEN_STRONG_INLINE Packet4f preduxp<Packet4f>(const Packet4f* vecs)
   return sum;
 }
 
-template<> EIGEN_STRONG_INLINE int predux<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, sum;
 
@@ -408,7 +408,7 @@ template<> EIGEN_STRONG_INLINE float predux_mul<Packet4f>(const Packet4f& a)
 
   return vget_lane_f32(prod, 0);
 }
-template<> EIGEN_STRONG_INLINE int predux_mul<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux_mul<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, prod;
 
@@ -436,7 +436,7 @@ template<> EIGEN_STRONG_INLINE float predux_min<Packet4f>(const Packet4f& a)
   return vget_lane_f32(min, 0);
 }
 
-template<> EIGEN_STRONG_INLINE int predux_min<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux_min<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, min;
 
@@ -461,7 +461,7 @@ template<> EIGEN_STRONG_INLINE float predux_max<Packet4f>(const Packet4f& a)
   return vget_lane_f32(max, 0);
 }
 
-template<> EIGEN_STRONG_INLINE int predux_max<Packet4i>(const Packet4i& a)
+template<> EIGEN_STRONG_INLINE int32_t predux_max<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, max;
 
diff --git a/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h b/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h
index 5cd2794a4..7122efa60 100644
--- a/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h
+++ b/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h
@@ -148,7 +148,7 @@ struct tribb_kernel
     ResMapper res(_res, resStride);
     gebp_kernel<LhsScalar, RhsScalar, Index, ResMapper, mr, nr, ConjLhs, ConjRhs> gebp_kernel;
 
-    Matrix<ResScalar,BlockSize,BlockSize,ColMajor> buffer;
+    Matrix<ResScalar,BlockSize,BlockSize,ColMajor> buffer((internal::constructor_without_unaligned_array_assert()));
 
     // let's process the block per panel of actual_mc x BlockSize,
     // again, each is split into three parts, etc.
diff --git a/Eigen/src/Core/products/TriangularMatrixMatrix.h b/Eigen/src/Core/products/TriangularMatrixMatrix.h
index 8a2f7cd78..6ec5a8a0b 100644
--- a/Eigen/src/Core/products/TriangularMatrixMatrix.h
+++ b/Eigen/src/Core/products/TriangularMatrixMatrix.h
@@ -137,7 +137,7 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix<Scalar,Index,Mode,true,
     ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA());
     ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB());
 
-    Matrix<Scalar,SmallPanelWidth,SmallPanelWidth,LhsStorageOrder> triangularBuffer;
+    Matrix<Scalar,SmallPanelWidth,SmallPanelWidth,LhsStorageOrder> triangularBuffer((internal::constructor_without_unaligned_array_assert()));
     triangularBuffer.setZero();
     if((Mode&ZeroDiag)==ZeroDiag)
       triangularBuffer.diagonal().setZero();
@@ -284,7 +284,7 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix<Scalar,Index,Mode,false,
     ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA());
     ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB());
 
-    Matrix<Scalar,SmallPanelWidth,SmallPanelWidth,RhsStorageOrder> triangularBuffer;
+    Matrix<Scalar,SmallPanelWidth,SmallPanelWidth,RhsStorageOrder> triangularBuffer((internal::constructor_without_unaligned_array_assert()));
     triangularBuffer.setZero();
     if((Mode&ZeroDiag)==ZeroDiag)
       triangularBuffer.diagonal().setZero();
diff --git a/Eigen/src/Core/util/Macros.h b/Eigen/src/Core/util/Macros.h
index 12531e342..29c796647 100644
--- a/Eigen/src/Core/util/Macros.h
+++ b/Eigen/src/Core/util/Macros.h
@@ -837,7 +837,7 @@ namespace Eigen {
 // just an empty macro !
 #define EIGEN_EMPTY
 
