bump to 3.1.4

(transplanted from 4020d4286f
)

.krazy

@@ -0,0 +1,3 @@
+SKIP /disabled/
+SKIP /bench/
+SKIP /build/

CMakeLists.txt

@@ -1,5 +1,6 @@
 project(Eigen)
-cmake_minimum_required(VERSION 2.8.5)
+
+cmake_minimum_required(VERSION 2.6.2)
 
 # guard against in-source builds
 
@@ -54,7 +55,6 @@ endif(EIGEN_HG_CHANGESET)
 
 
 include(CheckCXXCompilerFlag)
-include(GNUInstallDirs)
 
 set(CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake)
 
@@ -105,66 +105,26 @@ if(EIGEN_DEFAULT_TO_ROW_MAJOR)
   add_definitions("-DEIGEN_DEFAULT_TO_ROW_MAJOR")
 endif()
 
+add_definitions("-DEIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS")
+
 set(EIGEN_TEST_MAX_SIZE "320" CACHE STRING "Maximal matrix/vector size, default is 320")
 
-macro(ei_add_cxx_compiler_flag FLAG)
-  string(REGEX REPLACE "-" "" SFLAG ${FLAG})
-  check_cxx_compiler_flag(${FLAG} COMPILER_SUPPORT_${SFLAG})
-  if(COMPILER_SUPPORT_${SFLAG})
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${FLAG}")
-  endif()
-endmacro(ei_add_cxx_compiler_flag)
-
-if(NOT MSVC)
-  # We assume that other compilers are partly compatible with GNUCC
-  
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fexceptions")
+if(CMAKE_COMPILER_IS_GNUCXX)
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wnon-virtual-dtor -Wno-long-long -ansi -Wundef -Wcast-align -Wchar-subscripts -Wall -W -Wpointer-arith -Wwrite-strings -Wformat-security -fexceptions -fno-check-new -fno-common -fstrict-aliasing")
   set(CMAKE_CXX_FLAGS_DEBUG "-g3")
   set(CMAKE_CXX_FLAGS_RELEASE "-g0 -O2")
-  
-  # clang outputs some warnings for unknwon flags that are not caught by check_cxx_compiler_flag
-  # adding -Werror turns such warnings into errors
-  check_cxx_compiler_flag("-Werror" COMPILER_SUPPORT_WERROR)
-  if(COMPILER_SUPPORT_WERROR)
-    set(CMAKE_REQUIRED_FLAGS "-Werror")
+
+  check_cxx_compiler_flag("-Wno-variadic-macros" COMPILER_SUPPORT_WNOVARIADICMACRO)
+  if(COMPILER_SUPPORT_WNOVARIADICMACRO)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-variadic-macros")
   endif()
-  
-  ei_add_cxx_compiler_flag("-pedantic")
-  ei_add_cxx_compiler_flag("-Wall")
-  ei_add_cxx_compiler_flag("-Wextra")
-  #ei_add_cxx_compiler_flag("-Weverything")              # clang
-  
-  ei_add_cxx_compiler_flag("-Wundef")
-  ei_add_cxx_compiler_flag("-Wcast-align")
-  ei_add_cxx_compiler_flag("-Wchar-subscripts")
-  ei_add_cxx_compiler_flag("-Wnon-virtual-dtor")
-  ei_add_cxx_compiler_flag("-Wunused-local-typedefs")
-  ei_add_cxx_compiler_flag("-Wpointer-arith")
-  ei_add_cxx_compiler_flag("-Wwrite-strings")
-  ei_add_cxx_compiler_flag("-Wformat-security")
-  
-  ei_add_cxx_compiler_flag("-Wno-psabi")
-  ei_add_cxx_compiler_flag("-Wno-variadic-macros")
-  ei_add_cxx_compiler_flag("-Wno-long-long")
-  
-  ei_add_cxx_compiler_flag("-fno-check-new")
-  ei_add_cxx_compiler_flag("-fno-common")
-  ei_add_cxx_compiler_flag("-fstrict-aliasing")
-  ei_add_cxx_compiler_flag("-wd981")                    # disable ICC's "operands are evaluated in unspecified order" remark
-  ei_add_cxx_compiler_flag("-wd2304")                   # disbale ICC's "warning #2304: non-explicit constructor with single argument may cause implicit type conversion" produced by -Wnon-virtual-dtor
-  
-  # The -ansi flag must be added last, otherwise it is also used as a linker flag by check_cxx_compiler_flag making it fails
-  # Moreover we should not set both -strict-ansi and -ansi
-  check_cxx_compiler_flag("-strict-ansi" COMPILER_SUPPORT_STRICTANSI)
-  ei_add_cxx_compiler_flag("-Qunused-arguments")        # disable clang warning: argument unused during compilation: '-ansi'
-  
-  if(COMPILER_SUPPORT_STRICTANSI)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -strict-ansi")
-  else()
-    ei_add_cxx_compiler_flag("-ansi")
+
+  check_cxx_compiler_flag("-Wextra" COMPILER_SUPPORT_WEXTRA)
+  if(COMPILER_SUPPORT_WEXTRA)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wextra")
   endif()
-  
-  set(CMAKE_REQUIRED_FLAGS "")
+
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pedantic")
 
   option(EIGEN_TEST_SSE2 "Enable/Disable SSE2 in tests/examples" OFF)
   if(EIGEN_TEST_SSE2)
@@ -217,8 +177,9 @@ if(NOT MSVC)
     endif()
   endif()
 
-else(NOT MSVC)
+endif(CMAKE_COMPILER_IS_GNUCXX)
 
+if(MSVC)
   # C4127 - conditional expression is constant
   # C4714 - marked as __forceinline not inlined (I failed to deactivate it selectively)
   #         We can disable this warning in the unit tests since it is clear that it occurs
@@ -248,7 +209,7 @@ else(NOT MSVC)
     endif(NOT CMAKE_CL_64)
     message(STATUS "Enabling SSE2 in tests/examples")
   endif(EIGEN_TEST_SSE2)
-endif(NOT MSVC)
+endif(MSVC)
 
 option(EIGEN_TEST_NO_EXPLICIT_VECTORIZATION "Disable explicit vectorization in tests/examples" OFF)
 option(EIGEN_TEST_X87 "Force using X87 instructions. Implies no vectorization." OFF)
@@ -288,26 +249,25 @@ option(EIGEN_TEST_C++0x "Enables all C++0x features." OFF)
 
 include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR})
 
-# Backward compatibility support for EIGEN_INCLUDE_INSTALL_DIR
-if(EIGEN_INCLUDE_INSTALL_DIR AND NOT INCLUDE_INSTALL_DIR)
-  set(INCLUDE_INSTALL_DIR ${EIGEN_INCLUDE_INSTALL_DIR}
-      CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed")
+# the user modifiable install path for header files
+set(EIGEN_INCLUDE_INSTALL_DIR ${EIGEN_INCLUDE_INSTALL_DIR} CACHE PATH "The directory where we install the header files (optional)")
+
+# set the internal install path for header files which depends on wether the user modifiable
+# EIGEN_INCLUDE_INSTALL_DIR has been set by the user or not.
+if(EIGEN_INCLUDE_INSTALL_DIR)
+  set(INCLUDE_INSTALL_DIR
+    ${EIGEN_INCLUDE_INSTALL_DIR}
+    CACHE INTERNAL
+    "The directory where we install the header files (internal)"
+  )
 else()
   set(INCLUDE_INSTALL_DIR
-      "${CMAKE_INSTALL_INCLUDEDIR}/eigen3"
-      CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed"
-      )
+    "${CMAKE_INSTALL_PREFIX}/include/eigen3"
+    CACHE INTERNAL
+    "The directory where we install the header files (internal)"
+  )
 endif()
 
-set(CMAKEPACKAGE_INSTALL_DIR
-    "${CMAKE_INSTALL_LIBDIR}/cmake/eigen3"
-    CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen3Config.cmake is installed"
-    )
-set(PKGCONFIG_INSTALL_DIR
-    "${CMAKE_INSTALL_DATADIR}/pkgconfig"
-    CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where eigen3.pc is installed"
-    )
-
 # similar to set_target_properties but append the property instead of overwriting it
 macro(ei_add_target_property target prop value)
 
@@ -325,9 +285,21 @@ install(FILES
   )
 
 if(EIGEN_BUILD_PKGCONFIG)
-    configure_file(eigen3.pc.in eigen3.pc @ONLY)
+    SET(path_separator ":")
+    STRING(REPLACE ${path_separator} ";" pkg_config_libdir_search "$ENV{PKG_CONFIG_LIBDIR}")
+    message(STATUS "searching for 'pkgconfig' directory in PKG_CONFIG_LIBDIR ( $ENV{PKG_CONFIG_LIBDIR} ), ${CMAKE_INSTALL_PREFIX}/share, and ${CMAKE_INSTALL_PREFIX}/lib")
+    FIND_PATH(pkg_config_libdir pkgconfig ${pkg_config_libdir_search} ${CMAKE_INSTALL_PREFIX}/share ${CMAKE_INSTALL_PREFIX}/lib ${pkg_config_libdir_search})
+    if(pkg_config_libdir)
+        SET(pkg_config_install_dir ${pkg_config_libdir})
+        message(STATUS "found ${pkg_config_libdir}/pkgconfig" )
+    else(pkg_config_libdir)
+        SET(pkg_config_install_dir ${CMAKE_INSTALL_PREFIX}/share)
+        message(STATUS "pkgconfig not found; installing in ${pkg_config_install_dir}" )
+    endif(pkg_config_libdir)
+
+    configure_file(eigen3.pc.in eigen3.pc)
     install(FILES ${CMAKE_CURRENT_BINARY_DIR}/eigen3.pc
-        DESTINATION ${PKGCONFIG_INSTALL_DIR}
+        DESTINATION ${pkg_config_install_dir}/pkgconfig
         )
 endif(EIGEN_BUILD_PKGCONFIG)
 
@@ -336,7 +308,6 @@ add_subdirectory(Eigen)
 add_subdirectory(doc EXCLUDE_FROM_ALL)
 
 include(EigenConfigureTesting)
-
 # fixme, not sure this line is still needed:
 enable_testing() # must be called from the root CMakeLists, see man page
 
@@ -371,8 +342,6 @@ if(NOT WIN32)
   add_subdirectory(bench/spbench EXCLUDE_FROM_ALL)
 endif(NOT WIN32)
 
-configure_file(scripts/cdashtesting.cmake.in cdashtesting.cmake @ONLY)
-
 ei_testing_print_summary()
 
 message(STATUS "")
@@ -390,15 +359,12 @@ if(cmake_generator_tolower MATCHES "makefile")
   message(STATUS "--------------+--------------------------------------------------------------")
   message(STATUS "Command       |   Description")
   message(STATUS "--------------+--------------------------------------------------------------")
-  message(STATUS "make install  | Install Eigen. Headers will be installed to:")
-  message(STATUS "              |     <CMAKE_INSTALL_PREFIX>/<INCLUDE_INSTALL_DIR>")
-  message(STATUS "              |   Using the following values:")
-  message(STATUS "              |     CMAKE_INSTALL_PREFIX: ${CMAKE_INSTALL_PREFIX}")
-  message(STATUS "              |     INCLUDE_INSTALL_DIR:  ${INCLUDE_INSTALL_DIR}")
-  message(STATUS "              |   Change the install location of Eigen headers using:")
-  message(STATUS "              |     cmake . -DCMAKE_INSTALL_PREFIX=yourprefix")
-  message(STATUS "              |   Or:")
-  message(STATUS "              |     cmake . -DINCLUDE_INSTALL_DIR=yourdir")
+  message(STATUS "make install  | Install to ${CMAKE_INSTALL_PREFIX}. To change that:")
+  message(STATUS "              |     cmake . -DCMAKE_INSTALL_PREFIX=yourpath")
+  message(STATUS "              |   Eigen headers will then be installed to:")
+  message(STATUS "              |     ${INCLUDE_INSTALL_DIR}")
+  message(STATUS "              |   To install Eigen headers to a separate location, do:")
+  message(STATUS "              |     cmake . -DEIGEN_INCLUDE_INSTALL_DIR=yourpath")
   message(STATUS "make doc      | Generate the API documentation, requires Doxygen & LaTeX")
   message(STATUS "make check    | Build and run the unit-tests. Read this page:")
   message(STATUS "              |   http://eigen.tuxfamily.org/index.php?title=Tests")

CTestConfig.cmake

@@ -4,10 +4,10 @@
 ## # The following are required to uses Dart and the Cdash dashboard
 ##   ENABLE_TESTING()
 ##   INCLUDE(CTest)
-set(CTEST_PROJECT_NAME "Eigen3.2")
+set(CTEST_PROJECT_NAME "Eigen3.1")
 set(CTEST_NIGHTLY_START_TIME "00:00:00 UTC")
 
 set(CTEST_DROP_METHOD "http")
 set(CTEST_DROP_SITE "manao.inria.fr")
-set(CTEST_DROP_LOCATION "/CDash/submit.php?project=Eigen3.2")
-set(CTEST_DROP_SITE_CDASH TRUE)
+set(CTEST_DROP_LOCATION "/CDash/submit.php?project=Eigen3.1")
+set(CTEST_DROP_SITE_CDASH TRUE)

CTestCustom.cmake.in

@@ -1,3 +1,4 @@
 
-set(CTEST_CUSTOM_MAXIMUM_NUMBER_OF_WARNINGS "2000")
-set(CTEST_CUSTOM_MAXIMUM_NUMBER_OF_ERRORS   "2000")
+## A tribute to Dynamic!
+set(CTEST_CUSTOM_MAXIMUM_NUMBER_OF_WARNINGS "33331")
+set(CTEST_CUSTOM_MAXIMUM_NUMBER_OF_ERRORS "33331")

Eigen/CholmodSupport

@@ -12,7 +12,7 @@ extern "C" {
 /** \ingroup Support_modules
   * \defgroup CholmodSupport_Module CholmodSupport module
   *
-  * This module provides an interface to the Cholmod library which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
+  * This module provides an interface to the Cholmod library which is part of the <a href="http://www.cise.ufl.edu/research/sparse/SuiteSparse/">suitesparse</a> package.
   * It provides the two following main factorization classes:
   * - class CholmodSupernodalLLT: a supernodal LLT Cholesky factorization.
   * - class CholmodDecomposiiton: a general L(D)LT Cholesky factorization with automatic or explicit runtime selection of the underlying factorization method (supernodal or simplicial).

Eigen/Core

@@ -19,12 +19,6 @@
 // defined e.g. EIGEN_DONT_ALIGN) so it needs to be done before we do anything with vectorization.
 #include "src/Core/util/Macros.h"
 
-// Disable the ipa-cp-clone optimization flag with MinGW 6.x or newer (enabled by default with -O3)
-// See http://eigen.tuxfamily.org/bz/show_bug.cgi?id=556 for details.
-#if defined(__MINGW32__) && EIGEN_GNUC_AT_LEAST(4,6)
-  #pragma GCC optimize ("-fno-ipa-cp-clone")
-#endif
-
 #include <complex>
 
 // this include file manages BLAS and MKL related macros
@@ -93,25 +87,19 @@
     // so, to avoid compile errors when windows.h is included after Eigen/Core, ensure intrinsics are extern "C" here too.
     // notice that since these are C headers, the extern "C" is theoretically needed anyways.
     extern "C" {
-      // In theory we should only include immintrin.h and not the other *mmintrin.h header files directly.
-      // Doing so triggers some issues with ICC. However old gcc versions seems to not have this file, thus:
-      #if defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1110
-        #include <immintrin.h>
-      #else
-        #include <emmintrin.h>
-        #include <xmmintrin.h>
-        #ifdef  EIGEN_VECTORIZE_SSE3
-        #include <pmmintrin.h>
-        #endif
-        #ifdef EIGEN_VECTORIZE_SSSE3
-        #include <tmmintrin.h>
-        #endif
-        #ifdef EIGEN_VECTORIZE_SSE4_1
-        #include <smmintrin.h>
-        #endif
-        #ifdef EIGEN_VECTORIZE_SSE4_2
-        #include <nmmintrin.h>
-        #endif
+      #include <emmintrin.h>
+      #include <xmmintrin.h>
+      #ifdef  EIGEN_VECTORIZE_SSE3
+      #include <pmmintrin.h>
+      #endif
+      #ifdef EIGEN_VECTORIZE_SSSE3
+      #include <tmmintrin.h>
+      #endif
+      #ifdef EIGEN_VECTORIZE_SSE4_1
+      #include <smmintrin.h>
+      #endif
+      #ifdef EIGEN_VECTORIZE_SSE4_2
+      #include <nmmintrin.h>
       #endif
     } // end extern "C"
   #elif defined __ALTIVEC__
@@ -123,7 +111,7 @@
     #undef bool
     #undef vector
     #undef pixel
-  #elif defined  __ARM_NEON
+  #elif defined  __ARM_NEON__
     #define EIGEN_VECTORIZE
     #define EIGEN_VECTORIZE_NEON
     #include <arm_neon.h>
@@ -165,7 +153,7 @@
 #endif
 
 // required for __cpuid, needs to be included after cmath
-#if defined(_MSC_VER) && (defined(_M_IX86)||defined(_M_X64)) && (!defined(_WIN32_WCE))
+#if defined(_MSC_VER) && (defined(_M_IX86)||defined(_M_X64))
   #include <intrin.h>
 #endif
 
@@ -248,11 +236,15 @@ using std::ptrdiff_t;
   * \endcode
   */
 
+/** \defgroup Support_modules Support modules [category]
+  * Category of modules which add support for external libraries.
+  */
+
 #include "src/Core/util/Constants.h"
 #include "src/Core/util/ForwardDeclarations.h"
 #include "src/Core/util/Meta.h"
-#include "src/Core/util/StaticAssert.h"
 #include "src/Core/util/XprHelper.h"
+#include "src/Core/util/StaticAssert.h"
 #include "src/Core/util/Memory.h"
 
 #include "src/Core/NumTraits.h"
@@ -305,7 +297,6 @@ using std::ptrdiff_t;
 #include "src/Core/Map.h"
 #include "src/Core/Block.h"
 #include "src/Core/VectorBlock.h"
-#include "src/Core/Ref.h"
 #include "src/Core/Transpose.h"
 #include "src/Core/DiagonalMatrix.h"
 #include "src/Core/Diagonal.h"
@@ -339,7 +330,6 @@ using std::ptrdiff_t;
 #include "src/Core/products/TriangularSolverMatrix.h"
 #include "src/Core/products/TriangularSolverVector.h"
 #include "src/Core/BandMatrix.h"
-#include "src/Core/CoreIterators.h"
 
 #include "src/Core/BooleanRedux.h"
 #include "src/Core/Select.h"

Eigen/Eigen2Support

@@ -14,25 +14,12 @@
 #error Eigen2 support must be enabled by defining EIGEN2_SUPPORT before including any Eigen header
 #endif
 
-#ifndef EIGEN_NO_EIGEN2_DEPRECATED_WARNING
-
-#if defined(__GNUC__) || defined(__INTEL_COMPILER) || defined(__clang__)
-#warning "Eigen2 support is deprecated in Eigen 3.2.x and it will be removed in Eigen 3.3. (Define EIGEN_NO_EIGEN2_DEPRECATED_WARNING to disable this warning)"
-#else
-#pragma message ("Eigen2 support is deprecated in Eigen 3.2.x and it will be removed in Eigen 3.3. (Define EIGEN_NO_EIGEN2_DEPRECATED_WARNING to disable this warning)")
-#endif
-
-#endif // EIGEN_NO_EIGEN2_DEPRECATED_WARNING
-
 #include "src/Core/util/DisableStupidWarnings.h"
 
 /** \ingroup Support_modules
   * \defgroup Eigen2Support_Module Eigen2 support module
-  *
-  * \warning Eigen2 support is deprecated in Eigen 3.2.x and it will be removed in Eigen 3.3.
-  *
   * This module provides a couple of deprecated functions improving the compatibility with Eigen2.
-  * 
+  *
   * To use it, define EIGEN2_SUPPORT before including any Eigen header
   * \code
   * #define EIGEN2_SUPPORT

Eigen/Eigenvalues

@@ -33,8 +33,6 @@
 #include "src/Eigenvalues/HessenbergDecomposition.h"
 #include "src/Eigenvalues/ComplexSchur.h"
 #include "src/Eigenvalues/ComplexEigenSolver.h"
-#include "src/Eigenvalues/RealQZ.h"
-#include "src/Eigenvalues/GeneralizedEigenSolver.h"
 #include "src/Eigenvalues/MatrixBaseEigenvalues.h"
 #ifdef EIGEN_USE_LAPACKE
 #include "src/Eigenvalues/RealSchur_MKL.h"

Eigen/IterativeLinearSolvers

@@ -6,7 +6,7 @@
 
 #include "src/Core/util/DisableStupidWarnings.h"
 
-/** 
+/** \ingroup Sparse_modules
   * \defgroup IterativeLinearSolvers_Module IterativeLinearSolvers module
   *
   * This module currently provides iterative methods to solve problems of the form \c A \c x = \c b, where \c A is a squared matrix, usually very large and sparse.

Eigen/MetisSupport

@@ -1,28 +0,0 @@
-#ifndef EIGEN_METISSUPPORT_MODULE_H
-#define EIGEN_METISSUPPORT_MODULE_H
-
-#include "SparseCore"
-
-#include "src/Core/util/DisableStupidWarnings.h"
-
-extern "C" {
-#include <metis.h>
-}
-
-
-/** \ingroup Support_modules
-  * \defgroup MetisSupport_Module MetisSupport module
-  *
-  * \code
-  * #include <Eigen/MetisSupport>
-  * \endcode
-  * This module defines an interface to the METIS reordering package (http://glaros.dtc.umn.edu/gkhome/views/metis). 
-  * It can be used just as any other built-in method as explained in \link OrderingMethods_Module here. \endlink
-  */
-
-
-#include "src/MetisSupport/MetisSupport.h"
-
-#include "src/Core/util/ReenableStupidWarnings.h"
-
-#endif // EIGEN_METISSUPPORT_MODULE_H

Eigen/OrderingMethods

@@ -5,62 +5,19 @@
 
 #include "src/Core/util/DisableStupidWarnings.h"
 
-/** 
+/** \ingroup Sparse_modules
   * \defgroup OrderingMethods_Module OrderingMethods module
   *
-  * This module is currently for internal use only
-  * 
-  * It defines various built-in and external ordering methods for sparse matrices. 
-  * They are typically used to reduce the number of elements during 
-  * the sparse matrix decomposition (LLT, LU, QR).
-  * Precisely, in a preprocessing step, a permutation matrix P is computed using 
-  * those ordering methods and applied to the columns of the matrix. 
-  * Using for instance the sparse Cholesky decomposition, it is expected that 
-  * the nonzeros elements in LLT(A*P) will be much smaller than that in LLT(A).
-  * 
-  * 
-  * Usage : 
+  * This module is currently for internal use only.
+  *
+  *
   * \code
   * #include <Eigen/OrderingMethods>
   * \endcode
-  * 
-  * A simple usage is as a template parameter in the sparse decomposition classes : 
-  * 
-  * \code 
-  * SparseLU<MatrixType, COLAMDOrdering<int> > solver;
-  * \endcode 
-  * 
-  * \code 
-  * SparseQR<MatrixType, COLAMDOrdering<int> > solver;
-  * \endcode
-  * 
-  * It is possible as well to call directly a particular ordering method for your own purpose, 
-  * \code 
-  * AMDOrdering<int> ordering;
-  * PermutationMatrix<Dynamic, Dynamic, int> perm;
-  * SparseMatrix<double> A; 
-  * //Fill the matrix ...
-  * 
-  * ordering(A, perm); // Call AMD
-  * \endcode
-  * 
-  * \note Some of these methods (like AMD or METIS), need the sparsity pattern 
-  * of the input matrix to be symmetric. When the matrix is structurally unsymmetric, 
-  * Eigen computes internally the pattern of \f$A^T*A\f$ before calling the method.
-  * If your matrix is already symmetric (at leat in structure), you can avoid that
-  * by calling the method with a SelfAdjointView type.
-  * 
-  * \code
-  *  // Call the ordering on the pattern of the lower triangular matrix A
-  * ordering(A.selfadjointView<Lower>(), perm);
-  * \endcode
   */
 
-#ifndef EIGEN_MPL2_ONLY
 #include "src/OrderingMethods/Amd.h"
-#endif
 
-#include "src/OrderingMethods/Ordering.h"
 #include "src/Core/util/ReenableStupidWarnings.h"
 
 #endif // EIGEN_ORDERINGMETHODS_MODULE_H

Eigen/SPQRSupport

@@ -1,29 +0,0 @@
-#ifndef EIGEN_SPQRSUPPORT_MODULE_H
-#define EIGEN_SPQRSUPPORT_MODULE_H
-
-#include "SparseCore"
-
-#include "src/Core/util/DisableStupidWarnings.h"
-
-#include "SuiteSparseQR.hpp"
-
-/** \ingroup Support_modules
-  * \defgroup SPQRSupport_Module SuiteSparseQR module
-  * 
-  * This module provides an interface to the SPQR library, which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
-  *
-  * \code
-  * #include <Eigen/SPQRSupport>
-  * \endcode
-  *
-  * In order to use this module, the SPQR headers must be accessible from the include paths, and your binary must be linked to the SPQR library and its dependencies (Cholmod, AMD, COLAMD,...).
-  * For a cmake based project, you can use our FindSPQR.cmake and FindCholmod.Cmake modules
-  *
-  */
-
-#include "src/misc/Solve.h"
-#include "src/misc/SparseSolve.h"
-#include "src/CholmodSupport/CholmodSupport.h"
-#include "src/SPQRSupport/SuiteSparseQRSupport.h"
-
-#endif

Eigen/Sparse

@@ -1,15 +1,13 @@
 #ifndef EIGEN_SPARSE_MODULE_H
 #define EIGEN_SPARSE_MODULE_H
 
-/** \defgroup Sparse_Module Sparse meta-module
+/** \defgroup Sparse_modules Sparse modules
   *
   * Meta-module including all related modules:
-  * - \ref SparseCore_Module
-  * - \ref OrderingMethods_Module
-  * - \ref SparseCholesky_Module
-  * - \ref SparseLU_Module
-  * - \ref SparseQR_Module
-  * - \ref IterativeLinearSolvers_Module
+  * - SparseCore
+  * - OrderingMethods
+  * - SparseCholesky
+  * - IterativeLinearSolvers
   *
   * \code
   * #include <Eigen/Sparse>
@@ -19,8 +17,6 @@
 #include "SparseCore"
 #include "OrderingMethods"
 #include "SparseCholesky"
-#include "SparseLU"
-#include "SparseQR"
 #include "IterativeLinearSolvers"
 
 #endif // EIGEN_SPARSE_MODULE_H

Eigen/SparseCholesky

@@ -1,12 +1,3 @@
-// This file is part of Eigen, a lightweight C++ template library
-// for linear algebra.
-//
-// Copyright (C) 2008-2013 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/.
-
 #ifndef EIGEN_SPARSECHOLESKY_MODULE_H
 #define EIGEN_SPARSECHOLESKY_MODULE_H
 
@@ -15,7 +6,7 @@
 
 #include "src/Core/util/DisableStupidWarnings.h"
 
-/** 
+/** \ingroup Sparse_modules
   * \defgroup SparseCholesky_Module SparseCholesky module
   *
   * This module currently provides two variants of the direct sparse Cholesky decomposition for selfadjoint (hermitian) matrices.
@@ -30,17 +21,10 @@
   * \endcode
   */
 
-#ifdef EIGEN_MPL2_ONLY
-#error The SparseCholesky module has nothing to offer in MPL2 only mode
-#endif
-
 #include "src/misc/Solve.h"
 #include "src/misc/SparseSolve.h"
-#include "src/SparseCholesky/SimplicialCholesky.h"
 
-#ifndef EIGEN_MPL2_ONLY
-#include "src/SparseCholesky/SimplicialCholesky_impl.h"
-#endif
+#include "src/SparseCholesky/SimplicialCholesky.h"
 
 #include "src/Core/util/ReenableStupidWarnings.h"
 

Eigen/SparseCore

@@ -11,10 +11,10 @@
 #include <cstring>
 #include <algorithm>
 
-/** 
+/** \ingroup Sparse_modules
   * \defgroup SparseCore_Module SparseCore module
   *
-  * This module provides a sparse matrix representation, and basic associated matrix manipulations
+  * This module provides a sparse matrix representation, and basic associatd matrix manipulations
   * and operations.
   *
   * See the \ref TutorialSparse "Sparse tutorial"
@@ -40,12 +40,14 @@ struct Sparse {};
 #include "src/SparseCore/SparseMatrix.h"
 #include "src/SparseCore/MappedSparseMatrix.h"
 #include "src/SparseCore/SparseVector.h"
+#include "src/SparseCore/CoreIterators.h"
 #include "src/SparseCore/SparseBlock.h"
 #include "src/SparseCore/SparseTranspose.h"
 #include "src/SparseCore/SparseCwiseUnaryOp.h"
 #include "src/SparseCore/SparseCwiseBinaryOp.h"
 #include "src/SparseCore/SparseDot.h"
 #include "src/SparseCore/SparsePermutation.h"
+#include "src/SparseCore/SparseAssign.h"
 #include "src/SparseCore/SparseRedux.h"
 #include "src/SparseCore/SparseFuzzy.h"
 #include "src/SparseCore/ConservativeSparseSparseProduct.h"

Eigen/SparseLU

@@ -1,49 +0,0 @@
-// This file is part of Eigen, a lightweight C++ template library
-// for linear algebra.
-//
-// Copyright (C) 2012 Désiré Nuentsa-Wakam <desire.nuentsa_wakam@inria.fr>
-// Copyright (C) 2012 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/.
-
-#ifndef EIGEN_SPARSELU_MODULE_H
-#define EIGEN_SPARSELU_MODULE_H
-
-#include "SparseCore"
-
-/** 
-  * \defgroup SparseLU_Module SparseLU module
-  * This module defines a supernodal factorization of general sparse matrices.
-  * The code is fully optimized for supernode-panel updates with specialized kernels.
-  * Please, see the documentation of the SparseLU class for more details.
-  */
-
-#include "src/misc/Solve.h"
-#include "src/misc/SparseSolve.h"
-
-// Ordering interface
-#include "OrderingMethods"
-
-#include "src/SparseLU/SparseLU_gemm_kernel.h"
-
-#include "src/SparseLU/SparseLU_Structs.h"
-#include "src/SparseLU/SparseLU_SupernodalMatrix.h"
-#include "src/SparseLU/SparseLUImpl.h"
-#include "src/SparseCore/SparseColEtree.h"
-#include "src/SparseLU/SparseLU_Memory.h"
-#include "src/SparseLU/SparseLU_heap_relax_snode.h"
-#include "src/SparseLU/SparseLU_relax_snode.h"
-#include "src/SparseLU/SparseLU_pivotL.h"
-#include "src/SparseLU/SparseLU_panel_dfs.h"
-#include "src/SparseLU/SparseLU_kernel_bmod.h"
-#include "src/SparseLU/SparseLU_panel_bmod.h"
-#include "src/SparseLU/SparseLU_column_dfs.h"
-#include "src/SparseLU/SparseLU_column_bmod.h"
-#include "src/SparseLU/SparseLU_copy_to_ucol.h"
-#include "src/SparseLU/SparseLU_pruneL.h"
-#include "src/SparseLU/SparseLU_Utils.h"
-#include "src/SparseLU/SparseLU.h"
-
-#endif // EIGEN_SPARSELU_MODULE_H

Eigen/SparseQR

@@ -1,33 +0,0 @@
-#ifndef EIGEN_SPARSEQR_MODULE_H
-#define EIGEN_SPARSEQR_MODULE_H
-
-#include "SparseCore"
-#include "OrderingMethods"
-#include "src/Core/util/DisableStupidWarnings.h"
-
-/** \defgroup SparseQR_Module SparseQR module
-  * \brief Provides QR decomposition for sparse matrices
-  * 
-  * This module provides a simplicial version of the left-looking Sparse QR decomposition. 
-  * The columns of the input matrix should be reordered to limit the fill-in during the 
-  * decomposition. Built-in methods (COLAMD, AMD) or external  methods (METIS) can be used to this end.
-  * See the \link OrderingMethods_Module OrderingMethods\endlink module for the list 
-  * of built-in and external ordering methods.
-  * 
-  * \code
-  * #include <Eigen/SparseQR>
-  * \endcode
-  * 
-  * 
-  */
-
-#include "src/misc/Solve.h"
-#include "src/misc/SparseSolve.h"
-
-#include "OrderingMethods"
-#include "src/SparseCore/SparseColEtree.h"
-#include "src/SparseQR/SparseQR.h"
-
-#include "src/Core/util/ReenableStupidWarnings.h"
-
-#endif

Eigen/UmfPackSupport

@@ -12,7 +12,7 @@ extern "C" {
 /** \ingroup Support_modules
   * \defgroup UmfPackSupport_Module UmfPackSupport module
   *
-  * This module provides an interface to the UmfPack library which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
+  * This module provides an interface to the UmfPack library which is part of the <a href="http://www.cise.ufl.edu/research/sparse/SuiteSparse/">suitesparse</a> package.
   * It provides the following factorization class:
   * - class UmfPackLU: a multifrontal sequential LU factorization.
   *

Eigen/src/Cholesky/LDLT.h

@@ -16,10 +16,7 @@
 namespace Eigen { 
 
 namespace internal {
-  template<typename MatrixType, int UpLo> struct LDLT_Traits;
-
-  // PositiveSemiDef means positive semi-definite and non-zero; same for NegativeSemiDef
-  enum SignMatrix { PositiveSemiDef, NegativeSemiDef, ZeroSign, Indefinite };
+template<typename MatrixType, int UpLo> struct LDLT_Traits;
 }
 
 /** \ingroup Cholesky_Module
@@ -72,12 +69,7 @@ template<typename _MatrixType, int _UpLo> class LDLT
       * The default constructor is useful in cases in which the user intends to
       * perform decompositions via LDLT::compute(const MatrixType&).
       */
-    LDLT() 
-      : m_matrix(), 
-        m_transpositions(), 
-        m_sign(internal::ZeroSign),
-        m_isInitialized(false) 
-    {}
+    LDLT() : m_matrix(), m_transpositions(), m_isInitialized(false) {}
 
     /** \brief Default Constructor with memory preallocation
       *
@@ -89,7 +81,6 @@ template<typename _MatrixType, int _UpLo> class LDLT
       : m_matrix(size, size),
         m_transpositions(size),
         m_temporary(size),
-        m_sign(internal::ZeroSign),
         m_isInitialized(false)
     {}
 
@@ -102,7 +93,6 @@ template<typename _MatrixType, int _UpLo> class LDLT
       : m_matrix(matrix.rows(), matrix.cols()),
         m_transpositions(matrix.rows()),
         m_temporary(matrix.rows()),
-        m_sign(internal::ZeroSign),
         m_isInitialized(false)
     {
       compute(matrix);
@@ -149,7 +139,7 @@ template<typename _MatrixType, int _UpLo> class LDLT
     inline bool isPositive() const
     {
       eigen_assert(m_isInitialized && "LDLT is not initialized.");
-      return m_sign == internal::PositiveSemiDef || m_sign == internal::ZeroSign;
+      return m_sign == 1;
     }
     
     #ifdef EIGEN2_SUPPORT
@@ -163,7 +153,7 @@ template<typename _MatrixType, int _UpLo> class LDLT
     inline bool isNegative(void) const
     {
       eigen_assert(m_isInitialized && "LDLT is not initialized.");
-      return m_sign == internal::NegativeSemiDef || m_sign == internal::ZeroSign;
+      return m_sign == -1;
     }
 
     /** \returns a solution x of \f$ A x = b \f$ using the current decomposition of A.
@@ -206,7 +196,7 @@ template<typename _MatrixType, int _UpLo> class LDLT
     LDLT& compute(const MatrixType& matrix);
 
     template <typename Derived>
-    LDLT& rankUpdate(const MatrixBase<Derived>& w, const RealScalar& alpha=1);
+    LDLT& rankUpdate(const MatrixBase<Derived>& w,RealScalar alpha=1);
 
     /** \returns the internal LDLT decomposition matrix
       *
@@ -235,11 +225,6 @@ template<typename _MatrixType, int _UpLo> class LDLT
     }
 
   protected:
-    
-    static void check_template_parameters()
-    {
-      EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
-    }
 
     /** \internal
       * Used to compute and store the Cholesky decomposition A = L D L^* = U^* D U.
@@ -250,7 +235,7 @@ template<typename _MatrixType, int _UpLo> class LDLT
     MatrixType m_matrix;
     TranspositionType m_transpositions;
     TmpMatrixType m_temporary;
-    internal::SignMatrix m_sign;
+    int m_sign;
     bool m_isInitialized;
 };
 
@@ -261,9 +246,8 @@ template<int UpLo> struct ldlt_inplace;
 template<> struct ldlt_inplace<Lower>
 {
   template<typename MatrixType, typename TranspositionType, typename Workspace>
-  static bool unblocked(MatrixType& mat, TranspositionType& transpositions, Workspace& temp, SignMatrix& sign)
+  static bool unblocked(MatrixType& mat, TranspositionType& transpositions, Workspace& temp, int* sign=0)
   {
-    using std::abs;
     typedef typename MatrixType::Scalar Scalar;
     typedef typename MatrixType::RealScalar RealScalar;
     typedef typename MatrixType::Index Index;
@@ -273,19 +257,36 @@ template<> struct ldlt_inplace<Lower>
     if (size <= 1)
     {
       transpositions.setIdentity();
-      if (numext::real(mat.coeff(0,0)) > 0) sign = PositiveSemiDef;
-      else if (numext::real(mat.coeff(0,0)) < 0) sign = NegativeSemiDef;
-      else sign = ZeroSign;
+      if(sign)
+        *sign = real(mat.coeff(0,0))>0 ? 1:-1;
       return true;
     }
 