-#if EIGEN_COMP_MSVC_STRICT && (EIGEN_COMP_MSVC < 1900 ||  __CUDACC_VER__) // for older MSVC versions, as well as 1900 && CUDA 8, using the base operator is sufficient (cf Bugs 1000, 1324)
+#if EIGEN_COMP_MSVC_STRICT && (EIGEN_COMP_MSVC < 1900 ||  defined(__CUDACC_VER__)) // for older MSVC versions, as well as 1900 && CUDA 8, using the base operator is sufficient (cf Bugs 1000, 1324)
   #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \
     using Base::operator =;
 #elif EIGEN_COMP_CLANG // workaround clang bug (see http://forum.kde.org/viewtopic.php?f=74&t=102653)
diff --git a/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h b/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h
index a9f56c4f5..9ddd553f2 100644
--- a/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h
+++ b/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h
@@ -414,7 +414,8 @@ SelfAdjointEigenSolver<MatrixType>& SelfAdjointEigenSolver<MatrixType>
 
   if(n==1)
   {
-    m_eivalues.coeffRef(0,0) = numext::real(matrix.diagonal()[0]);
+    m_eivec = matrix;
+    m_eivalues.coeffRef(0,0) = numext::real(m_eivec.coeff(0,0));
     if(computeEigenvectors)
       m_eivec.setOnes(n,n);
     m_info = Success;
diff --git a/Eigen/src/Householder/BlockHouseholder.h b/Eigen/src/Householder/BlockHouseholder.h
index 39bf8c83d..01a7ed188 100644
--- a/Eigen/src/Householder/BlockHouseholder.h
+++ b/Eigen/src/Householder/BlockHouseholder.h
@@ -87,7 +87,8 @@ void apply_block_householder_on_the_left(MatrixType& mat, const VectorsType& vec
   const TriangularView<const VectorsType, UnitLower> V(vectors);
 
   // A -= V T V^* A
-  Matrix<typename MatrixType::Scalar,VectorsType::ColsAtCompileTime,MatrixType::ColsAtCompileTime,0,
+  Matrix<typename MatrixType::Scalar,VectorsType::ColsAtCompileTime,MatrixType::ColsAtCompileTime,
+         (VectorsType::MaxColsAtCompileTime==1 && MatrixType::MaxColsAtCompileTime!=1)?RowMajor:ColMajor,
          VectorsType::MaxColsAtCompileTime,MatrixType::MaxColsAtCompileTime> tmp = V.adjoint() * mat;
   // FIXME add .noalias() once the triangular product can work inplace
   if(forward) tmp = T.template triangularView<Upper>()           * tmp;
diff --git a/Eigen/src/SVD/JacobiSVD.h b/Eigen/src/SVD/JacobiSVD.h
index 1337ae987..43488b1e0 100644
--- a/Eigen/src/SVD/JacobiSVD.h
+++ b/Eigen/src/SVD/JacobiSVD.h
@@ -112,9 +112,11 @@ public:
     ColsAtCompileTime = MatrixType::ColsAtCompileTime,
     MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime,
     MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime,
-    Options = MatrixType::Options
+    TrOptions = RowsAtCompileTime==1 ? (MatrixType::Options & ~(RowMajor))
+              : ColsAtCompileTime==1 ? (MatrixType::Options |   RowMajor)
+              : MatrixType::Options
   };
-  typedef Matrix<Scalar, ColsAtCompileTime, RowsAtCompileTime, Options, MaxColsAtCompileTime, MaxRowsAtCompileTime>
+  typedef Matrix<Scalar, ColsAtCompileTime, RowsAtCompileTime, TrOptions, MaxColsAtCompileTime, MaxRowsAtCompileTime>
           TransposeTypeWithSameStorageOrder;
 