+    RealScalar cutoff(0), biggest_in_corner;
+
     for (Index k = 0; k < size; ++k)
     {
       // Find largest diagonal element
       Index index_of_biggest_in_corner;
-      mat.diagonal().tail(size-k).cwiseAbs().maxCoeff(&index_of_biggest_in_corner);
+      biggest_in_corner = mat.diagonal().tail(size-k).cwiseAbs().maxCoeff(&index_of_biggest_in_corner);
       index_of_biggest_in_corner += k;
 
+      if(k == 0)
+      {
+        // The biggest overall is the point of reference to which further diagonals
+        // are compared; if any diagonal is negligible compared
+        // to the largest overall, the algorithm bails.
+        cutoff = abs(NumTraits<Scalar>::epsilon() * biggest_in_corner);
+      }
+
+      // Finish early if the matrix is not full rank.
+      if(biggest_in_corner < cutoff)
+      {
+        for(Index i = k; i < size; i++) transpositions.coeffRef(i) = i;
+        if(sign) *sign = 0;
+        break;
+      }
+
       transpositions.coeffRef(k) = index_of_biggest_in_corner;
       if(k != index_of_biggest_in_corner)
       {
@@ -298,11 +299,11 @@ template<> struct ldlt_inplace<Lower>
         for(int i=k+1;i<index_of_biggest_in_corner;++i)
         {
           Scalar tmp = mat.coeffRef(i,k);
-          mat.coeffRef(i,k) = numext::conj(mat.coeffRef(index_of_biggest_in_corner,i));
-          mat.coeffRef(index_of_biggest_in_corner,i) = numext::conj(tmp);
+          mat.coeffRef(i,k) = conj(mat.coeffRef(index_of_biggest_in_corner,i));
+          mat.coeffRef(index_of_biggest_in_corner,i) = conj(tmp);
         }
         if(NumTraits<Scalar>::IsComplex)
-          mat.coeffRef(index_of_biggest_in_corner,k) = numext::conj(mat.coeff(index_of_biggest_in_corner,k));
+          mat.coeffRef(index_of_biggest_in_corner,k) = conj(mat.coeff(index_of_biggest_in_corner,k));
       }
 
       // partition the matrix:
@@ -316,27 +317,22 @@ template<> struct ldlt_inplace<Lower>
 
       if(k>0)
       {
-        temp.head(k) = mat.diagonal().real().head(k).asDiagonal() * A10.adjoint();
+        temp.head(k) = mat.diagonal().head(k).asDiagonal() * A10.adjoint();
         mat.coeffRef(k,k) -= (A10 * temp.head(k)).value();
         if(rs>0)
           A21.noalias() -= A20 * temp.head(k);
       }
+      if((rs>0) && (abs(mat.coeffRef(k,k)) > cutoff))
+        A21 /= mat.coeffRef(k,k);
       
-      // In some previous versions of Eigen (e.g., 3.2.1), the scaling was omitted if the pivot
-      // was smaller than the cutoff value. However, soince LDLT is not rank-revealing
-      // we should only make sure we do not introduce INF or NaN values.
-      // LAPACK also uses 0 as the cutoff value.
-      RealScalar realAkk = numext::real(mat.coeffRef(k,k));
-      if((rs>0) && (abs(realAkk) > RealScalar(0)))
-        A21 /= realAkk;
-
-      if (sign == PositiveSemiDef) {
-        if (realAkk < 0) sign = Indefinite;
-      } else if (sign == NegativeSemiDef) {
-        if (realAkk > 0) sign = Indefinite;
-      } else if (sign == ZeroSign) {
-        if (realAkk > 0) sign = PositiveSemiDef;
-        else if (realAkk < 0) sign = NegativeSemiDef;
+      if(sign)
+      {
+        // LDLT is not guaranteed to work for indefinite matrices, but let's try to get the sign right
+        int newSign = real(mat.diagonal().coeff(index_of_biggest_in_corner)) > 0;
+        if(k == 0)
+          *sign = newSign;
+        else if(*sign != newSign)
+          *sign = 0;
       }
     }
 
@@ -351,9 +347,9 @@ template<> struct ldlt_inplace<Lower>
   // Here only rank-1 updates are implemented, to reduce the
   // requirement for intermediate storage and improve accuracy
   template<typename MatrixType, typename WDerived>
-  static bool updateInPlace(MatrixType& mat, MatrixBase<WDerived>& w, const typename MatrixType::RealScalar& sigma=1)
+  static bool updateInPlace(MatrixType& mat, MatrixBase<WDerived>& w, typename MatrixType::RealScalar sigma=1)
   {
-    using numext::isfinite;
+    using internal::isfinite;
     typedef typename MatrixType::Scalar Scalar;
     typedef typename MatrixType::RealScalar RealScalar;
     typedef typename MatrixType::Index Index;
@@ -371,9 +367,9 @@ template<> struct ldlt_inplace<Lower>
         break;
 
       // Update the diagonal terms
-      RealScalar dj = numext::real(mat.coeff(j,j));
+      RealScalar dj = real(mat.coeff(j,j));
       Scalar wj = w.coeff(j);
-      RealScalar swj2 = sigma*numext::abs2(wj);
+      RealScalar swj2 = sigma*abs2(wj);
       RealScalar gamma = dj*alpha + swj2;
 
       mat.coeffRef(j,j) += swj2/alpha;
@@ -384,13 +380,13 @@ template<> struct ldlt_inplace<Lower>
       Index rs = size-j-1;
       w.tail(rs) -= wj * mat.col(j).tail(rs);
       if(gamma != 0)
-        mat.col(j).tail(rs) += (sigma*numext::conj(wj)/gamma)*w.tail(rs);
+        mat.col(j).tail(rs) += (sigma*conj(wj)/gamma)*w.tail(rs);
     }
     return true;
   }
 
   template<typename MatrixType, typename TranspositionType, typename Workspace, typename WType>
-  static bool update(MatrixType& mat, const TranspositionType& transpositions, Workspace& tmp, const WType& w, const typename MatrixType::RealScalar& sigma=1)
+  static bool update(MatrixType& mat, const TranspositionType& transpositions, Workspace& tmp, const WType& w, typename MatrixType::RealScalar sigma=1)
   {
     // Apply the permutation to the input w
     tmp = transpositions * w;
@@ -402,14 +398,14 @@ template<> struct ldlt_inplace<Lower>
 template<> struct ldlt_inplace<Upper>
 {
   template<typename MatrixType, typename TranspositionType, typename Workspace>
-  static EIGEN_STRONG_INLINE bool unblocked(MatrixType& mat, TranspositionType& transpositions, Workspace& temp, SignMatrix& sign)
+  static EIGEN_STRONG_INLINE bool unblocked(MatrixType& mat, TranspositionType& transpositions, Workspace& temp, int* sign=0)
   {
     Transpose<MatrixType> matt(mat);
     return ldlt_inplace<Lower>::unblocked(matt, transpositions, temp, sign);
   }
 
   template<typename MatrixType, typename TranspositionType, typename Workspace, typename WType>
-  static EIGEN_STRONG_INLINE bool update(MatrixType& mat, TranspositionType& transpositions, Workspace& tmp, WType& w, const typename MatrixType::RealScalar& sigma=1)
+  static EIGEN_STRONG_INLINE bool update(MatrixType& mat, TranspositionType& transpositions, Workspace& tmp, WType& w, typename MatrixType::RealScalar sigma=1)
   {
     Transpose<MatrixType> matt(mat);
     return ldlt_inplace<Lower>::update(matt, transpositions, tmp, w.conjugate(), sigma);
@@ -439,8 +435,6 @@ template<typename MatrixType> struct LDLT_Traits<MatrixType,Upper>
 template<typename MatrixType, int _UpLo>
 LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::compute(const MatrixType& a)
 {
-  check_template_parameters();
-  
   eigen_assert(a.rows()==a.cols());
   const Index size = a.rows();
 
@@ -449,9 +443,8 @@ LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::compute(const MatrixType& a)
   m_transpositions.resize(size);
   m_isInitialized = false;
   m_temporary.resize(size);
-  m_sign = internal::ZeroSign;
 
-  internal::ldlt_inplace<UpLo>::unblocked(m_matrix, m_transpositions, m_temporary, m_sign);
+  internal::ldlt_inplace<UpLo>::unblocked(m_matrix, m_transpositions, m_temporary, &m_sign);
 
   m_isInitialized = true;
   return *this;
@@ -464,7 +457,7 @@ LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::compute(const MatrixType& a)
   */
 template<typename MatrixType, int _UpLo>
 template<typename Derived>
-LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::rankUpdate(const MatrixBase<Derived>& w, const typename LDLT<MatrixType,_UpLo>::RealScalar& sigma)
+LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::rankUpdate(const MatrixBase<Derived>& w,typename NumTraits<typename MatrixType::Scalar>::Real sigma)
 {
   const Index size = w.rows();
   if (m_isInitialized)
@@ -479,7 +472,7 @@ LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::rankUpdate(const MatrixBase<Deri
     for (Index i = 0; i < size; i++)
       m_transpositions.coeffRef(i) = i;
     m_temporary.resize(size);
-    m_sign = sigma>=0 ? internal::PositiveSemiDef : internal::NegativeSemiDef;
+    m_sign = sigma>=0 ? 1 : -1;
     m_isInitialized = true;
   }
 
@@ -510,21 +503,16 @@ struct solve_retval<LDLT<_MatrixType,_UpLo>, Rhs>
     using std::abs;
     using std::max;
     typedef typename LDLTType::MatrixType MatrixType;
+    typedef typename LDLTType::Scalar Scalar;
     typedef typename LDLTType::RealScalar RealScalar;
-    const typename Diagonal<const MatrixType>::RealReturnType vectorD(dec().vectorD());
-    // In some previous versions, tolerance was set to the max of 1/highest and the maximal diagonal entry * epsilon
-    // as motivated by LAPACK's xGELSS:
-    // RealScalar tolerance = (max)(vectorD.array().abs().maxCoeff() *NumTraits<RealScalar>::epsilon(),RealScalar(1) / NumTraits<RealScalar>::highest());
-    // However, LDLT is not rank revealing, and so adjusting the tolerance wrt to the highest
-    // diagonal element is not well justified and to numerical issues in some cases.
-    // Moreover, Lapack's xSYTRS routines use 0 for the tolerance.
-    RealScalar tolerance = RealScalar(1) / NumTraits<RealScalar>::highest();
-    
+    const Diagonal<const MatrixType> vectorD = dec().vectorD();
+    RealScalar tolerance = (max)(vectorD.array().abs().maxCoeff() * NumTraits<Scalar>::epsilon(),
+				 RealScalar(1) / NumTraits<RealScalar>::highest()); // motivated by LAPACK's xGELSS
     for (Index i = 0; i < vectorD.size(); ++i) {
       if(abs(vectorD(i)) > tolerance)
-        dst.row(i) /= vectorD(i);
+	dst.row(i) /= vectorD(i);
       else
-        dst.row(i).setZero();
+	dst.row(i).setZero();
     }
 
     // dst = L^-T (D^-1 L^-1 P b)
@@ -577,7 +565,7 @@ MatrixType LDLT<MatrixType,_UpLo>::reconstructedMatrix() const
   // L^* P
   res = matrixU() * res;
   // D(L^*P)
-  res = vectorD().real().asDiagonal() * res;
+  res = vectorD().asDiagonal() * res;
   // L(DL^*P)
   res = matrixL() * res;
   // P^T (LDL^*P)

Eigen/src/Cholesky/LLT.h

@@ -174,12 +174,6 @@ template<typename _MatrixType, int _UpLo> class LLT
     LLT rankUpdate(const VectorType& vec, const RealScalar& sigma = 1);
 
   protected:
-    
-    static void check_template_parameters()
-    {
-      EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
-    }
-    
     /** \internal
       * Used to compute and store L
       * The strict upper part is not used and even not initialized.
@@ -196,7 +190,6 @@ template<typename Scalar, int UpLo> struct llt_inplace;
 template<typename MatrixType, typename VectorType>
 static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma)
 {
-  using std::sqrt;
   typedef typename MatrixType::Scalar Scalar;
   typedef typename MatrixType::RealScalar RealScalar;
   typedef typename MatrixType::Index Index;
@@ -206,7 +199,7 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
   typedef Matrix<Scalar,Dynamic,1> TempVectorType;
   typedef typename TempVectorType::SegmentReturnType TempVecSegment;
 
-  Index n = mat.cols();
+  int n = mat.cols();
   eigen_assert(mat.rows()==n && vec.size()==n);
 
   TempVectorType temp;
@@ -218,12 +211,12 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
     // i.e., for sigma > 0
     temp = sqrt(sigma) * vec;
 
-    for(Index i=0; i<n; ++i)
+    for(int i=0; i<n; ++i)
     {
       JacobiRotation<Scalar> g;
       g.makeGivens(mat(i,i), -temp(i), &mat(i,i));
 
-      Index rs = n-i-1;
+      int rs = n-i-1;
       if(rs>0)
       {
         ColXprSegment x(mat.col(i).tail(rs));
@@ -236,12 +229,12 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
   {
     temp = vec;
     RealScalar beta = 1;
-    for(Index j=0; j<n; ++j)
+    for(int j=0; j<n; ++j)
     {
-      RealScalar Ljj = numext::real(mat.coeff(j,j));
-      RealScalar dj = numext::abs2(Ljj);
+      RealScalar Ljj = real(mat.coeff(j,j));
+      RealScalar dj = abs2(Ljj);
       Scalar wj = temp.coeff(j);
-      RealScalar swj2 = sigma*numext::abs2(wj);
+      RealScalar swj2 = sigma*abs2(wj);
       RealScalar gamma = dj*beta + swj2;
 
       RealScalar x = dj + swj2/beta;
@@ -257,7 +250,7 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
       {
         temp.tail(rs) -= (wj/Ljj) * mat.col(j).tail(rs);
         if(gamma != 0)
-          mat.col(j).tail(rs) = (nLjj/Ljj) * mat.col(j).tail(rs) + (nLjj * sigma*numext::conj(wj)/gamma)*temp.tail(rs);
+          mat.col(j).tail(rs) = (nLjj/Ljj) * mat.col(j).tail(rs) + (nLjj * sigma*conj(wj)/gamma)*temp.tail(rs);
       }
     }
   }
@@ -270,7 +263,6 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
   template<typename MatrixType>
   static typename MatrixType::Index unblocked(MatrixType& mat)
   {
-    using std::sqrt;
     typedef typename MatrixType::Index Index;
     
     eigen_assert(mat.rows()==mat.cols());
@@ -283,13 +275,13 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
       Block<MatrixType,1,Dynamic> A10(mat,k,0,1,k);
       Block<MatrixType,Dynamic,Dynamic> A20(mat,k+1,0,rs,k);
 
-      RealScalar x = numext::real(mat.coeff(k,k));
+      RealScalar x = real(mat.coeff(k,k));
       if (k>0) x -= A10.squaredNorm();
       if (x<=RealScalar(0))
         return k;
       mat.coeffRef(k,k) = x = sqrt(x);
       if (k>0 && rs>0) A21.noalias() -= A20 * A10.adjoint();
-      if (rs>0) A21 /= x;
+      if (rs>0) A21 *= RealScalar(1)/x;
     }
     return -1;
   }
@@ -390,8 +382,6 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Upper>
 template<typename MatrixType, int _UpLo>
 LLT<MatrixType,_UpLo>& LLT<MatrixType,_UpLo>::compute(const MatrixType& a)
 {
-  check_template_parameters();
-  
   eigen_assert(a.rows()==a.cols());
   const Index size = a.rows();
   m_matrix.resize(size, size);

Eigen/src/Cholesky/LLT_MKL.h

@@ -60,7 +60,7 @@ template<> struct mkl_llt<EIGTYPE> \
     lda = m.outerStride(); \
 \
     info = LAPACKE_##MKLPREFIX##potrf( matrix_order, uplo, size, (MKLTYPE*)a, lda ); \
-    info = (info==0) ? -1 : info>0 ? info-1 : size; \
+    info = (info==0) ? Success : NumericalIssue; \
     return info; \
   } \
 }; \

Eigen/src/CholmodSupport/CholmodSupport.h

@@ -51,6 +51,7 @@ void cholmod_configure_matrix(CholmodType& mat)
 template<typename _Scalar, int _Options, typename _Index>
 cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
 {
+  typedef SparseMatrix<_Scalar,_Options,_Index> MatrixType;
   cholmod_sparse res;
   res.nzmax   = mat.nonZeros();
   res.nrow    = mat.rows();;
@@ -58,12 +59,10 @@ cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
   res.p       = mat.outerIndexPtr();
   res.i       = mat.innerIndexPtr();
   res.x       = mat.valuePtr();
-  res.z       = 0;
   res.sorted  = 1;
   if(mat.isCompressed())
   {
     res.packed  = 1;
-    res.nz = 0;
   }
   else
   {
@@ -78,13 +77,9 @@ cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
   {
     res.itype = CHOLMOD_INT;
   }
-  else if (internal::is_same<_Index,SuiteSparse_long>::value)
-  {
-    res.itype = CHOLMOD_LONG;
-  }
   else
   {
-    eigen_assert(false && "Index type not supported yet");
+    eigen_assert(false && "Index type different than int is not supported yet");
   }
 
   // setup res.xtype
@@ -128,7 +123,7 @@ cholmod_dense viewAsCholmod(MatrixBase<Derived>& mat)
   res.ncol   = mat.cols();
   res.nzmax  = res.nrow * res.ncol;
   res.d      = Derived::IsVectorAtCompileTime ? mat.derived().size() : mat.derived().outerStride();
-  res.x      = (void*)(mat.derived().data());
+  res.x      = mat.derived().data();
   res.z      = 0;
 
   internal::cholmod_configure_matrix<Scalar>(res);
@@ -142,8 +137,8 @@ template<typename Scalar, int Flags, typename Index>
 MappedSparseMatrix<Scalar,Flags,Index> viewAsEigen(cholmod_sparse& cm)
 {
   return MappedSparseMatrix<Scalar,Flags,Index>
-         (cm.nrow, cm.ncol, static_cast<Index*>(cm.p)[cm.ncol],
-          static_cast<Index*>(cm.p), static_cast<Index*>(cm.i),static_cast<Scalar*>(cm.x) );
+         (cm.nrow, cm.ncol, reinterpret_cast<Index*>(cm.p)[cm.ncol],
+          reinterpret_cast<Index*>(cm.p), reinterpret_cast<Index*>(cm.i),reinterpret_cast<Scalar*>(cm.x) );
 }
 
 enum CholmodMode {
@@ -172,14 +167,12 @@ class CholmodBase : internal::noncopyable
     CholmodBase()
       : m_cholmodFactor(0), m_info(Success), m_isInitialized(false)
     {
-      m_shiftOffset[0] = m_shiftOffset[1] = RealScalar(0.0);
       cholmod_start(&m_cholmod);
     }
 
     CholmodBase(const MatrixType& matrix)
       : m_cholmodFactor(0), m_info(Success), m_isInitialized(false)
     {
-      m_shiftOffset[0] = m_shiftOffset[1] = RealScalar(0.0);
       cholmod_start(&m_cholmod);
       compute(matrix);
     }
@@ -244,7 +237,7 @@ class CholmodBase : internal::noncopyable
       return internal::sparse_solve_retval<CholmodBase, Rhs>(*this, b.derived());
     }
     
-    /** Performs a symbolic decomposition on the sparsity pattern of \a matrix.
+    /** Performs a symbolic decomposition on the sparcity of \a matrix.
       *
       * This function is particularly useful when solving for several problems having the same structure.
       * 
@@ -268,7 +261,7 @@ class CholmodBase : internal::noncopyable
     
     /** Performs a numeric decomposition of \a matrix
       *
-      * The given matrix must have the same sparsity pattern as the matrix on which the symbolic decomposition has been performed.
+      * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed.
       *
       * \sa analyzePattern()
       */
@@ -276,10 +269,9 @@ class CholmodBase : internal::noncopyable
     {
       eigen_assert(m_analysisIsOk && "You must first call analyzePattern()");
       cholmod_sparse A = viewAsCholmod(matrix.template selfadjointView<UpLo>());
-      cholmod_factorize_p(&A, m_shiftOffset, 0, 0, m_cholmodFactor, &m_cholmod);
+      cholmod_factorize(&A, m_cholmodFactor, &m_cholmod);
       
-      // If the factorization failed, minor is the column at which it did. On success minor == n.
-      this->m_info = (m_cholmodFactor->minor == m_cholmodFactor->n ? Success : NumericalIssue);
+      this->m_info = Success;
       m_factorizationIsOk = true;
     }
     
@@ -294,18 +286,16 @@ class CholmodBase : internal::noncopyable
     {
       eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()");
       const Index size = m_cholmodFactor->n;
-      EIGEN_UNUSED_VARIABLE(size);
       eigen_assert(size==b.rows());
 
       // note: cd stands for Cholmod Dense
-      Rhs& b_ref(b.const_cast_derived());
-      cholmod_dense b_cd = viewAsCholmod(b_ref);
+      cholmod_dense b_cd = viewAsCholmod(b.const_cast_derived());
       cholmod_dense* x_cd = cholmod_solve(CHOLMOD_A, m_cholmodFactor, &b_cd, &m_cholmod);
       if(!x_cd)
       {
         this->m_info = NumericalIssue;
       }
-      // TODO optimize this copy by swapping when possible (be careful with alignment, etc.)
+      // TODO optimize this copy by swapping when possible (be carreful with alignment, etc.)
       dest = Matrix<Scalar,Dest::RowsAtCompileTime,Dest::ColsAtCompileTime>::Map(reinterpret_cast<Scalar*>(x_cd->x),b.rows(),b.cols());
       cholmod_free_dense(&x_cd, &m_cholmod);
     }
@@ -316,7 +306,6 @@ class CholmodBase : internal::noncopyable
     {
       eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()");
       const Index size = m_cholmodFactor->n;
-      EIGEN_UNUSED_VARIABLE(size);
       eigen_assert(size==b.rows());
 
       // note: cs stands for Cholmod Sparse
@@ -326,36 +315,19 @@ class CholmodBase : internal::noncopyable
       {
         this->m_info = NumericalIssue;
       }
-      // TODO optimize this copy by swapping when possible (be careful with alignment, etc.)
+      // TODO optimize this copy by swapping when possible (be carreful with alignment, etc.)
       dest = viewAsEigen<DestScalar,DestOptions,DestIndex>(*x_cs);
       cholmod_free_sparse(&x_cs, &m_cholmod);
     }
     #endif // EIGEN_PARSED_BY_DOXYGEN
     
-    
-    /** Sets the shift parameter that will be used to adjust the diagonal coefficients during the numerical factorization.
-      *
-      * During the numerical factorization, an offset term is added to the diagonal coefficients:\n
-      * \c d_ii = \a offset + \c d_ii
-      *
-      * The default is \a offset=0.
-      *
-      * \returns a reference to \c *this.
-      */
-    Derived& setShift(const RealScalar& offset)
-    {
-      m_shiftOffset[0] = offset;
-      return derived();
-    }
-    
     template<typename Stream>
-    void dumpMemory(Stream& /*s*/)
+    void dumpMemory(Stream& s)
     {}
     
   protected:
     mutable cholmod_common m_cholmod;
     cholmod_factor* m_cholmodFactor;
-    RealScalar m_shiftOffset[2];
     mutable ComputationInfo m_info;
     bool m_isInitialized;
     int m_factorizationIsOk;
@@ -368,8 +340,8 @@ class CholmodBase : internal::noncopyable
   *
   * This class allows to solve for A.X = B sparse linear problems via a simplicial LL^T Cholesky factorization
   * using the Cholmod library.
-  * This simplicial variant is equivalent to Eigen's built-in SimplicialLLT class. Therefore, it has little practical interest.
-  * The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices
+  * This simplicial variant is equivalent to Eigen's built-in SimplicialLLT class. Thefore, it has little practical interest.
+  * The sparse matrix A must be selfajoint and positive definite. The vectors or matrices
   * X and B can be either dense or sparse.
   *
   * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<>
@@ -395,7 +367,7 @@ class CholmodSimplicialLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimpl
     CholmodSimplicialLLT(const MatrixType& matrix) : Base()
     {
       init();
-      Base::compute(matrix);
+      compute(matrix);
     }
 
     ~CholmodSimplicialLLT() {}
@@ -415,8 +387,8 @@ class CholmodSimplicialLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimpl
   *
   * This class allows to solve for A.X = B sparse linear problems via a simplicial LDL^T Cholesky factorization
   * using the Cholmod library.
-  * This simplicial variant is equivalent to Eigen's built-in SimplicialLDLT class. Therefore, it has little practical interest.
-  * The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices
+  * This simplicial variant is equivalent to Eigen's built-in SimplicialLDLT class. Thefore, it has little practical interest.
+  * The sparse matrix A must be selfajoint and positive definite. The vectors or matrices
   * X and B can be either dense or sparse.
   *
   * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<>
@@ -442,7 +414,7 @@ class CholmodSimplicialLDLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimp
     CholmodSimplicialLDLT(const MatrixType& matrix) : Base()
     {
       init();
-      Base::compute(matrix);
+      compute(matrix);
     }
 
     ~CholmodSimplicialLDLT() {}
@@ -461,7 +433,7 @@ class CholmodSimplicialLDLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimp
   * This class allows to solve for A.X = B sparse linear problems via a supernodal LL^T Cholesky factorization
   * using the Cholmod library.
   * This supernodal variant performs best on dense enough problems, e.g., 3D FEM, or very high order 2D FEM.
-  * The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices
+  * The sparse matrix A must be selfajoint and positive definite. The vectors or matrices
   * X and B can be either dense or sparse.
   *
   * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<>
@@ -487,7 +459,7 @@ class CholmodSupernodalLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSuper
     CholmodSupernodalLLT(const MatrixType& matrix) : Base()
     {
       init();
-      Base::compute(matrix);
+      compute(matrix);
     }
 
     ~CholmodSupernodalLLT() {}
@@ -504,7 +476,7 @@ class CholmodSupernodalLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSuper
   * \brief A general Cholesky factorization and solver based on Cholmod
   *
   * This class allows to solve for A.X = B sparse linear problems via a LL^T or LDL^T Cholesky factorization
-  * using the Cholmod library. The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices
+  * using the Cholmod library. The sparse matrix A must be selfajoint and positive definite. The vectors or matrices
   * X and B can be either dense or sparse.
   *
   * This variant permits to change the underlying Cholesky method at runtime.
@@ -534,7 +506,7 @@ class CholmodDecomposition : public CholmodBase<_MatrixType, _UpLo, CholmodDecom
     CholmodDecomposition(const MatrixType& matrix) : Base()
     {
       init();
-      Base::compute(matrix);
+      compute(matrix);
     }
 
     ~CholmodDecomposition() {}

Eigen/src/Core/Array.h

@@ -107,10 +107,10 @@ class Array
       *
       * \sa resize(Index,Index)
       */
-    EIGEN_STRONG_INLINE Array() : Base()
+    EIGEN_STRONG_INLINE explicit Array() : Base()
     {
       Base::_check_template_params();
-      EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+      EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 
 #ifndef EIGEN_PARSED_BY_DOXYGEN
@@ -120,22 +120,7 @@ class Array
       : Base(internal::constructor_without_unaligned_array_assert())
     {
       Base::_check_template_params();
-      EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
-    }
-#endif
-
-#ifdef EIGEN_HAVE_RVALUE_REFERENCES
-    Array(Array&& other)
-      : Base(std::move(other))
-    {
-      Base::_check_template_params();
-      if (RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic)
-        Base::_set_noalias(other);
-    }
-    Array& operator=(Array&& other)
-    {
-      other.swap(*this);
-      return *this;
+      EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 #endif
 
@@ -152,15 +137,15 @@ class Array
       EIGEN_STATIC_ASSERT_VECTOR_ONLY(Array)
       eigen_assert(dim >= 0);
       eigen_assert(SizeAtCompileTime == Dynamic || SizeAtCompileTime == dim);
-      EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+      EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 
     #ifndef EIGEN_PARSED_BY_DOXYGEN
     template<typename T0, typename T1>
-    EIGEN_STRONG_INLINE Array(const T0& val0, const T1& val1)
+    EIGEN_STRONG_INLINE Array(const T0& x, const T1& y)
     {
       Base::_check_template_params();
-      this->template _init2<T0,T1>(val0, val1);
+      this->template _init2<T0,T1>(x, y);
     }
     #else
     /** constructs an uninitialized matrix with \a rows rows and \a cols columns.
@@ -170,27 +155,27 @@ class Array
       * Matrix() instead. */
     Array(Index rows, Index cols);
     /** constructs an initialized 2D vector with given coefficients */
-    Array(const Scalar& val0, const Scalar& val1);
+    Array(const Scalar& x, const Scalar& y);
     #endif
 
     /** constructs an initialized 3D vector with given coefficients */
-    EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2)
+    EIGEN_STRONG_INLINE Array(const Scalar& x, const Scalar& y, const Scalar& z)
     {
       Base::_check_template_params();
       EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 3)
-      m_storage.data()[0] = val0;
-      m_storage.data()[1] = val1;
-      m_storage.data()[2] = val2;
+      m_storage.data()[0] = x;
+      m_storage.data()[1] = y;
+      m_storage.data()[2] = z;
     }
     /** constructs an initialized 4D vector with given coefficients */
-    EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2, const Scalar& val3)
+    EIGEN_STRONG_INLINE Array(const Scalar& x, const Scalar& y, const Scalar& z, const Scalar& w)
     {
       Base::_check_template_params();
       EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 4)
-      m_storage.data()[0] = val0;
-      m_storage.data()[1] = val1;
-      m_storage.data()[2] = val2;
-      m_storage.data()[3] = val3;
+      m_storage.data()[0] = x;
+      m_storage.data()[1] = y;
+      m_storage.data()[2] = z;
+      m_storage.data()[3] = w;
     }
 
     explicit Array(const Scalar *data);
@@ -225,7 +210,7 @@ class Array
       : Base(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
     {
       Base::_check_template_params();
-      Base::_resize_to_match(other);
+      Base::resize(other.rows(), other.cols());
       *this = other;
     }
 

Eigen/src/Core/ArrayBase.h

@@ -46,6 +46,9 @@ template<typename Derived> class ArrayBase
 
     typedef ArrayBase Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl;
 
+    using internal::special_scalar_op_base<Derived,typename internal::traits<Derived>::Scalar,
+                typename NumTraits<typename internal::traits<Derived>::Scalar>::Real>::operator*;
+
     typedef typename internal::traits<Derived>::StorageKind StorageKind;
     typedef typename internal::traits<Derived>::Index Index;
     typedef typename internal::traits<Derived>::Scalar Scalar;
@@ -53,7 +56,6 @@ template<typename Derived> class ArrayBase
     typedef typename NumTraits<Scalar>::Real RealScalar;
 
     typedef DenseBase<Derived> Base;
-    using Base::operator*;
     using Base::RowsAtCompileTime;
     using Base::ColsAtCompileTime;
     using Base::SizeAtCompileTime;
@@ -141,7 +143,7 @@ template<typename Derived> class ArrayBase
     ArrayBase<Derived>& array() { return *this; }
     const ArrayBase<Derived>& array() const { return *this; }
 
-    /** \returns an \link Eigen::MatrixBase Matrix \endlink expression of this array
+    /** \returns an \link MatrixBase Matrix \endlink expression of this array
       * \sa MatrixBase::array() */
     MatrixWrapper<Derived> matrix() { return derived(); }
     const MatrixWrapper<const Derived> matrix() const { return derived(); }

Eigen/src/Core/ArrayWrapper.h

@@ -29,11 +29,6 @@ struct traits<ArrayWrapper<ExpressionType> >
   : public traits<typename remove_all<typename ExpressionType::Nested>::type >
 {
   typedef ArrayXpr XprKind;
-  // Let's remove NestByRefBit
-  enum {
-    Flags0 = traits<typename remove_all<typename ExpressionType::Nested>::type >::Flags,
-    Flags = Flags0 & ~NestByRefBit
-  };
 };
 }
 
@@ -63,19 +58,19 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
     inline ScalarWithConstIfNotLvalue* data() { return m_expression.const_cast_derived().data(); }
     inline const Scalar* data() const { return m_expression.data(); }
 
-    inline CoeffReturnType coeff(Index rowId, Index colId) const
+    inline CoeffReturnType coeff(Index row, Index col) const
     {
-      return m_expression.coeff(rowId, colId);
+      return m_expression.coeff(row, col);
     }
 
-    inline Scalar& coeffRef(Index rowId, Index colId)
+    inline Scalar& coeffRef(Index row, Index col)
     {
-      return m_expression.const_cast_derived().coeffRef(rowId, colId);
+      return m_expression.const_cast_derived().coeffRef(row, col);
     }
 
-    inline const Scalar& coeffRef(Index rowId, Index colId) const
+    inline const Scalar& coeffRef(Index row, Index col) const
     {
-      return m_expression.const_cast_derived().coeffRef(rowId, colId);
+      return m_expression.const_cast_derived().coeffRef(row, col);
     }
 
     inline CoeffReturnType coeff(Index index) const
@@ -94,15 +89,15 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
     }
 
     template<int LoadMode>
-    inline const PacketScalar packet(Index rowId, Index colId) const
+    inline const PacketScalar packet(Index row, Index col) const
     {
-      return m_expression.template packet<LoadMode>(rowId, colId);
+      return m_expression.template packet<LoadMode>(row, col);
     }
 
     template<int LoadMode>
-    inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
+    inline void writePacket(Index row, Index col, const PacketScalar& x)
     {
-      m_expression.const_cast_derived().template writePacket<LoadMode>(rowId, colId, val);
+      m_expression.const_cast_derived().template writePacket<LoadMode>(row, col, x);
     }
 
     template<int LoadMode>
@@ -112,9 +107,9 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
     }
 
     template<int LoadMode>
-    inline void writePacket(Index index, const PacketScalar& val)
+    inline void writePacket(Index index, const PacketScalar& x)
     {
-      m_expression.const_cast_derived().template writePacket<LoadMode>(index, val);
+      m_expression.const_cast_derived().template writePacket<LoadMode>(index, x);
     }
 
     template<typename Dest>
@@ -154,11 +149,6 @@ struct traits<MatrixWrapper<ExpressionType> >
  : public traits<typename remove_all<typename ExpressionType::Nested>::type >
 {
   typedef MatrixXpr XprKind;
-  // Let's remove NestByRefBit
-  enum {
-    Flags0 = traits<typename remove_all<typename ExpressionType::Nested>::type >::Flags,
-    Flags = Flags0 & ~NestByRefBit
-  };
 };
 }
 
@@ -178,7 +168,7 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
 
     typedef typename internal::nested<ExpressionType>::type NestedExpressionType;
 
-    inline MatrixWrapper(ExpressionType& a_matrix) : m_expression(a_matrix) {}
+    inline MatrixWrapper(ExpressionType& matrix) : m_expression(matrix) {}
 
     inline Index rows() const { return m_expression.rows(); }
     inline Index cols() const { return m_expression.cols(); }
@@ -188,19 +178,19 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
     inline ScalarWithConstIfNotLvalue* data() { return m_expression.const_cast_derived().data(); }
     inline const Scalar* data() const { return m_expression.data(); }
 
-    inline CoeffReturnType coeff(Index rowId, Index colId) const
+    inline CoeffReturnType coeff(Index row, Index col) const
     {
-      return m_expression.coeff(rowId, colId);
+      return m_expression.coeff(row, col);
     }
 
-    inline Scalar& coeffRef(Index rowId, Index colId)
+    inline Scalar& coeffRef(Index row, Index col)
     {
-      return m_expression.const_cast_derived().coeffRef(rowId, colId);
+      return m_expression.const_cast_derived().coeffRef(row, col);
     }
 
-    inline const Scalar& coeffRef(Index rowId, Index colId) const
+    inline const Scalar& coeffRef(Index row, Index col) const
     {
-      return m_expression.derived().coeffRef(rowId, colId);
+      return m_expression.derived().coeffRef(row, col);
     }
 
     inline CoeffReturnType coeff(Index index) const
@@ -219,15 +209,15 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
     }
 
     template<int LoadMode>
-    inline const PacketScalar packet(Index rowId, Index colId) const
+    inline const PacketScalar packet(Index row, Index col) const
     {
-      return m_expression.template packet<LoadMode>(rowId, colId);
+      return m_expression.template packet<LoadMode>(row, col);
     }
 
     template<int LoadMode>
-    inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
+    inline void writePacket(Index row, Index col, const PacketScalar& x)
     {
-      m_expression.const_cast_derived().template writePacket<LoadMode>(rowId, colId, val);
+      m_expression.const_cast_derived().template writePacket<LoadMode>(row, col, x);
     }
 
     template<int LoadMode>
@@ -237,9 +227,9 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
     }
 
     template<int LoadMode>
-    inline void writePacket(Index index, const PacketScalar& val)
+    inline void writePacket(Index index, const PacketScalar& x)
     {
-      m_expression.const_cast_derived().template writePacket<LoadMode>(index, val);
+      m_expression.const_cast_derived().template writePacket<LoadMode>(index, x);
     }
 
     const typename internal::remove_all<NestedExpressionType>::type& 

Eigen/src/Core/Assign.h

@@ -155,7 +155,7 @@ struct assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, Stop, Stop>
 template<typename Derived1, typename Derived2, int Index, int Stop>
 struct assign_DefaultTraversal_InnerUnrolling
 {
-  static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src, typename Derived1::Index outer)
+  static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src, int outer)
   {
     dst.copyCoeffByOuterInner(outer, Index, src);
     assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src, outer);
@@ -165,7 +165,7 @@ struct assign_DefaultTraversal_InnerUnrolling
 template<typename Derived1, typename Derived2, int Stop>
 struct assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, Stop, Stop>
 {
-  static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &, typename Derived1::Index) {}
+  static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &, int) {}
 };
 