   void allocate(const JacobiSVD<MatrixType, FullPivHouseholderQRPreconditioner>& svd)
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index 84a21b3df..d337594f5 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -151,6 +151,7 @@ ei_add_test(packetmath "-DEIGEN_FAST_MATH=1")
 ei_add_test(unalignedassert)
 ei_add_test(vectorization_logic)
 ei_add_test(basicstuff)
+ei_add_test(constructor)
 ei_add_test(linearstructure)
 ei_add_test(integer_types)
 ei_add_test(unalignedcount)
diff --git a/test/constructor.cpp b/test/constructor.cpp
new file mode 100644
index 000000000..eec9e2192
--- /dev/null
+++ b/test/constructor.cpp
@@ -0,0 +1,84 @@
+// This file is part of Eigen, a lightweight C++ template library
+// for linear algebra.
+//
+// Copyright (C) 2017 Gael Guennebaud <gael.guennebaud@inria.fr>
+//
+// This Source Code Form is subject to the terms of the Mozilla
+// Public License v. 2.0. If a copy of the MPL was not distributed
+// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
+
+
+#define TEST_ENABLE_TEMPORARY_TRACKING
+
+#include "main.h"
+
+template<typename MatrixType> struct Wrapper
+{
+  MatrixType m_mat;
+  inline Wrapper(const MatrixType &x) : m_mat(x) {}
+  inline operator const MatrixType& () const { return m_mat; }
+  inline operator MatrixType& () { return m_mat; }
+};
+
+template<typename MatrixType> void ctor_init1(const MatrixType& m)
+{
+  // Check logic in PlainObjectBase::_init1
+  Index rows = m.rows();
+  Index cols = m.cols();
+
+  MatrixType m0 = MatrixType::Random(rows,cols);
+
+  VERIFY_EVALUATION_COUNT( MatrixType m1(m0), 1);
+  VERIFY_EVALUATION_COUNT( MatrixType m2(m0+m0), 1);
+  VERIFY_EVALUATION_COUNT( MatrixType m2(m0.block(0,0,rows,cols)) , 1);
+
+  Wrapper<MatrixType> wrapper(m0);
+  VERIFY_EVALUATION_COUNT( MatrixType m3(wrapper) , 1);
+}
+
+
+void test_constructor()
+{
+  for(int i = 0; i < g_repeat; i++) {
+    CALL_SUBTEST_1( ctor_init1(Matrix<float, 1, 1>()) );
+    CALL_SUBTEST_1( ctor_init1(Matrix4d()) );
+    CALL_SUBTEST_1( ctor_init1(MatrixXcf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
+    CALL_SUBTEST_1( ctor_init1(MatrixXi(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
+  }
+  {
+    Matrix<Index,1,1> a(123);
+    VERIFY_IS_EQUAL(a[0], 123);
+  }
+  {
+    Matrix<Index,1,1> a(123.0);
+    VERIFY_IS_EQUAL(a[0], 123);
+  }
+  {
+    Matrix<float,1,1> a(123);
+    VERIFY_IS_EQUAL(a[0], 123.f);
+  }
+  {
+    Array<Index,1,1> a(123);
+    VERIFY_IS_EQUAL(a[0], 123);
+  }
+  {
+    Array<Index,1,1> a(123.0);
+    VERIFY_IS_EQUAL(a[0], 123);
+  }
+  {
+    Array<float,1,1> a(123);
+    VERIFY_IS_EQUAL(a[0], 123.f);
+  }
+  {
+    Array<Index,3,3> a(123);
+    VERIFY_IS_EQUAL(a(4), 123);
+  }
+  {
+    Array<Index,3,3> a(123.0);
+    VERIFY_IS_EQUAL(a(4), 123);
+  }
+  {
+    Array<float,3,3> a(123);
+    VERIFY_IS_EQUAL(a(4), 123.f);
+  }
+}
diff --git a/test/jacobisvd.cpp b/test/jacobisvd.cpp
index 3d8d0203d..7f5f71562 100644
--- a/test/jacobisvd.cpp
+++ b/test/jacobisvd.cpp
@@ -101,6 +101,12 @@ void test_jacobisvd()
     // Test on inf/nan matrix
     CALL_SUBTEST_7(  (svd_inf_nan<JacobiSVD<MatrixXf>, MatrixXf>()) );
     CALL_SUBTEST_10( (svd_inf_nan<JacobiSVD<MatrixXd>, MatrixXd>()) );
+
+    // bug1395 test compile-time vectors as input
+    CALL_SUBTEST_13(( jacobisvd_verify_assert(Matrix<double,6,1>()) ));
+    CALL_SUBTEST_13(( jacobisvd_verify_assert(Matrix<double,1,6>()) ));
+    CALL_SUBTEST_13(( jacobisvd_verify_assert(Matrix<double,Dynamic,1>(r)) ));
+    CALL_SUBTEST_13(( jacobisvd_verify_assert(Matrix<double,1,Dynamic>(c)) ));
   }
 