 /***********************
@@ -218,7 +218,7 @@ struct assign_innervec_CompleteUnrolling<Derived1, Derived2, Stop, Stop>
 template<typename Derived1, typename Derived2, int Index, int Stop>
 struct assign_innervec_InnerUnrolling
 {
-  static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src, typename Derived1::Index outer)
+  static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src, int outer)
   {
     dst.template copyPacketByOuterInner<Derived2, Aligned, Aligned>(outer, Index, src);
     assign_innervec_InnerUnrolling<Derived1, Derived2,
@@ -229,7 +229,7 @@ struct assign_innervec_InnerUnrolling
 template<typename Derived1, typename Derived2, int Stop>
 struct assign_innervec_InnerUnrolling<Derived1, Derived2, Stop, Stop>
 {
-  static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &, typename Derived1::Index) {}
+  static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &, int) {}
 };
 
 /***************************************************************************
@@ -439,26 +439,19 @@ struct assign_impl<Derived1, Derived2, SliceVectorizedTraversal, NoUnrolling, Ve
   typedef typename Derived1::Index Index;
   static inline void run(Derived1 &dst, const Derived2 &src)
   {
-    typedef typename Derived1::Scalar Scalar;
-    typedef packet_traits<Scalar> PacketTraits;
+    typedef packet_traits<typename Derived1::Scalar> PacketTraits;
     enum {
       packetSize = PacketTraits::size,
       alignable = PacketTraits::AlignedOnScalar,
-      dstIsAligned = assign_traits<Derived1,Derived2>::DstIsAligned,
-      dstAlignment = alignable ? Aligned : int(dstIsAligned),
+      dstAlignment = alignable ? Aligned : int(assign_traits<Derived1,Derived2>::DstIsAligned) ,
       srcAlignment = assign_traits<Derived1,Derived2>::JointAlignment
     };
-    const Scalar *dst_ptr = &dst.coeffRef(0,0);
-    if((!bool(dstIsAligned)) && (size_t(dst_ptr) % sizeof(Scalar))>0)
-    {
-      // the pointer is not aligend-on scalar, so alignment is not possible
-      return assign_impl<Derived1,Derived2,DefaultTraversal,NoUnrolling>::run(dst, src);
-    }
     const Index packetAlignedMask = packetSize - 1;
     const Index innerSize = dst.innerSize();
     const Index outerSize = dst.outerSize();
     const Index alignedStep = alignable ? (packetSize - dst.outerStride() % packetSize) & packetAlignedMask : 0;
-    Index alignedStart = ((!alignable) || bool(dstIsAligned)) ? 0 : internal::first_aligned(dst_ptr, innerSize);
+    Index alignedStart = ((!alignable) || assign_traits<Derived1,Derived2>::DstIsAligned) ? 0
+                       : internal::first_aligned(&dst.coeffRef(0,0), innerSize);
 
     for(Index outer = 0; outer < outerSize; ++outer)
     {
@@ -514,19 +507,19 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
 namespace internal {
 
 template<typename Derived, typename OtherDerived,
-         bool EvalBeforeAssigning = (int(internal::traits<OtherDerived>::Flags) & EvalBeforeAssigningBit) != 0,
-         bool NeedToTranspose = ((int(Derived::RowsAtCompileTime) == 1 && int(OtherDerived::ColsAtCompileTime) == 1)
-                              |   // FIXME | instead of || to please GCC 4.4.0 stupid warning "suggest parentheses around &&".
-                                  // revert to || as soon as not needed anymore.
-                                  (int(Derived::ColsAtCompileTime) == 1 && int(OtherDerived::RowsAtCompileTime) == 1))
-                              && int(Derived::SizeAtCompileTime) != 1>
+         bool EvalBeforeAssigning = (int(OtherDerived::Flags) & EvalBeforeAssigningBit) != 0,
+         bool NeedToTranspose = Derived::IsVectorAtCompileTime
+                && OtherDerived::IsVectorAtCompileTime
+                && ((int(Derived::RowsAtCompileTime) == 1 && int(OtherDerived::ColsAtCompileTime) == 1)
+                      |  // FIXME | instead of || to please GCC 4.4.0 stupid warning "suggest parentheses around &&".
+                         // revert to || as soon as not needed anymore.
+                    (int(Derived::ColsAtCompileTime) == 1 && int(OtherDerived::RowsAtCompileTime) == 1))
+                && int(Derived::SizeAtCompileTime) != 1>
 struct assign_selector;
 
 template<typename Derived, typename OtherDerived>
 struct assign_selector<Derived,OtherDerived,false,false> {
   static EIGEN_STRONG_INLINE Derived& run(Derived& dst, const OtherDerived& other) { return dst.lazyAssign(other.derived()); }
-  template<typename ActualDerived, typename ActualOtherDerived>
-  static EIGEN_STRONG_INLINE Derived& evalTo(ActualDerived& dst, const ActualOtherDerived& other) { other.evalTo(dst); return dst; }
 };
 template<typename Derived, typename OtherDerived>
 struct assign_selector<Derived,OtherDerived,true,false> {
@@ -535,8 +528,6 @@ struct assign_selector<Derived,OtherDerived,true,false> {
 template<typename Derived, typename OtherDerived>
 struct assign_selector<Derived,OtherDerived,false,true> {
   static EIGEN_STRONG_INLINE Derived& run(Derived& dst, const OtherDerived& other) { return dst.lazyAssign(other.transpose()); }
-  template<typename ActualDerived, typename ActualOtherDerived>
-  static EIGEN_STRONG_INLINE Derived& evalTo(ActualDerived& dst, const ActualOtherDerived& other) { Transpose<ActualDerived> dstTrans(dst); other.evalTo(dstTrans); return dst; }
 };
 template<typename Derived, typename OtherDerived>
 struct assign_selector<Derived,OtherDerived,true,true> {
@@ -575,14 +566,16 @@ template<typename Derived>
 template <typename OtherDerived>
 EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::operator=(const EigenBase<OtherDerived>& other)
 {
-  return internal::assign_selector<Derived,OtherDerived,false>::evalTo(derived(), other.derived());
+  other.derived().evalTo(derived());
+  return derived();
 }
 
 template<typename Derived>
 template<typename OtherDerived>
 EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::operator=(const ReturnByValue<OtherDerived>& other)
 {
-  return internal::assign_selector<Derived,OtherDerived,false>::evalTo(derived(), other.derived());
+  other.evalTo(derived());
+  return derived();
 }
 
 } // end namespace Eigen

Eigen/src/Core/Block.h

@@ -21,6 +21,7 @@ namespace Eigen {
   * \param XprType the type of the expression in which we are taking a block
   * \param BlockRows the number of rows of the block we are taking at compile time (optional)
   * \param BlockCols the number of columns of the block we are taking at compile time (optional)
+  * \param _DirectAccessStatus \internal used for partial specialization
   *
   * This class represents an expression of either a fixed-size or dynamic-size block. It is the return
   * type of DenseBase::block(Index,Index,Index,Index) and DenseBase::block<int,int>(Index,Index) and
@@ -46,8 +47,8 @@ namespace Eigen {
   */
 
 namespace internal {
-template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel>
-struct traits<Block<XprType, BlockRows, BlockCols, InnerPanel> > : traits<XprType>
+template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool HasDirectAccess>
+struct traits<Block<XprType, BlockRows, BlockCols, InnerPanel, HasDirectAccess> > : traits<XprType>
 {
   typedef typename traits<XprType>::Scalar Scalar;
   typedef typename traits<XprType>::StorageKind StorageKind;
@@ -81,7 +82,7 @@ struct traits<Block<XprType, BlockRows, BlockCols, InnerPanel> > : traits<XprTyp
                        && (InnerStrideAtCompileTime == 1)
                         ? PacketAccessBit : 0,
     MaskAlignedBit = (InnerPanel && (OuterStrideAtCompileTime!=Dynamic) && (((OuterStrideAtCompileTime * int(sizeof(Scalar))) % 16) == 0)) ? AlignedBit : 0,
-    FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1 || (InnerPanel && (traits<XprType>::Flags&LinearAccessBit))) ? LinearAccessBit : 0,
+    FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1) ? LinearAccessBit : 0,
     FlagsLvalueBit = is_lvalue<XprType>::value ? LvalueBit : 0,
     FlagsRowMajorBit = IsRowMajor ? RowMajorBit : 0,
     Flags0 = traits<XprType>::Flags & ( (HereditaryBits & ~RowMajorBit) |
@@ -91,92 +92,21 @@ struct traits<Block<XprType, BlockRows, BlockCols, InnerPanel> > : traits<XprTyp
     Flags = Flags0 | FlagsLinearAccessBit | FlagsLvalueBit | FlagsRowMajorBit
   };
 };
+}
 
-template<typename XprType, int BlockRows=Dynamic, int BlockCols=Dynamic, bool InnerPanel = false,
-         bool HasDirectAccess = internal::has_direct_access<XprType>::ret> class BlockImpl_dense;
-         
-} // end namespace internal
-
-template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, typename StorageKind> class BlockImpl;
-
-template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel> class Block
-  : public BlockImpl<XprType, BlockRows, BlockCols, InnerPanel, typename internal::traits<XprType>::StorageKind>
+template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool HasDirectAccess> class Block
+  : public internal::dense_xpr_base<Block<XprType, BlockRows, BlockCols, InnerPanel, HasDirectAccess> >::type
 {
-    typedef BlockImpl<XprType, BlockRows, BlockCols, InnerPanel, typename internal::traits<XprType>::StorageKind> Impl;
-  public:
-    //typedef typename Impl::Base Base;
-    typedef Impl Base;
-    EIGEN_GENERIC_PUBLIC_INTERFACE(Block)
-    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Block)
-  
-    /** Column or Row constructor
-      */
-    inline Block(XprType& xpr, Index i) : Impl(xpr,i)
-    {
-      eigen_assert( (i>=0) && (
-          ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && i<xpr.rows())
-        ||((BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) && i<xpr.cols())));
-    }
-
-    /** Fixed-size constructor
-      */
-    inline Block(XprType& xpr, Index a_startRow, Index a_startCol)
-      : Impl(xpr, a_startRow, a_startCol)
-    {
-      EIGEN_STATIC_ASSERT(RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic,THIS_METHOD_IS_ONLY_FOR_FIXED_SIZE)
-      eigen_assert(a_startRow >= 0 && BlockRows >= 1 && a_startRow + BlockRows <= xpr.rows()
-             && a_startCol >= 0 && BlockCols >= 1 && a_startCol + BlockCols <= xpr.cols());
-    }
-
-    /** Dynamic-size constructor
-      */
-    inline Block(XprType& xpr,
-          Index a_startRow, Index a_startCol,
-          Index blockRows, Index blockCols)
-      : Impl(xpr, a_startRow, a_startCol, blockRows, blockCols)
-    {
-      eigen_assert((RowsAtCompileTime==Dynamic || RowsAtCompileTime==blockRows)
-          && (ColsAtCompileTime==Dynamic || ColsAtCompileTime==blockCols));
-      eigen_assert(a_startRow >= 0 && blockRows >= 0 && a_startRow  <= xpr.rows() - blockRows
-          && a_startCol >= 0 && blockCols >= 0 && a_startCol <= xpr.cols() - blockCols);
-    }
-};
-         
-// The generic default implementation for dense block simplu forward to the internal::BlockImpl_dense
-// that must be specialized for direct and non-direct access...
-template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel>
-class BlockImpl<XprType, BlockRows, BlockCols, InnerPanel, Dense>
-  : public internal::BlockImpl_dense<XprType, BlockRows, BlockCols, InnerPanel>
-{
-    typedef internal::BlockImpl_dense<XprType, BlockRows, BlockCols, InnerPanel> Impl;
-    typedef typename XprType::Index Index;
-  public:
-    typedef Impl Base;
-    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl)
-    inline BlockImpl(XprType& xpr, Index i) : Impl(xpr,i) {}
-    inline BlockImpl(XprType& xpr, Index a_startRow, Index a_startCol) : Impl(xpr, a_startRow, a_startCol) {}
-    inline BlockImpl(XprType& xpr, Index a_startRow, Index a_startCol, Index blockRows, Index blockCols)
-      : Impl(xpr, a_startRow, a_startCol, blockRows, blockCols) {}
-};
-
-namespace internal {
-
-/** \internal Internal implementation of dense Blocks in the general case. */
-template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool HasDirectAccess> class BlockImpl_dense
-  : public internal::dense_xpr_base<Block<XprType, BlockRows, BlockCols, InnerPanel> >::type
-{
-    typedef Block<XprType, BlockRows, BlockCols, InnerPanel> BlockType;
   public:
 
-    typedef typename internal::dense_xpr_base<BlockType>::type Base;
-    EIGEN_DENSE_PUBLIC_INTERFACE(BlockType)
-    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense)
+    typedef typename internal::dense_xpr_base<Block>::type Base;
+    EIGEN_DENSE_PUBLIC_INTERFACE(Block)
 
     class InnerIterator;
 
     /** Column or Row constructor
       */
-    inline BlockImpl_dense(XprType& xpr, Index i)
+    inline Block(XprType& xpr, Index i)
       : m_xpr(xpr),
         // It is a row if and only if BlockRows==1 and BlockCols==XprType::ColsAtCompileTime,
         // and it is a column if and only if BlockRows==XprType::RowsAtCompileTime and BlockCols==1,
@@ -186,43 +116,58 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
         m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0),
         m_blockRows(BlockRows==1 ? 1 : xpr.rows()),
         m_blockCols(BlockCols==1 ? 1 : xpr.cols())
-    {}
+    {
+      eigen_assert( (i>=0) && (
+          ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && i<xpr.rows())
+        ||((BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) && i<xpr.cols())));
+    }
 
     /** Fixed-size constructor
       */
-    inline BlockImpl_dense(XprType& xpr, Index a_startRow, Index a_startCol)
-      : m_xpr(xpr), m_startRow(a_startRow), m_startCol(a_startCol),
-                    m_blockRows(BlockRows), m_blockCols(BlockCols)
-    {}
+    inline Block(XprType& xpr, Index startRow, Index startCol)
+      : m_xpr(xpr), m_startRow(startRow), m_startCol(startCol),
+        m_blockRows(BlockRows), m_blockCols(BlockCols)
+    {
+      EIGEN_STATIC_ASSERT(RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic,THIS_METHOD_IS_ONLY_FOR_FIXED_SIZE)
+      eigen_assert(startRow >= 0 && BlockRows >= 1 && startRow + BlockRows <= xpr.rows()
+             && startCol >= 0 && BlockCols >= 1 && startCol + BlockCols <= xpr.cols());
+    }
 
     /** Dynamic-size constructor
       */
-    inline BlockImpl_dense(XprType& xpr,
-          Index a_startRow, Index a_startCol,
+    inline Block(XprType& xpr,
+          Index startRow, Index startCol,
           Index blockRows, Index blockCols)
-      : m_xpr(xpr), m_startRow(a_startRow), m_startCol(a_startCol),
-                    m_blockRows(blockRows), m_blockCols(blockCols)
-    {}
+      : m_xpr(xpr), m_startRow(startRow), m_startCol(startCol),
+                          m_blockRows(blockRows), m_blockCols(blockCols)
+    {
+      eigen_assert((RowsAtCompileTime==Dynamic || RowsAtCompileTime==blockRows)
+          && (ColsAtCompileTime==Dynamic || ColsAtCompileTime==blockCols));
+      eigen_assert(startRow >= 0 && blockRows >= 0 && startRow + blockRows <= xpr.rows()
+          && startCol >= 0 && blockCols >= 0 && startCol + blockCols <= xpr.cols());
+    }
+
+    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Block)
 
     inline Index rows() const { return m_blockRows.value(); }
     inline Index cols() const { return m_blockCols.value(); }
 
-    inline Scalar& coeffRef(Index rowId, Index colId)
+    inline Scalar& coeffRef(Index row, Index col)
     {
       EIGEN_STATIC_ASSERT_LVALUE(XprType)
       return m_xpr.const_cast_derived()
-               .coeffRef(rowId + m_startRow.value(), colId + m_startCol.value());
+               .coeffRef(row + m_startRow.value(), col + m_startCol.value());
     }
 
-    inline const Scalar& coeffRef(Index rowId, Index colId) const
+    inline const Scalar& coeffRef(Index row, Index col) const
     {
       return m_xpr.derived()
-               .coeffRef(rowId + m_startRow.value(), colId + m_startCol.value());
+               .coeffRef(row + m_startRow.value(), col + m_startCol.value());
     }
 
-    EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index row, Index col) const
     {
-      return m_xpr.coeff(rowId + m_startRow.value(), colId + m_startCol.value());
+      return m_xpr.coeff(row + m_startRow.value(), col + m_startCol.value());
     }
 
     inline Scalar& coeffRef(Index index)
@@ -248,17 +193,17 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
     }
 
     template<int LoadMode>
-    inline PacketScalar packet(Index rowId, Index colId) const
+    inline PacketScalar packet(Index row, Index col) const
     {
       return m_xpr.template packet<Unaligned>
-              (rowId + m_startRow.value(), colId + m_startCol.value());
+              (row + m_startRow.value(), col + m_startCol.value());
     }
 
     template<int LoadMode>
-    inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
+    inline void writePacket(Index row, Index col, const PacketScalar& x)
     {
       m_xpr.const_cast_derived().template writePacket<Unaligned>
-              (rowId + m_startRow.value(), colId + m_startCol.value(), val);
+              (row + m_startRow.value(), col + m_startCol.value(), x);
     }
 
     template<int LoadMode>
@@ -270,11 +215,11 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
     }
 
     template<int LoadMode>
-    inline void writePacket(Index index, const PacketScalar& val)
+    inline void writePacket(Index index, const PacketScalar& x)
     {
       m_xpr.const_cast_derived().template writePacket<Unaligned>
          (m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
-          m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0), val);
+          m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0), x);
     }
 
     #ifdef EIGEN_PARSED_BY_DOXYGEN
@@ -308,21 +253,21 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
     const internal::variable_if_dynamic<Index, ColsAtCompileTime> m_blockCols;
 };
 
-/** \internal Internal implementation of dense Blocks in the direct access case.*/
+/** \internal */
 template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel>
-class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
-  : public MapBase<Block<XprType, BlockRows, BlockCols, InnerPanel> >
+class Block<XprType,BlockRows,BlockCols, InnerPanel,true>
+  : public MapBase<Block<XprType, BlockRows, BlockCols, InnerPanel, true> >
 {
-    typedef Block<XprType, BlockRows, BlockCols, InnerPanel> BlockType;
   public:
 
-    typedef MapBase<BlockType> Base;
-    EIGEN_DENSE_PUBLIC_INTERFACE(BlockType)
-    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense)
+    typedef MapBase<Block> Base;
+    EIGEN_DENSE_PUBLIC_INTERFACE(Block)
+
+    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Block)
 
     /** Column or Row constructor
       */
-    inline BlockImpl_dense(XprType& xpr, Index i)
+    inline Block(XprType& xpr, Index i)
       : Base(internal::const_cast_ptr(&xpr.coeffRef(
               (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0,
               (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0)),
@@ -330,25 +275,34 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
              BlockCols==1 ? 1 : xpr.cols()),
         m_xpr(xpr)
     {
+      eigen_assert( (i>=0) && (
+          ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && i<xpr.rows())
+        ||((BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) && i<xpr.cols())));
       init();
     }
 
     /** Fixed-size constructor
       */
-    inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol)
+    inline Block(XprType& xpr, Index startRow, Index startCol)
       : Base(internal::const_cast_ptr(&xpr.coeffRef(startRow,startCol))), m_xpr(xpr)
     {
+      eigen_assert(startRow >= 0 && BlockRows >= 1 && startRow + BlockRows <= xpr.rows()
+             && startCol >= 0 && BlockCols >= 1 && startCol + BlockCols <= xpr.cols());
       init();
     }
 
     /** Dynamic-size constructor
       */
-    inline BlockImpl_dense(XprType& xpr,
+    inline Block(XprType& xpr,
           Index startRow, Index startCol,
           Index blockRows, Index blockCols)
       : Base(internal::const_cast_ptr(&xpr.coeffRef(startRow,startCol)), blockRows, blockCols),
         m_xpr(xpr)
     {
+      eigen_assert((RowsAtCompileTime==Dynamic || RowsAtCompileTime==blockRows)
+             && (ColsAtCompileTime==Dynamic || ColsAtCompileTime==blockCols));
+      eigen_assert(startRow >= 0 && blockRows >= 0 && startRow + blockRows <= xpr.rows()
+             && startCol >= 0 && blockCols >= 0 && startCol + blockCols <= xpr.cols());
       init();
     }
 
@@ -360,7 +314,7 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
     /** \sa MapBase::innerStride() */
     inline Index innerStride() const
     {
-      return internal::traits<BlockType>::HasSameStorageOrderAsXprType
+      return internal::traits<Block>::HasSameStorageOrderAsXprType
              ? m_xpr.innerStride()
              : m_xpr.outerStride();
     }
@@ -379,7 +333,7 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
 
     #ifndef EIGEN_PARSED_BY_DOXYGEN
     /** \internal used by allowAligned() */
-    inline BlockImpl_dense(XprType& xpr, const Scalar* data, Index blockRows, Index blockCols)
+    inline Block(XprType& xpr, const Scalar* data, Index blockRows, Index blockCols)
       : Base(data, blockRows, blockCols), m_xpr(xpr)
     {
       init();
@@ -389,7 +343,7 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
   protected:
     void init()
     {
-      m_outerStride = internal::traits<BlockType>::HasSameStorageOrderAsXprType
+      m_outerStride = internal::traits<Block>::HasSameStorageOrderAsXprType
                     ? m_xpr.outerStride()
                     : m_xpr.innerStride();
     }
@@ -398,8 +352,6 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
     Index m_outerStride;
 };
 
-} // end namespace internal
-
 } // end namespace Eigen
 
 #endif // EIGEN_BLOCK_H

Eigen/src/Core/BooleanRedux.h

@@ -29,9 +29,9 @@ struct all_unroller
 };
 
 template<typename Derived>
-struct all_unroller<Derived, 0>
+struct all_unroller<Derived, 1>
 {
-  static inline bool run(const Derived &/*mat*/) { return true; }
+  static inline bool run(const Derived &mat) { return mat.coeff(0, 0); }
 };
 
 template<typename Derived>
@@ -55,9 +55,9 @@ struct any_unroller
 };
 
 template<typename Derived>
-struct any_unroller<Derived, 0>
+struct any_unroller<Derived, 1>
 {
-  static inline bool run(const Derived & /*mat*/) { return false; }
+  static inline bool run(const Derived &mat) { return mat.coeff(0, 0); }
 };
 
 template<typename Derived>
@@ -85,7 +85,9 @@ inline bool DenseBase<Derived>::all() const
           && SizeAtCompileTime * (CoeffReadCost + NumTraits<Scalar>::AddCost) <= EIGEN_UNROLLING_LIMIT
   };
   if(unroll)
-    return internal::all_unroller<Derived, unroll ? int(SizeAtCompileTime) : Dynamic>::run(derived());
+    return internal::all_unroller<Derived,
+                           unroll ? int(SizeAtCompileTime) : Dynamic
+     >::run(derived());
   else
   {
     for(Index j = 0; j < cols(); ++j)
@@ -109,7 +111,9 @@ inline bool DenseBase<Derived>::any() const
           && SizeAtCompileTime * (CoeffReadCost + NumTraits<Scalar>::AddCost) <= EIGEN_UNROLLING_LIMIT
   };
   if(unroll)
-    return internal::any_unroller<Derived, unroll ? int(SizeAtCompileTime) : Dynamic>::run(derived());
+    return internal::any_unroller<Derived,
+                           unroll ? int(SizeAtCompileTime) : Dynamic
+           >::run(derived());
   else
   {
     for(Index j = 0; j < cols(); ++j)
@@ -129,26 +133,6 @@ inline typename DenseBase<Derived>::Index DenseBase<Derived>::count() const
   return derived().template cast<bool>().template cast<Index>().sum();
 }
 
-/** \returns true is \c *this contains at least one Not A Number (NaN).
-  *
-  * \sa allFinite()
-  */
-template<typename Derived>
-inline bool DenseBase<Derived>::hasNaN() const
-{
-  return !((derived().array()==derived().array()).all());
-}
-
-/** \returns true if \c *this contains only finite numbers, i.e., no NaN and no +/-INF values.
-  *
-  * \sa hasNaN()
-  */
-template<typename Derived>
-inline bool DenseBase<Derived>::allFinite() const
-{
-  return !((derived()-derived()).hasNaN());
-}
-    
 } // end namespace Eigen
 
 #endif // EIGEN_ALLANDANY_H

Eigen/src/Core/CommaInitializer.h

@@ -43,17 +43,6 @@ struct CommaInitializer
     m_xpr.block(0, 0, other.rows(), other.cols()) = other;
   }
 
-  /* Copy/Move constructor which transfers ownership. This is crucial in 
-   * absence of return value optimization to avoid assertions during destruction. */
-  // FIXME in C++11 mode this could be replaced by a proper RValue constructor
-  inline CommaInitializer(const CommaInitializer& o)
-  : m_xpr(o.m_xpr), m_row(o.m_row), m_col(o.m_col), m_currentBlockRows(o.m_currentBlockRows) {
-    // Mark original object as finished. In absence of R-value references we need to const_cast:
-    const_cast<CommaInitializer&>(o).m_row = m_xpr.rows();
-    const_cast<CommaInitializer&>(o).m_col = m_xpr.cols();
-    const_cast<CommaInitializer&>(o).m_currentBlockRows = 0;
-  }
-
   /* inserts a scalar value in the target matrix */
   CommaInitializer& operator,(const Scalar& s)
   {
@@ -76,7 +65,9 @@ struct CommaInitializer
   template<typename OtherDerived>
   CommaInitializer& operator,(const DenseBase<OtherDerived>& other)
   {
-    if (m_col==m_xpr.cols() && (other.cols()!=0 || other.rows()!=m_currentBlockRows))
+    if(other.cols()==0 || other.rows()==0)
+      return *this;
+    if (m_col==m_xpr.cols())
     {
       m_row+=m_currentBlockRows;
       m_col = 0;
@@ -84,18 +75,24 @@ struct CommaInitializer
       eigen_assert(m_row+m_currentBlockRows<=m_xpr.rows()
         && "Too many rows passed to comma initializer (operator<<)");
     }
-    eigen_assert((m_col + other.cols() <= m_xpr.cols())
+    eigen_assert(m_col<m_xpr.cols()
       && "Too many coefficients passed to comma initializer (operator<<)");
     eigen_assert(m_currentBlockRows==other.rows());
-    m_xpr.template block<OtherDerived::RowsAtCompileTime, OtherDerived::ColsAtCompileTime>
-                    (m_row, m_col, other.rows(), other.cols()) = other;
+    if (OtherDerived::SizeAtCompileTime != Dynamic)
+      m_xpr.template block<OtherDerived::RowsAtCompileTime != Dynamic ? OtherDerived::RowsAtCompileTime : 1,
+                              OtherDerived::ColsAtCompileTime != Dynamic ? OtherDerived::ColsAtCompileTime : 1>
+                    (m_row, m_col) = other;
+    else
+      m_xpr.block(m_row, m_col, other.rows(), other.cols()) = other;
     m_col += other.cols();
     return *this;
   }
 
   inline ~CommaInitializer()
   {
-      finished();
+    eigen_assert((m_row+m_currentBlockRows) == m_xpr.rows()
+         && m_col == m_xpr.cols()
+         && "Too few coefficients passed to comma initializer (operator<<)");
   }
 
   /** \returns the built matrix once all its coefficients have been set.
@@ -105,12 +102,7 @@ struct CommaInitializer
     * quaternion.fromRotationMatrix((Matrix3f() << axis0, axis1, axis2).finished());
     * \endcode
     */
-  inline XprType& finished() {
-      eigen_assert(((m_row+m_currentBlockRows) == m_xpr.rows() || m_xpr.cols() == 0)
-           && m_col == m_xpr.cols()
-           && "Too few coefficients passed to comma initializer (operator<<)");
-      return m_xpr;
-  }
+  inline XprType& finished() { return m_xpr; }
 
   XprType& m_xpr;   // target expression
   Index m_row;              // current row id
@@ -126,8 +118,6 @@ struct CommaInitializer
   *
   * Example: \include MatrixBase_set.cpp
   * Output: \verbinclude MatrixBase_set.out
-  * 
-  * \note According the c++ standard, the argument expressions of this comma initializer are evaluated in arbitrary order.
   *
   * \sa CommaInitializer::finished(), class CommaInitializer
   */

Eigen/src/Core/CwiseBinaryOp.h

@@ -81,8 +81,7 @@ struct traits<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >
         )
      ),
     Flags = (Flags0 & ~RowMajorBit) | (LhsFlags & RowMajorBit),
-    Cost0 = EIGEN_ADD_COST(LhsCoeffReadCost,RhsCoeffReadCost),
-    CoeffReadCost = EIGEN_ADD_COST(Cost0,functor_traits<BinaryOp>::Cost)
+    CoeffReadCost = LhsCoeffReadCost + RhsCoeffReadCost + functor_traits<BinaryOp>::Cost
   };
 };
 } // end namespace internal
@@ -95,8 +94,8 @@ struct traits<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >
 // So allowing mixing different types gives very unexpected errors when enabling vectorization, when the user tries to
 // add together a float matrix and a double matrix.
 #define EIGEN_CHECK_BINARY_COMPATIBILIY(BINOP,LHS,RHS) \
-  EIGEN_STATIC_ASSERT((internal::functor_is_product_like<BINOP>::ret \
-                        ? int(internal::scalar_product_traits<LHS, RHS>::Defined) \
+  EIGEN_STATIC_ASSERT((internal::functor_allows_mixing_real_and_complex<BINOP>::ret \
+                        ? int(internal::is_same<typename NumTraits<LHS>::Real, typename NumTraits<RHS>::Real>::value) \
                         : int(internal::is_same<LHS, RHS>::value)), \
     YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
 
@@ -123,13 +122,13 @@ class CwiseBinaryOp : internal::no_assignment_operator,
     typedef typename internal::remove_reference<LhsNested>::type _LhsNested;
     typedef typename internal::remove_reference<RhsNested>::type _RhsNested;
 
-    EIGEN_STRONG_INLINE CwiseBinaryOp(const Lhs& aLhs, const Rhs& aRhs, const BinaryOp& func = BinaryOp())
-      : m_lhs(aLhs), m_rhs(aRhs), m_functor(func)
+    EIGEN_STRONG_INLINE CwiseBinaryOp(const Lhs& lhs, const Rhs& rhs, const BinaryOp& func = BinaryOp())
+      : m_lhs(lhs), m_rhs(rhs), m_functor(func)
     {
       EIGEN_CHECK_BINARY_COMPATIBILIY(BinaryOp,typename Lhs::Scalar,typename Rhs::Scalar);
       // require the sizes to match
       EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Lhs, Rhs)
-      eigen_assert(aLhs.rows() == aRhs.rows() && aLhs.cols() == aRhs.cols());
+      eigen_assert(lhs.rows() == rhs.rows() && lhs.cols() == rhs.cols());
     }
 
     EIGEN_STRONG_INLINE Index rows() const {
@@ -170,17 +169,17 @@ class CwiseBinaryOpImpl<BinaryOp, Lhs, Rhs, Dense>
     typedef typename internal::dense_xpr_base<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >::type Base;
     EIGEN_DENSE_PUBLIC_INTERFACE( Derived )
 
-    EIGEN_STRONG_INLINE const Scalar coeff(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const
     {
-      return derived().functor()(derived().lhs().coeff(rowId, colId),
-                                 derived().rhs().coeff(rowId, colId));
+      return derived().functor()(derived().lhs().coeff(row, col),
+                                 derived().rhs().coeff(row, col));
     }
 
     template<int LoadMode>
-    EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE PacketScalar packet(Index row, Index col) const
     {
-      return derived().functor().packetOp(derived().lhs().template packet<LoadMode>(rowId, colId),
-                                          derived().rhs().template packet<LoadMode>(rowId, colId));
+      return derived().functor().packetOp(derived().lhs().template packet<LoadMode>(row, col),
+                                          derived().rhs().template packet<LoadMode>(row, col));
     }
 
     EIGEN_STRONG_INLINE const Scalar coeff(Index index) const

Eigen/src/Core/CwiseNullaryOp.h

@@ -54,27 +54,27 @@ class CwiseNullaryOp : internal::no_assignment_operator,
     typedef typename internal::dense_xpr_base<CwiseNullaryOp>::type Base;
     EIGEN_DENSE_PUBLIC_INTERFACE(CwiseNullaryOp)
 
-    CwiseNullaryOp(Index nbRows, Index nbCols, const NullaryOp& func = NullaryOp())
-      : m_rows(nbRows), m_cols(nbCols), m_functor(func)
+    CwiseNullaryOp(Index rows, Index cols, const NullaryOp& func = NullaryOp())
+      : m_rows(rows), m_cols(cols), m_functor(func)
     {
-      eigen_assert(nbRows >= 0
-            && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == nbRows)
-            &&  nbCols >= 0
-            && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == nbCols));
+      eigen_assert(rows >= 0
+            && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows)
+            &&  cols >= 0
+            && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols));
     }
 
     EIGEN_STRONG_INLINE Index rows() const { return m_rows.value(); }
     EIGEN_STRONG_INLINE Index cols() const { return m_cols.value(); }
 
-    EIGEN_STRONG_INLINE const Scalar coeff(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE const Scalar coeff(Index rows, Index cols) const
     {
-      return m_functor(rowId, colId);
+      return m_functor(rows, cols);
     }
 
     template<int LoadMode>
-    EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE PacketScalar packet(Index row, Index col) const
     {
-      return m_functor.packetOp(rowId, colId);
+      return m_functor.packetOp(row, col);
     }
 
     EIGEN_STRONG_INLINE const Scalar coeff(Index index) const
@@ -163,11 +163,11 @@ DenseBase<Derived>::NullaryExpr(const CustomNullaryOp& func)
 
 /** \returns an expression of a constant matrix of value \a value
   *
-  * The parameters \a nbRows and \a nbCols are the number of rows and of columns of
+  * The parameters \a rows and \a cols are the number of rows and of columns of
   * the returned matrix. Must be compatible with this DenseBase type.
   *
   * This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
-  * it is redundant to pass \a nbRows and \a nbCols as arguments, so Zero() should be used
+  * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used
   * instead.
   *
   * The template parameter \a CustomNullaryOp is the type of the functor.
@@ -176,9 +176,9 @@ DenseBase<Derived>::NullaryExpr(const CustomNullaryOp& func)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
-DenseBase<Derived>::Constant(Index nbRows, Index nbCols, const Scalar& value)
+DenseBase<Derived>::Constant(Index rows, Index cols, const Scalar& value)
 {
-  return DenseBase<Derived>::NullaryExpr(nbRows, nbCols, internal::scalar_constant_op<Scalar>(value));
+  return DenseBase<Derived>::NullaryExpr(rows, cols, internal::scalar_constant_op<Scalar>(value));
 }
 
 /** \returns an expression of a constant matrix of value \a value
@@ -292,14 +292,14 @@ DenseBase<Derived>::LinSpaced(const Scalar& low, const Scalar& high)
   return DenseBase<Derived>::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op<Scalar,true>(low,high,Derived::SizeAtCompileTime));
 }
 
-/** \returns true if all coefficients in this matrix are approximately equal to \a val, to within precision \a prec */
+/** \returns true if all coefficients in this matrix are approximately equal to \a value, to within precision \a prec */
 template<typename Derived>
 bool DenseBase<Derived>::isApproxToConstant
-(const Scalar& val, const RealScalar& prec) const
+(const Scalar& value, RealScalar prec) const
 {
   for(Index j = 0; j < cols(); ++j)
     for(Index i = 0; i < rows(); ++i)
-      if(!internal::isApprox(this->coeff(i, j), val, prec))
+      if(!internal::isApprox(this->coeff(i, j), value, prec))
         return false;
   return true;
 }
@@ -309,19 +309,19 @@ bool DenseBase<Derived>::isApproxToConstant
   * \returns true if all coefficients in this matrix are approximately equal to \a value, to within precision \a prec */
 template<typename Derived>
 bool DenseBase<Derived>::isConstant
-(const Scalar& val, const RealScalar& prec) const
+(const Scalar& value, RealScalar prec) const
 {
-  return isApproxToConstant(val, prec);
+  return isApproxToConstant(value, prec);
 }
 
-/** Alias for setConstant(): sets all coefficients in this expression to \a val.
+/** Alias for setConstant(): sets all coefficients in this expression to \a value.
   *
   * \sa setConstant(), Constant(), class CwiseNullaryOp
   */
 template<typename Derived>
-EIGEN_STRONG_INLINE void DenseBase<Derived>::fill(const Scalar& val)
+EIGEN_STRONG_INLINE void DenseBase<Derived>::fill(const Scalar& value)
 {
-  setConstant(val);
+  setConstant(value);
 }
 
 /** Sets all coefficients in this expression to \a value.
@@ -329,9 +329,9 @@ EIGEN_STRONG_INLINE void DenseBase<Derived>::fill(const Scalar& val)
   * \sa fill(), setConstant(Index,const Scalar&), setConstant(Index,Index,const Scalar&), setZero(), setOnes(), Constant(), class CwiseNullaryOp, setZero(), setOnes()
   */
 template<typename Derived>
-EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setConstant(const Scalar& val)
+EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setConstant(const Scalar& value)
 {
-  return derived() = Constant(rows(), cols(), val);
+  return derived() = Constant(rows(), cols(), value);
 }
 
 /** Resizes to the given \a size, and sets all coefficients in this expression to the given \a value.
@@ -345,17 +345,17 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setConstant(const Scalar& val)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setConstant(Index size, const Scalar& val)
+PlainObjectBase<Derived>::setConstant(Index size, const Scalar& value)
 {
   resize(size);
-  return setConstant(val);
+  return setConstant(value);
 }
 