   CALL_SUBTEST_7(( jacobisvd<MatrixXf>(MatrixXf(internal::random<int>(EIGEN_TEST_MAX_SIZE/4, EIGEN_TEST_MAX_SIZE/2), internal::random<int>(EIGEN_TEST_MAX_SIZE/4, EIGEN_TEST_MAX_SIZE/2))) ));
diff --git a/test/main.h b/test/main.h
index 1d5bdc1c4..25d2dcf43 100644
--- a/test/main.h
+++ b/test/main.h
@@ -41,6 +41,7 @@
 #include <complex>
 #include <deque>
 #include <queue>
+#include <cassert>
 #include <list>
 #if __cplusplus >= 201103L
 #include <random>
@@ -79,10 +80,12 @@
 #ifdef TEST_ENABLE_TEMPORARY_TRACKING
 
 static long int nb_temporaries;
+static long int nb_temporaries_on_assert = -1;
 
 inline void on_temporary_creation(long int size) {
   // here's a great place to set a breakpoint when debugging failures in this test!
   if(size!=0) nb_temporaries++;
+  if(nb_temporaries_on_assert>0) assert(nb_temporaries<nb_temporaries_on_assert);
 }
 
 #define EIGEN_DENSE_STORAGE_CTOR_PLUGIN { on_temporary_creation(size); }
diff --git a/test/permutationmatrices.cpp b/test/permutationmatrices.cpp
index 70b469ebc..db1266579 100644
--- a/test/permutationmatrices.cpp
+++ b/test/permutationmatrices.cpp
@@ -37,8 +37,7 @@ template<typename MatrixType> void permutationmatrices(const MatrixType& m)
   RightPermutationType rp(rv);
   MatrixType m_permuted = MatrixType::Random(rows,cols);
   
-  const int one_if_dynamic = MatrixType::SizeAtCompileTime==Dynamic ? 1 : 0;
-  VERIFY_EVALUATION_COUNT(m_permuted = lp * m_original * rp, one_if_dynamic); // 1 temp for sub expression "lp * m_original"
+  VERIFY_EVALUATION_COUNT(m_permuted = lp * m_original * rp, 1); // 1 temp for sub expression "lp * m_original"
 
   for (int i=0; i<rows; i++)
     for (int j=0; j<cols; j++)
@@ -50,7 +49,7 @@ template<typename MatrixType> void permutationmatrices(const MatrixType& m)
   VERIFY_IS_APPROX(m_permuted, lm*m_original*rm);
   
   m_permuted = m_original;
-  VERIFY_EVALUATION_COUNT(m_permuted = lp * m_permuted * rp, one_if_dynamic);
+  VERIFY_EVALUATION_COUNT(m_permuted = lp * m_permuted * rp, 1);
   VERIFY_IS_APPROX(m_permuted, lm*m_original*rm);
   
   VERIFY_IS_APPROX(lp.inverse()*m_permuted*rp.inverse(), m_original);
@@ -75,19 +74,19 @@ template<typename MatrixType> void permutationmatrices(const MatrixType& m)
   
   // check inplace permutations
   m_permuted = m_original;
-  VERIFY_EVALUATION_COUNT(m_permuted.noalias()= lp.inverse() * m_permuted, one_if_dynamic); // 1 temp to allocate the mask
+  VERIFY_EVALUATION_COUNT(m_permuted.noalias()= lp.inverse() * m_permuted, 1); // 1 temp to allocate the mask
   VERIFY_IS_APPROX(m_permuted, lp.inverse()*m_original);
   
   m_permuted = m_original;
-  VERIFY_EVALUATION_COUNT(m_permuted.noalias() = m_permuted * rp.inverse(), one_if_dynamic); // 1 temp to allocate the mask
+  VERIFY_EVALUATION_COUNT(m_permuted.noalias() = m_permuted * rp.inverse(), 1); // 1 temp to allocate the mask
   VERIFY_IS_APPROX(m_permuted, m_original*rp.inverse());
   
   m_permuted = m_original;
-  VERIFY_EVALUATION_COUNT(m_permuted.noalias() = lp * m_permuted, one_if_dynamic); // 1 temp to allocate the mask
+  VERIFY_EVALUATION_COUNT(m_permuted.noalias() = lp * m_permuted, 1); // 1 temp to allocate the mask
   VERIFY_IS_APPROX(m_permuted, lp*m_original);
   