 /** Resizes to the given size, and sets all coefficients in this expression to the given \a value.
   *
-  * \param nbRows the new number of rows
-  * \param nbCols the new number of columns
-  * \param val the value to which all coefficients are set
+  * \param rows the new number of rows
+  * \param cols the new number of columns
+  * \param value the value to which all coefficients are set
   *
   * Example: \include Matrix_setConstant_int_int.cpp
   * Output: \verbinclude Matrix_setConstant_int_int.out
@@ -364,10 +364,10 @@ PlainObjectBase<Derived>::setConstant(Index size, const Scalar& val)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setConstant(Index nbRows, Index nbCols, const Scalar& val)
+PlainObjectBase<Derived>::setConstant(Index rows, Index cols, const Scalar& value)
 {
-  resize(nbRows, nbCols);
-  return setConstant(val);
+  resize(rows, cols);
+  return setConstant(value);
 }
 
 /**
@@ -384,10 +384,10 @@ PlainObjectBase<Derived>::setConstant(Index nbRows, Index nbCols, const Scalar&
   * \sa CwiseNullaryOp
   */
 template<typename Derived>
-EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(Index newSize, const Scalar& low, const Scalar& high)
+EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(Index size, const Scalar& low, const Scalar& high)
 {
   EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
-  return derived() = Derived::NullaryExpr(newSize, internal::linspaced_op<Scalar,false>(low,high,newSize));
+  return derived() = Derived::NullaryExpr(size, internal::linspaced_op<Scalar,false>(low,high,size));
 }
 
 /**
@@ -425,9 +425,9 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(const Scalar& low,
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
-DenseBase<Derived>::Zero(Index nbRows, Index nbCols)
+DenseBase<Derived>::Zero(Index rows, Index cols)
 {
-  return Constant(nbRows, nbCols, Scalar(0));
+  return Constant(rows, cols, Scalar(0));
 }
 
 /** \returns an expression of a zero vector.
@@ -479,7 +479,7 @@ DenseBase<Derived>::Zero()
   * \sa class CwiseNullaryOp, Zero()
   */
 template<typename Derived>
-bool DenseBase<Derived>::isZero(const RealScalar& prec) const
+bool DenseBase<Derived>::isZero(RealScalar prec) const
 {
   for(Index j = 0; j < cols(); ++j)
     for(Index i = 0; i < rows(); ++i)
@@ -512,16 +512,16 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setZero()
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setZero(Index newSize)
+PlainObjectBase<Derived>::setZero(Index size)
 {
-  resize(newSize);
+  resize(size);
   return setConstant(Scalar(0));
 }
 
 /** Resizes to the given size, and sets all coefficients in this expression to zero.
   *
-  * \param nbRows the new number of rows
-  * \param nbCols the new number of columns
+  * \param rows the new number of rows
+  * \param cols the new number of columns
   *
   * Example: \include Matrix_setZero_int_int.cpp
   * Output: \verbinclude Matrix_setZero_int_int.out
@@ -530,9 +530,9 @@ PlainObjectBase<Derived>::setZero(Index newSize)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setZero(Index nbRows, Index nbCols)
+PlainObjectBase<Derived>::setZero(Index rows, Index cols)
 {
-  resize(nbRows, nbCols);
+  resize(rows, cols);
   return setConstant(Scalar(0));
 }
 
@@ -540,7 +540,7 @@ PlainObjectBase<Derived>::setZero(Index nbRows, Index nbCols)
 
 /** \returns an expression of a matrix where all coefficients equal one.
   *
-  * The parameters \a nbRows and \a nbCols are the number of rows and of columns of
+  * The parameters \a rows and \a cols are the number of rows and of columns of
   * the returned matrix. Must be compatible with this MatrixBase type.
   *
   * This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
@@ -554,14 +554,14 @@ PlainObjectBase<Derived>::setZero(Index nbRows, Index nbCols)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
-DenseBase<Derived>::Ones(Index nbRows, Index nbCols)
+DenseBase<Derived>::Ones(Index rows, Index cols)
 {
-  return Constant(nbRows, nbCols, Scalar(1));
+  return Constant(rows, cols, Scalar(1));
 }
 
 /** \returns an expression of a vector where all coefficients equal one.
   *
-  * The parameter \a newSize is the size of the returned vector.
+  * The parameter \a size is the size of the returned vector.
   * Must be compatible with this MatrixBase type.
   *
   * \only_for_vectors
@@ -577,9 +577,9 @@ DenseBase<Derived>::Ones(Index nbRows, Index nbCols)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
-DenseBase<Derived>::Ones(Index newSize)
+DenseBase<Derived>::Ones(Index size)
 {
-  return Constant(newSize, Scalar(1));
+  return Constant(size, Scalar(1));
 }
 
 /** \returns an expression of a fixed-size matrix or vector where all coefficients equal one.
@@ -609,7 +609,7 @@ DenseBase<Derived>::Ones()
   */
 template<typename Derived>
 bool DenseBase<Derived>::isOnes
-(const RealScalar& prec) const
+(RealScalar prec) const
 {
   return isApproxToConstant(Scalar(1), prec);
 }
@@ -627,7 +627,7 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setOnes()
   return setConstant(Scalar(1));
 }
 
-/** Resizes to the given \a newSize, and sets all coefficients in this expression to one.
+/** Resizes to the given \a size, and sets all coefficients in this expression to one.
   *
   * \only_for_vectors
   *
@@ -638,16 +638,16 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setOnes()
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setOnes(Index newSize)
+PlainObjectBase<Derived>::setOnes(Index size)
 {
-  resize(newSize);
+  resize(size);
   return setConstant(Scalar(1));
 }
 
 /** Resizes to the given size, and sets all coefficients in this expression to one.
   *
-  * \param nbRows the new number of rows
-  * \param nbCols the new number of columns
+  * \param rows the new number of rows
+  * \param cols the new number of columns
   *
   * Example: \include Matrix_setOnes_int_int.cpp
   * Output: \verbinclude Matrix_setOnes_int_int.out
@@ -656,9 +656,9 @@ PlainObjectBase<Derived>::setOnes(Index newSize)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setOnes(Index nbRows, Index nbCols)
+PlainObjectBase<Derived>::setOnes(Index rows, Index cols)
 {
-  resize(nbRows, nbCols);
+  resize(rows, cols);
   return setConstant(Scalar(1));
 }
 
@@ -666,7 +666,7 @@ PlainObjectBase<Derived>::setOnes(Index nbRows, Index nbCols)
 
 /** \returns an expression of the identity matrix (not necessarily square).
   *
-  * The parameters \a nbRows and \a nbCols are the number of rows and of columns of
+  * The parameters \a rows and \a cols are the number of rows and of columns of
   * the returned matrix. Must be compatible with this MatrixBase type.
   *
   * This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
@@ -680,9 +680,9 @@ PlainObjectBase<Derived>::setOnes(Index nbRows, Index nbCols)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::IdentityReturnType
-MatrixBase<Derived>::Identity(Index nbRows, Index nbCols)
+MatrixBase<Derived>::Identity(Index rows, Index cols)
 {
-  return DenseBase<Derived>::NullaryExpr(nbRows, nbCols, internal::scalar_identity_op<Scalar>());
+  return DenseBase<Derived>::NullaryExpr(rows, cols, internal::scalar_identity_op<Scalar>());
 }
 
 /** \returns an expression of the identity matrix (not necessarily square).
@@ -714,7 +714,7 @@ MatrixBase<Derived>::Identity()
   */
 template<typename Derived>
 bool MatrixBase<Derived>::isIdentity
-(const RealScalar& prec) const
+(RealScalar prec) const
 {
   for(Index j = 0; j < cols(); ++j)
   {
@@ -776,8 +776,8 @@ EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity()
 
 /** \brief Resizes to the given size, and writes the identity expression (not necessarily square) into *this.
   *
-  * \param nbRows the new number of rows
-  * \param nbCols the new number of columns
+  * \param rows the new number of rows
+  * \param cols the new number of columns
   *
   * Example: \include Matrix_setIdentity_int_int.cpp
   * Output: \verbinclude Matrix_setIdentity_int_int.out
@@ -785,9 +785,9 @@ EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity()
   * \sa MatrixBase::setIdentity(), class CwiseNullaryOp, MatrixBase::Identity()
   */
 template<typename Derived>
-EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity(Index nbRows, Index nbCols)
+EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity(Index rows, Index cols)
 {
-  derived().resize(nbRows, nbCols);
+  derived().resize(rows, cols);
   return setIdentity();
 }
 
@@ -798,10 +798,10 @@ EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity(Index nbRows, Inde
   * \sa MatrixBase::Unit(Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
   */
 template<typename Derived>
-EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::Unit(Index newSize, Index i)
+EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::Unit(Index size, Index i)
 {
   EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
-  return BasisReturnType(SquareMatrixType::Identity(newSize,newSize), i);
+  return BasisReturnType(SquareMatrixType::Identity(size,size), i);
 }
 
 /** \returns an expression of the i-th unit (basis) vector.

Eigen/src/Core/CwiseUnaryOp.h

@@ -47,7 +47,7 @@ struct traits<CwiseUnaryOp<UnaryOp, XprType> >
     Flags = _XprTypeNested::Flags & (
       HereditaryBits | LinearAccessBit | AlignedBit
       | (functor_traits<UnaryOp>::PacketAccess ? PacketAccessBit : 0)),
-    CoeffReadCost = EIGEN_ADD_COST(_XprTypeNested::CoeffReadCost, functor_traits<UnaryOp>::Cost)
+    CoeffReadCost = _XprTypeNested::CoeffReadCost + functor_traits<UnaryOp>::Cost
   };
 };
 }
@@ -98,15 +98,15 @@ class CwiseUnaryOpImpl<UnaryOp,XprType,Dense>
     typedef typename internal::dense_xpr_base<CwiseUnaryOp<UnaryOp, XprType> >::type Base;
     EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
 
-    EIGEN_STRONG_INLINE const Scalar coeff(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const
     {
-      return derived().functor()(derived().nestedExpression().coeff(rowId, colId));
+      return derived().functor()(derived().nestedExpression().coeff(row, col));
     }
 
     template<int LoadMode>
-    EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE PacketScalar packet(Index row, Index col) const
     {
-      return derived().functor().packetOp(derived().nestedExpression().template packet<LoadMode>(rowId, colId));
+      return derived().functor().packetOp(derived().nestedExpression().template packet<LoadMode>(row, col));
     }
 
     EIGEN_STRONG_INLINE const Scalar coeff(Index index) const

Eigen/src/Core/CwiseUnaryView.h

@@ -38,7 +38,7 @@ struct traits<CwiseUnaryView<ViewOp, MatrixType> >
   typedef typename remove_all<MatrixTypeNested>::type _MatrixTypeNested;
   enum {
     Flags = (traits<_MatrixTypeNested>::Flags & (HereditaryBits | LvalueBit | LinearAccessBit | DirectAccessBit)),
-    CoeffReadCost = EIGEN_ADD_COST(traits<_MatrixTypeNested>::CoeffReadCost, functor_traits<ViewOp>::Cost),
+    CoeffReadCost = traits<_MatrixTypeNested>::CoeffReadCost + functor_traits<ViewOp>::Cost,
     MatrixTypeInnerStride =  inner_stride_at_compile_time<MatrixType>::ret,
     // need to cast the sizeof's from size_t to int explicitly, otherwise:
     // "error: no integral type can represent all of the enumerator values

Eigen/src/Core/DenseBase.h

@@ -13,16 +13,6 @@
 
 namespace Eigen {
 
-namespace internal {
-  
-// The index type defined by EIGEN_DEFAULT_DENSE_INDEX_TYPE must be a signed type.
-// This dummy function simply aims at checking that at compile time.
-static inline void check_DenseIndex_is_signed() {
-  EIGEN_STATIC_ASSERT(NumTraits<DenseIndex>::IsSigned,THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE); 
-}
-
-} // end namespace internal
-  
 /** \class DenseBase
   * \ingroup Core_Module
   *
@@ -40,14 +30,15 @@ static inline void check_DenseIndex_is_signed() {
   */
 template<typename Derived> class DenseBase
 #ifndef EIGEN_PARSED_BY_DOXYGEN
-  : public internal::special_scalar_op_base<Derived, typename internal::traits<Derived>::Scalar,
-                                            typename NumTraits<typename internal::traits<Derived>::Scalar>::Real,
-                                            DenseCoeffsBase<Derived> >
+  : public internal::special_scalar_op_base<Derived,typename internal::traits<Derived>::Scalar,
+                                     typename NumTraits<typename internal::traits<Derived>::Scalar>::Real>
 #else
   : public DenseCoeffsBase<Derived>
 #endif // not EIGEN_PARSED_BY_DOXYGEN
 {
   public:
+    using internal::special_scalar_op_base<Derived,typename internal::traits<Derived>::Scalar,
+                typename NumTraits<typename internal::traits<Derived>::Scalar>::Real>::operator*;
 
     class InnerIterator;
 
@@ -62,9 +53,8 @@ template<typename Derived> class DenseBase
     typedef typename internal::traits<Derived>::Scalar Scalar;
     typedef typename internal::packet_traits<Scalar>::type PacketScalar;
     typedef typename NumTraits<Scalar>::Real RealScalar;
-    typedef internal::special_scalar_op_base<Derived,Scalar,RealScalar, DenseCoeffsBase<Derived> > Base;
 
-    using Base::operator*;
+    typedef DenseCoeffsBase<Derived> Base;
     using Base::derived;
     using Base::const_cast_derived;
     using Base::rows;
@@ -183,6 +173,10 @@ template<typename Derived> class DenseBase
     /** \returns the number of nonzero coefficients which is in practice the number
       * of stored coefficients. */
     inline Index nonZeros() const { return size(); }
+    /** \returns true if either the number of rows or the number of columns is equal to 1.
+      * In other words, this function returns
+      * \code rows()==1 || cols()==1 \endcode
+      * \sa rows(), cols(), IsVectorAtCompileTime. */
 
     /** \returns the outer size.
       *
@@ -210,21 +204,21 @@ template<typename Derived> class DenseBase
       * Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does
       * nothing else.
       */
-    void resize(Index newSize)
+    void resize(Index size)
     {
-      EIGEN_ONLY_USED_FOR_DEBUG(newSize);
-      eigen_assert(newSize == this->size()
+      EIGEN_ONLY_USED_FOR_DEBUG(size);
+      eigen_assert(size == this->size()
                 && "DenseBase::resize() does not actually allow to resize.");
     }
     /** Only plain matrices/arrays, not expressions, may be resized; therefore the only useful resize methods are
       * Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does
       * nothing else.
       */
-    void resize(Index nbRows, Index nbCols)
+    void resize(Index rows, Index cols)
     {
-      EIGEN_ONLY_USED_FOR_DEBUG(nbRows);
-      EIGEN_ONLY_USED_FOR_DEBUG(nbCols);
-      eigen_assert(nbRows == this->rows() && nbCols == this->cols()
+      EIGEN_ONLY_USED_FOR_DEBUG(rows);
+      EIGEN_ONLY_USED_FOR_DEBUG(cols);
+      eigen_assert(rows == this->rows() && cols == this->cols()
                 && "DenseBase::resize() does not actually allow to resize.");
     }
 
@@ -262,13 +256,11 @@ template<typename Derived> class DenseBase
     template<typename OtherDerived>
     Derived& operator=(const ReturnByValue<OtherDerived>& func);
 
-    /** \internal Copies \a other into *this without evaluating other. \returns a reference to *this. */
+#ifndef EIGEN_PARSED_BY_DOXYGEN
+    /** Copies \a other into *this without evaluating other. \returns a reference to *this. */
     template<typename OtherDerived>
     Derived& lazyAssign(const DenseBase<OtherDerived>& other);
-
-    /** \internal Evaluates \a other into *this. \returns a reference to *this. */
-    template<typename OtherDerived>
-    Derived& lazyAssign(const ReturnByValue<OtherDerived>& other);
+#endif // not EIGEN_PARSED_BY_DOXYGEN
 
     CommaInitializer<Derived> operator<< (const Scalar& s);
 
@@ -279,7 +271,7 @@ template<typename Derived> class DenseBase
     CommaInitializer<Derived> operator<< (const DenseBase<OtherDerived>& other);
 
     Eigen::Transpose<Derived> transpose();
-	typedef typename internal::add_const<Transpose<const Derived> >::type ConstTransposeReturnType;
+    typedef const Transpose<const Derived> ConstTransposeReturnType;
     ConstTransposeReturnType transpose() const;
     void transposeInPlace();
 #ifndef EIGEN_NO_DEBUG
@@ -289,6 +281,29 @@ template<typename Derived> class DenseBase
   public:
 #endif
 
+    typedef VectorBlock<Derived> SegmentReturnType;
+    typedef const VectorBlock<const Derived> ConstSegmentReturnType;
+    template<int Size> struct FixedSegmentReturnType { typedef VectorBlock<Derived, Size> Type; };
+    template<int Size> struct ConstFixedSegmentReturnType { typedef const VectorBlock<const Derived, Size> Type; };
+    
+    // Note: The "DenseBase::" prefixes are added to help MSVC9 to match these declarations with the later implementations.
+    SegmentReturnType segment(Index start, Index size);
+    typename DenseBase::ConstSegmentReturnType segment(Index start, Index size) const;
+
+    SegmentReturnType head(Index size);
+    typename DenseBase::ConstSegmentReturnType head(Index size) const;
+
+    SegmentReturnType tail(Index size);
+    typename DenseBase::ConstSegmentReturnType tail(Index size) const;
+
+    template<int Size> typename FixedSegmentReturnType<Size>::Type head();
+    template<int Size> typename ConstFixedSegmentReturnType<Size>::Type head() const;
+
+    template<int Size> typename FixedSegmentReturnType<Size>::Type tail();
+    template<int Size> typename ConstFixedSegmentReturnType<Size>::Type tail() const;
+
+    template<int Size> typename FixedSegmentReturnType<Size>::Type segment(Index start);
+    template<int Size> typename ConstFixedSegmentReturnType<Size>::Type segment(Index start) const;
 
     static const ConstantReturnType
     Constant(Index rows, Index cols, const Scalar& value);
@@ -333,20 +348,17 @@ template<typename Derived> class DenseBase
 
     template<typename OtherDerived>
     bool isApprox(const DenseBase<OtherDerived>& other,
-                  const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
+                  RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
     bool isMuchSmallerThan(const RealScalar& other,
-                           const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
+                           RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
     template<typename OtherDerived>
     bool isMuchSmallerThan(const DenseBase<OtherDerived>& other,
-                           const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
+                           RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
 
-    bool isApproxToConstant(const Scalar& value, const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    bool isConstant(const Scalar& value, const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    bool isZero(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    bool isOnes(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    
-    inline bool hasNaN() const;
-    inline bool allFinite() const;
+    bool isApproxToConstant(const Scalar& value, RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isConstant(const Scalar& value, RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isZero(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isOnes(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
 
     inline Derived& operator*=(const Scalar& other);
     inline Derived& operator/=(const Scalar& other);
@@ -426,6 +438,8 @@ template<typename Derived> class DenseBase
       return derived().coeff(0,0);
     }
 