   m_permuted = m_original;
-  VERIFY_EVALUATION_COUNT(m_permuted.noalias() = m_permuted * rp, one_if_dynamic); // 1 temp to allocate the mask
+  VERIFY_EVALUATION_COUNT(m_permuted.noalias() = m_permuted * rp, 1); // 1 temp to allocate the mask
   VERIFY_IS_APPROX(m_permuted, m_original*rp);
 
   if(rows>1 && cols>1)
diff --git a/test/redux.cpp b/test/redux.cpp
index 6ddc59c18..989e1057b 100644
--- a/test/redux.cpp
+++ b/test/redux.cpp
@@ -70,10 +70,10 @@ template<typename MatrixType> void matrixRedux(const MatrixType& m)
   VERIFY_IS_APPROX(m1.block(r0,c0,0,0).prod(),  Scalar(1));
 
   // test nesting complex expression
-  VERIFY_EVALUATION_COUNT( (m1.matrix()*m1.matrix().transpose()).sum(), (MatrixType::SizeAtCompileTime==Dynamic ? 1 : 0) );
+  VERIFY_EVALUATION_COUNT( (m1.matrix()*m1.matrix().transpose()).sum(), (MatrixType::IsVectorAtCompileTime && MatrixType::SizeAtCompileTime!=1 ? 0 : 1) );
   Matrix<Scalar, MatrixType::RowsAtCompileTime, MatrixType::RowsAtCompileTime> m2(rows,rows);
   m2.setRandom();
-  VERIFY_EVALUATION_COUNT( ((m1.matrix()*m1.matrix().transpose())+m2).sum(), (MatrixType::SizeAtCompileTime==Dynamic ? 1 : 0) );
+  VERIFY_EVALUATION_COUNT( ((m1.matrix()*m1.matrix().transpose())+m2).sum(),(MatrixType::IsVectorAtCompileTime && MatrixType::SizeAtCompileTime!=1 ? 0 : 1));
 }
 
 template<typename VectorType> void vectorRedux(const VectorType& w)
@@ -156,8 +156,10 @@ void test_redux()
     CALL_SUBTEST_1( matrixRedux(Array<float, 1, 1>()) );
     CALL_SUBTEST_2( matrixRedux(Matrix2f()) );
     CALL_SUBTEST_2( matrixRedux(Array2f()) );
+    CALL_SUBTEST_2( matrixRedux(Array22f()) );
     CALL_SUBTEST_3( matrixRedux(Matrix4d()) );
     CALL_SUBTEST_3( matrixRedux(Array4d()) );
+    CALL_SUBTEST_3( matrixRedux(Array44d()) );
     CALL_SUBTEST_4( matrixRedux(MatrixXcf(internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
     CALL_SUBTEST_4( matrixRedux(ArrayXXcf(internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
     CALL_SUBTEST_5( matrixRedux(MatrixXd (internal::random<int>(1,maxsize), internal::random<int>(1,maxsize))) );
diff --git a/test/vectorwiseop.cpp b/test/vectorwiseop.cpp
index 739eacaf3..f3ab561ee 100644
--- a/test/vectorwiseop.cpp
+++ b/test/vectorwiseop.cpp
@@ -231,12 +231,12 @@ template<typename MatrixType> void vectorwiseop_matrix(const MatrixType& m)
   Matrix<Scalar,MatrixType::RowsAtCompileTime,MatrixType::RowsAtCompileTime> m1m1 = m1 * m1.transpose();
   VERIFY_IS_APPROX( (m1 * m1.transpose()).colwise().sum(), m1m1.colwise().sum());
   Matrix<Scalar,1,MatrixType::RowsAtCompileTime> tmp(rows);
-  VERIFY_EVALUATION_COUNT( tmp = (m1 * m1.transpose()).colwise().sum(), (MatrixType::RowsAtCompileTime==Dynamic ? 1 : 0));
+  VERIFY_EVALUATION_COUNT( tmp = (m1 * m1.transpose()).colwise().sum(), 1);
 
   m2 = m1.rowwise() - (m1.colwise().sum()/RealScalar(m1.rows())).eval();
   m1 = m1.rowwise() - (m1.colwise().sum()/RealScalar(m1.rows()));
   VERIFY_IS_APPROX( m1, m2 );
-  VERIFY_EVALUATION_COUNT( m2 = (m1.rowwise() - m1.colwise().sum()/RealScalar(m1.rows())), (MatrixType::RowsAtCompileTime==Dynamic && MatrixType::ColsAtCompileTime!=1 ? 1 : 0) );
+  VERIFY_EVALUATION_COUNT( m2 = (m1.rowwise() - m1.colwise().sum()/RealScalar(m1.rows())), (MatrixType::RowsAtCompileTime!=1 ? 1 : 0) );
 }
 
 void test_vectorwiseop()
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorBroadcasting.h b/unsupported/Eigen/CXX11/src/Tensor/TensorBroadcasting.h
index 4cfe300eb..23a74460e 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorBroadcasting.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorBroadcasting.h
@@ -54,7 +54,7 @@ struct is_input_scalar<Sizes<> > {
   static const bool value = true;
 };
 #ifndef EIGEN_EMULATE_CXX11_META_H
-template <typename std::size_t... Indices>
+template <typename std::ptrdiff_t... Indices>
 struct is_input_scalar<Sizes<Indices...> > {
   static const bool value = (Sizes<Indices...>::total_size == 1);
 };
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorStorage.h b/unsupported/Eigen/CXX11/src/Tensor/TensorStorage.h
index f8121d17b..2854a4a17 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorStorage.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorStorage.h
@@ -126,7 +126,7 @@ class TensorStorage<T, DSizes<IndexType, NumIndices_>, Options_>
 	}
 	else 
           m_data = 0;
-        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
+        EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
       }
       m_dimensions = nbDimensions;
     }
diff --git a/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h b/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h
index 9ad2b9cc8..bb6d9e1fe 100644
--- a/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h
+++ b/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h
@@ -61,10 +61,11 @@ struct MatrixExponentialScalingOp
  *  After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
  *  approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
  */
-template <typename MatrixType>
-void matrix_exp_pade3(const MatrixType &A, MatrixType &U, MatrixType &V)
+template <typename MatA, typename MatU, typename MatV>
+void matrix_exp_pade3(const MatA& A, MatU& U, MatV& V)
 {
-  typedef typename NumTraits<typename traits<MatrixType>::Scalar>::Real RealScalar;
+  typedef typename MatA::PlainObject MatrixType;
+  typedef typename NumTraits<typename traits<MatA>::Scalar>::Real RealScalar;
   const RealScalar b[] = {120.L, 60.L, 12.L, 1.L};
   const MatrixType A2 = A * A;
   const MatrixType tmp = b[3] * A2 + b[1] * MatrixType::Identity(A.rows(), A.cols());
@@ -77,9 +78,10 @@ void matrix_exp_pade3(const MatrixType &A, MatrixType &U, MatrixType &V)
  *  After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
  *  approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
  */
-template <typename MatrixType>
-void matrix_exp_pade5(const MatrixType &A, MatrixType &U, MatrixType &V)
+template <typename MatA, typename MatU, typename MatV>
+void matrix_exp_pade5(const MatA& A, MatU& U, MatV& V)
 {
+  typedef typename MatA::PlainObject MatrixType;
   typedef typename NumTraits<typename traits<MatrixType>::Scalar>::Real RealScalar;
   const RealScalar b[] = {30240.L, 15120.L, 3360.L, 420.L, 30.L, 1.L};
   const MatrixType A2 = A * A;
@@ -94,9 +96,10 @@ void matrix_exp_pade5(const MatrixType &A, MatrixType &U, MatrixType &V)
  *  After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
  *  approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
  */
-template <typename MatrixType>
-void matrix_exp_pade7(const MatrixType &A, MatrixType &U, MatrixType &V)
+template <typename MatA, typename MatU, typename MatV>
+void matrix_exp_pade7(const MatA& A, MatU& U, MatV& V)
 {
+  typedef typename MatA::PlainObject MatrixType;
   typedef typename NumTraits<typename traits<MatrixType>::Scalar>::Real RealScalar;
   const RealScalar b[] = {17297280.L, 8648640.L, 1995840.L, 277200.L, 25200.L, 1512.L, 56.L, 1.L};
   const MatrixType A2 = A * A;
@@ -114,9 +117,10 @@ void matrix_exp_pade7(const MatrixType &A, MatrixType &U, MatrixType &V)
  *  After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
  *  approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
  */
-template <typename MatrixType>
-void matrix_exp_pade9(const MatrixType &A, MatrixType &U, MatrixType &V)
+template <typename MatA, typename MatU, typename MatV>
+void matrix_exp_pade9(const MatA& A, MatU& U, MatV& V)
 {
+  typedef typename MatA::PlainObject MatrixType;
   typedef typename NumTraits<typename traits<MatrixType>::Scalar>::Real RealScalar;
   const RealScalar b[] = {17643225600.L, 8821612800.L, 2075673600.L, 302702400.L, 30270240.L,
                           2162160.L, 110880.L, 3960.L, 90.L, 1.L};
@@ -135,9 +139,10 @@ void matrix_exp_pade9(const MatrixType &A, MatrixType &U, MatrixType &V)
  *  After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
  *  approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
  */
-template <typename MatrixType>
-void matrix_exp_pade13(const MatrixType &A, MatrixType &U, MatrixType &V)
+template <typename MatA, typename MatU, typename MatV>
+void matrix_exp_pade13(const MatA& A, MatU& U, MatV& V)
 {
+  typedef typename MatA::PlainObject MatrixType;
   typedef typename NumTraits<typename traits<MatrixType>::Scalar>::Real RealScalar;
   const RealScalar b[] = {64764752532480000.L, 32382376266240000.L, 7771770303897600.L,
                           1187353796428800.L, 129060195264000.L, 10559470521600.L, 670442572800.L,
@@ -162,9 +167,10 @@ void matrix_exp_pade13(const MatrixType &A, MatrixType &U, MatrixType &V)
  *  This function activates only if your long double is double-double or quadruple.
  */
 #if LDBL_MANT_DIG > 64
-template <typename MatrixType>
-void matrix_exp_pade17(const MatrixType &A, MatrixType &U, MatrixType &V)
+template <typename MatA, typename MatU, typename MatV>
+void matrix_exp_pade17(const MatA& A, MatU& U, MatV& V)
 {
+  typedef typename MatA::PlainObject MatrixType;
   typedef typename NumTraits<typename traits<MatrixType>::Scalar>::Real RealScalar;
   const RealScalar b[] = {830034394580628357120000.L, 415017197290314178560000.L,
                           100610229646136770560000.L, 15720348382208870400000.L,
@@ -342,9 +348,10 @@ struct matrix_exp_computeUV<MatrixType, long double>
  * \param arg    argument of matrix exponential (should be plain object)
  * \param result variable in which result will be stored
  */
-template <typename MatrixType, typename ResultType> 
-void matrix_exp_compute(const MatrixType& arg, ResultType &result)
+template <typename ArgType, typename ResultType>
+void matrix_exp_compute(const ArgType& arg, ResultType &result)
 {
+  typedef typename ArgType::PlainObject MatrixType;
 #if LDBL_MANT_DIG > 112 // rarely happens
   typedef typename traits<MatrixType>::Scalar Scalar;
   typedef typename NumTraits<Scalar>::Real RealScalar;
@@ -354,11 +361,11 @@ void matrix_exp_compute(const MatrixType& arg, ResultType &result)
     return;
   }
 #endif
-  typename MatrixType::PlainObject U, V;
+  MatrixType U, V;
   int squarings; 
   matrix_exp_computeUV<MatrixType>::run(arg, U, V, squarings); // Pade approximant is (U+V) / (-U+V)
-  typename MatrixType::PlainObject numer = U + V;
-  typename MatrixType::PlainObject denom = -U + V;
+  MatrixType numer = U + V;
+  MatrixType denom = -U + V;
   result = denom.partialPivLu().solve(numer);
   for (int i=0; i<squarings; i++)
     result *= result;   // undo scaling by repeated squaring