+/////////// Array module ///////////
+
     bool all(void) const;
     bool any(void) const;
     Index count() const;
@@ -451,19 +465,17 @@ template<typename Derived> class DenseBase
 
     template<typename ThenDerived>
     inline const Select<Derived,ThenDerived, typename ThenDerived::ConstantReturnType>
-    select(const DenseBase<ThenDerived>& thenMatrix, const typename ThenDerived::Scalar& elseScalar) const;
+    select(const DenseBase<ThenDerived>& thenMatrix, typename ThenDerived::Scalar elseScalar) const;
 
     template<typename ElseDerived>
     inline const Select<Derived, typename ElseDerived::ConstantReturnType, ElseDerived >
-    select(const typename ElseDerived::Scalar& thenScalar, const DenseBase<ElseDerived>& elseMatrix) const;
+    select(typename ElseDerived::Scalar thenScalar, const DenseBase<ElseDerived>& elseMatrix) const;
 
     template<int p> RealScalar lpNorm() const;
 
     template<int RowFactor, int ColFactor>
-    inline const Replicate<Derived,RowFactor,ColFactor> replicate() const;
-    
-    typedef Replicate<Derived,Dynamic,Dynamic> ReplicateReturnType;
-    inline const ReplicateReturnType replicate(Index rowFacor,Index colFactor) const;
+    const Replicate<Derived,RowFactor,ColFactor> replicate() const;
+    const Replicate<Derived,Dynamic,Dynamic> replicate(Index rowFacor,Index colFactor) const;
 
     typedef Reverse<Derived, BothDirections> ReverseReturnType;
     typedef const Reverse<const Derived, BothDirections> ConstReverseReturnType;

Eigen/src/Core/DenseCoeffsBase.h

@@ -427,22 +427,22 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
 
     template<int StoreMode>
     EIGEN_STRONG_INLINE void writePacket
-    (Index row, Index col, const typename internal::packet_traits<Scalar>::type& val)
+    (Index row, Index col, const typename internal::packet_traits<Scalar>::type& x)
     {
       eigen_internal_assert(row >= 0 && row < rows()
                         && col >= 0 && col < cols());
-      derived().template writePacket<StoreMode>(row,col,val);
+      derived().template writePacket<StoreMode>(row,col,x);
     }
 
 
     /** \internal */
     template<int StoreMode>
     EIGEN_STRONG_INLINE void writePacketByOuterInner
-    (Index outer, Index inner, const typename internal::packet_traits<Scalar>::type& val)
+    (Index outer, Index inner, const typename internal::packet_traits<Scalar>::type& x)
     {
       writePacket<StoreMode>(rowIndexByOuterInner(outer, inner),
                             colIndexByOuterInner(outer, inner),
-                            val);
+                            x);
     }
 
     /** \internal
@@ -456,10 +456,10 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
       */
     template<int StoreMode>
     EIGEN_STRONG_INLINE void writePacket
-    (Index index, const typename internal::packet_traits<Scalar>::type& val)
+    (Index index, const typename internal::packet_traits<Scalar>::type& x)
     {
       eigen_internal_assert(index >= 0 && index < size());
-      derived().template writePacket<StoreMode>(index,val);
+      derived().template writePacket<StoreMode>(index,x);
     }
 
 #ifndef EIGEN_PARSED_BY_DOXYGEN

Eigen/src/Core/DenseStorage.h

@@ -24,14 +24,6 @@ namespace internal {
 
 struct constructor_without_unaligned_array_assert {};
 
-template<typename T, int Size> void check_static_allocation_size()
-{
-  // if EIGEN_STACK_ALLOCATION_LIMIT is defined to 0, then no limit
-  #if EIGEN_STACK_ALLOCATION_LIMIT
-  EIGEN_STATIC_ASSERT(Size * sizeof(T) <= EIGEN_STACK_ALLOCATION_LIMIT, OBJECT_ALLOCATED_ON_STACK_IS_TOO_BIG);
-  #endif
-}
-
 /** \internal
   * Static array. If the MatrixOrArrayOptions require auto-alignment, the array will be automatically aligned:
   * to 16 bytes boundary if the total size is a multiple of 16 bytes.
@@ -43,16 +35,8 @@ template <typename T, int Size, int MatrixOrArrayOptions,
 struct plain_array
 {
   T array[Size];
-
-  plain_array() 
-  { 
-    check_static_allocation_size<T,Size>();
-  }
-
-  plain_array(constructor_without_unaligned_array_assert) 
-  { 
-    check_static_allocation_size<T,Size>();
-  }
+  plain_array() {}
+  plain_array(constructor_without_unaligned_array_assert) {}
 };
 
 #if defined(EIGEN_DISABLE_UNALIGNED_ARRAY_ASSERT)
@@ -80,17 +64,8 @@ template <typename T, int Size, int MatrixOrArrayOptions>
 struct plain_array<T, Size, MatrixOrArrayOptions, 16>
 {
   EIGEN_USER_ALIGN16 T array[Size];
-
-  plain_array() 
-  { 
-    EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(0xf);
-    check_static_allocation_size<T,Size>();
-  }
-
-  plain_array(constructor_without_unaligned_array_assert) 
-  { 
-    check_static_allocation_size<T,Size>();
-  }
+  plain_array() { EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(0xf) }
+  plain_array(constructor_without_unaligned_array_assert) {}
 };
 
 template <typename T, int MatrixOrArrayOptions, int Alignment>
@@ -122,41 +97,33 @@ template<typename T, int Size, int _Rows, int _Cols, int _Options> class DenseSt
 {
     internal::plain_array<T,Size,_Options> m_data;
   public:
-    DenseStorage() {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert)
+    inline explicit DenseStorage() {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert)
       : m_data(internal::constructor_without_unaligned_array_assert()) {}
-    DenseStorage(const DenseStorage& other) : m_data(other.m_data) {}
-    DenseStorage& operator=(const DenseStorage& other)
-    {
-      if (this != &other) m_data = other.m_data;
-      return *this;
-    }
-    DenseStorage(DenseIndex,DenseIndex,DenseIndex) {}
-    void swap(DenseStorage& other) { std::swap(m_data,other.m_data); }
-    static DenseIndex rows(void) {return _Rows;}
-    static DenseIndex cols(void) {return _Cols;}
-    void conservativeResize(DenseIndex,DenseIndex,DenseIndex) {}
-    void resize(DenseIndex,DenseIndex,DenseIndex) {}
-    const T *data() const { return m_data.array; }
-    T *data() { return m_data.array; }
+    inline DenseStorage(DenseIndex,DenseIndex,DenseIndex) {}
+    inline void swap(DenseStorage& other) { std::swap(m_data,other.m_data); }
+    static inline DenseIndex rows(void) {return _Rows;}
+    static inline DenseIndex cols(void) {return _Cols;}
+    inline void conservativeResize(DenseIndex,DenseIndex,DenseIndex) {}
+    inline void resize(DenseIndex,DenseIndex,DenseIndex) {}
+    inline const T *data() const { return m_data.array; }
+    inline T *data() { return m_data.array; }
 };
 
 // null matrix
 template<typename T, int _Rows, int _Cols, int _Options> class DenseStorage<T, 0, _Rows, _Cols, _Options>
 {
   public:
-    DenseStorage() {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert) {}
-    DenseStorage(const DenseStorage&) {}
-    DenseStorage& operator=(const DenseStorage&) { return *this; }
-    DenseStorage(DenseIndex,DenseIndex,DenseIndex) {}
-    void swap(DenseStorage& ) {}
-    static DenseIndex rows(void) {return _Rows;}
-    static DenseIndex cols(void) {return _Cols;}
-    void conservativeResize(DenseIndex,DenseIndex,DenseIndex) {}
-    void resize(DenseIndex,DenseIndex,DenseIndex) {}
-    const T *data() const { return 0; }
-    T *data() { return 0; }
+    inline explicit DenseStorage() {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert) {}
+    inline DenseStorage(DenseIndex,DenseIndex,DenseIndex) {}
+    inline void swap(DenseStorage& ) {}
+    static inline DenseIndex rows(void) {return _Rows;}
+    static inline DenseIndex cols(void) {return _Cols;}
+    inline void conservativeResize(DenseIndex,DenseIndex,DenseIndex) {}
+    inline void resize(DenseIndex,DenseIndex,DenseIndex) {}
+    inline const T *data() const { return 0; }
+    inline T *data() { return 0; }
 };
 
 // more specializations for null matrices; these are necessary to resolve ambiguities
@@ -176,29 +143,18 @@ template<typename T, int Size, int _Options> class DenseStorage<T, Size, Dynamic
     DenseIndex m_rows;
     DenseIndex m_cols;
   public:
-    DenseStorage() : m_rows(0), m_cols(0) {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert)
+    inline explicit DenseStorage() : m_rows(0), m_cols(0) {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert)
       : m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0), m_cols(0) {}
-    DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows), m_cols(other.m_cols) {}
-    DenseStorage& operator=(const DenseStorage& other)
-    {
-      if (this != &other)
-      {
-        m_data = other.m_data;
-        m_rows = other.m_rows;
-        m_cols = other.m_cols;
-      }
-      return *this;
-    }
-    DenseStorage(DenseIndex, DenseIndex nbRows, DenseIndex nbCols) : m_rows(nbRows), m_cols(nbCols) {}
-    void swap(DenseStorage& other)
+    inline DenseStorage(DenseIndex, DenseIndex rows, DenseIndex cols) : m_rows(rows), m_cols(cols) {}
+    inline void swap(DenseStorage& other)
     { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); }
-    DenseIndex rows() const {return m_rows;}
-    DenseIndex cols() const {return m_cols;}
-    void conservativeResize(DenseIndex, DenseIndex nbRows, DenseIndex nbCols) { m_rows = nbRows; m_cols = nbCols; }
-    void resize(DenseIndex, DenseIndex nbRows, DenseIndex nbCols) { m_rows = nbRows; m_cols = nbCols; }
-    const T *data() const { return m_data.array; }
-    T *data() { return m_data.array; }
+    inline DenseIndex rows(void) const {return m_rows;}
+    inline DenseIndex cols(void) const {return m_cols;}
+    inline void conservativeResize(DenseIndex, DenseIndex rows, DenseIndex cols) { m_rows = rows; m_cols = cols; }
+    inline void resize(DenseIndex, DenseIndex rows, DenseIndex cols) { m_rows = rows; m_cols = cols; }
+    inline const T *data() const { return m_data.array; }
+    inline T *data() { return m_data.array; }
 };
 
 // dynamic-size matrix with fixed-size storage and fixed width
@@ -207,27 +163,17 @@ template<typename T, int Size, int _Cols, int _Options> class DenseStorage<T, Si
     internal::plain_array<T,Size,_Options> m_data;
     DenseIndex m_rows;
   public:
-    DenseStorage() : m_rows(0) {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert)
+    inline explicit DenseStorage() : m_rows(0) {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert)
       : m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0) {}
-    DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows) {}
-    DenseStorage& operator=(const DenseStorage& other)
-    {
-      if (this != &other)
-      {
-        m_data = other.m_data;
-        m_rows = other.m_rows;
-      }
-      return *this;
-    }
-    DenseStorage(DenseIndex, DenseIndex nbRows, DenseIndex) : m_rows(nbRows) {}
-    void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
-    DenseIndex rows(void) const {return m_rows;}
-    DenseIndex cols(void) const {return _Cols;}
-    void conservativeResize(DenseIndex, DenseIndex nbRows, DenseIndex) { m_rows = nbRows; }
-    void resize(DenseIndex, DenseIndex nbRows, DenseIndex) { m_rows = nbRows; }
-    const T *data() const { return m_data.array; }
-    T *data() { return m_data.array; }
+    inline DenseStorage(DenseIndex, DenseIndex rows, DenseIndex) : m_rows(rows) {}
+    inline void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
+    inline DenseIndex rows(void) const {return m_rows;}
+    inline DenseIndex cols(void) const {return _Cols;}
+    inline void conservativeResize(DenseIndex, DenseIndex rows, DenseIndex) { m_rows = rows; }
+    inline void resize(DenseIndex, DenseIndex rows, DenseIndex) { m_rows = rows; }
+    inline const T *data() const { return m_data.array; }
+    inline T *data() { return m_data.array; }
 };
 
 // dynamic-size matrix with fixed-size storage and fixed height
@@ -236,27 +182,17 @@ template<typename T, int Size, int _Rows, int _Options> class DenseStorage<T, Si
     internal::plain_array<T,Size,_Options> m_data;
     DenseIndex m_cols;
   public:
-    DenseStorage() : m_cols(0) {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert)
+    inline explicit DenseStorage() : m_cols(0) {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert)
       : m_data(internal::constructor_without_unaligned_array_assert()), m_cols(0) {}
-    DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_cols(other.m_cols) {}
-    DenseStorage& operator=(const DenseStorage& other)
-    {
-      if (this != &other)
-      {
-        m_data = other.m_data;
-        m_cols = other.m_cols;
-      }
-      return *this;
-    }
-    DenseStorage(DenseIndex, DenseIndex, DenseIndex nbCols) : m_cols(nbCols) {}
-    void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
-    DenseIndex rows(void) const {return _Rows;}
-    DenseIndex cols(void) const {return m_cols;}
-    void conservativeResize(DenseIndex, DenseIndex, DenseIndex nbCols) { m_cols = nbCols; }
-    void resize(DenseIndex, DenseIndex, DenseIndex nbCols) { m_cols = nbCols; }
-    const T *data() const { return m_data.array; }
-    T *data() { return m_data.array; }
+    inline DenseStorage(DenseIndex, DenseIndex, DenseIndex cols) : m_cols(cols) {}
+    inline void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
+    inline DenseIndex rows(void) const {return _Rows;}
+    inline DenseIndex cols(void) const {return m_cols;}
+    inline void conservativeResize(DenseIndex, DenseIndex, DenseIndex cols) { m_cols = cols; }
+    inline void resize(DenseIndex, DenseIndex, DenseIndex cols) { m_cols = cols; }
+    inline const T *data() const { return m_data.array; }
+    inline T *data() { return m_data.array; }
 };
 
 // purely dynamic matrix.
@@ -266,41 +202,24 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
     DenseIndex m_rows;
     DenseIndex m_cols;
   public:
-    DenseStorage() : m_data(0), m_rows(0), m_cols(0) {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert)
+    inline explicit DenseStorage() : m_data(0), m_rows(0), m_cols(0) {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert)
        : m_data(0), m_rows(0), m_cols(0) {}
-    DenseStorage(DenseIndex size, DenseIndex nbRows, DenseIndex nbCols)
-      : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(nbRows), m_cols(nbCols)
+    inline DenseStorage(DenseIndex size, DenseIndex rows, DenseIndex 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 }
-#ifdef EIGEN_HAVE_RVALUE_REFERENCES
-    DenseStorage(DenseStorage&& other)
-      : m_data(std::move(other.m_data))
-      , m_rows(std::move(other.m_rows))
-      , m_cols(std::move(other.m_cols))
-    {
-      other.m_data = nullptr;
-    }
-    DenseStorage& operator=(DenseStorage&& other)
-    {
-      using std::swap;
-      swap(m_data, other.m_data);
-      swap(m_rows, other.m_rows);
-      swap(m_cols, other.m_cols);
-      return *this;
-    }
-#endif
-    ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, m_rows*m_cols); }
-    void swap(DenseStorage& other)
+    inline ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, m_rows*m_cols); }
+    inline void swap(DenseStorage& other)
     { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); }
-    DenseIndex rows(void) const {return m_rows;}
-    DenseIndex cols(void) const {return m_cols;}
-    void conservativeResize(DenseIndex size, DenseIndex nbRows, DenseIndex nbCols)
+    inline DenseIndex rows(void) const {return m_rows;}
+    inline DenseIndex cols(void) const {return m_cols;}
+    inline void conservativeResize(DenseIndex size, DenseIndex rows, DenseIndex cols)
     {
       m_data = internal::conditional_aligned_realloc_new_auto<T,(_Options&DontAlign)==0>(m_data, size, m_rows*m_cols);
-      m_rows = nbRows;
-      m_cols = nbCols;
+      m_rows = rows;
+      m_cols = cols;
     }
-    void resize(DenseIndex size, DenseIndex nbRows, DenseIndex nbCols)
+    void resize(DenseIndex size, DenseIndex rows, DenseIndex cols)
     {
       if(size != m_rows*m_cols)
       {
@@ -311,14 +230,11 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
           m_data = 0;
         EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
       }
-      m_rows = nbRows;
-      m_cols = nbCols;
+      m_rows = rows;
+      m_cols = cols;
     }
-    const T *data() const { return m_data; }
-    T *data() { return m_data; }
-  private:
-    DenseStorage(const DenseStorage&);
-    DenseStorage& operator=(const DenseStorage&);
+    inline const T *data() const { return m_data; }
+    inline T *data() { return m_data; }
 };
 
 // matrix with dynamic width and fixed height (so that matrix has dynamic size).
@@ -327,35 +243,20 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
     T *m_data;
     DenseIndex m_cols;
   public:
-    DenseStorage() : m_data(0), m_cols(0) {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {}
-    DenseStorage(DenseIndex size, DenseIndex, DenseIndex nbCols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_cols(nbCols)
+    inline explicit DenseStorage() : m_data(0), m_cols(0) {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {}
+    inline DenseStorage(DenseIndex size, DenseIndex, DenseIndex cols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_cols(cols)
     { EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN }
-#ifdef EIGEN_HAVE_RVALUE_REFERENCES
-    DenseStorage(DenseStorage&& other)
-      : m_data(std::move(other.m_data))
-      , m_cols(std::move(other.m_cols))
-    {
-      other.m_data = nullptr;
-    }
-    DenseStorage& operator=(DenseStorage&& other)
-    {
-      using std::swap;
-      swap(m_data, other.m_data);
-      swap(m_cols, other.m_cols);
-      return *this;
-    }
-#endif
-    ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Rows*m_cols); }
-    void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
-    static DenseIndex rows(void) {return _Rows;}
-    DenseIndex cols(void) const {return m_cols;}
-    void conservativeResize(DenseIndex size, DenseIndex, DenseIndex nbCols)
+    inline ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Rows*m_cols); }
+    inline void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
+    static inline DenseIndex rows(void) {return _Rows;}
+    inline DenseIndex cols(void) const {return m_cols;}
+    inline void conservativeResize(DenseIndex size, DenseIndex, DenseIndex cols)
     {
       m_data = internal::conditional_aligned_realloc_new_auto<T,(_Options&DontAlign)==0>(m_data, size, _Rows*m_cols);
-      m_cols = nbCols;
+      m_cols = cols;
     }
-    EIGEN_STRONG_INLINE void resize(DenseIndex size, DenseIndex, DenseIndex nbCols)
+    EIGEN_STRONG_INLINE void resize(DenseIndex size, DenseIndex, DenseIndex cols)
     {
       if(size != _Rows*m_cols)
       {
@@ -366,13 +267,10 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
           m_data = 0;
         EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
       }
-      m_cols = nbCols;
+      m_cols = cols;
     }
-    const T *data() const { return m_data; }
-    T *data() { return m_data; }
-  private:
-    DenseStorage(const DenseStorage&);
-    DenseStorage& operator=(const DenseStorage&);
+    inline const T *data() const { return m_data; }
+    inline T *data() { return m_data; }
 };
 
 // matrix with dynamic height and fixed width (so that matrix has dynamic size).
@@ -381,35 +279,20 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
     T *m_data;
     DenseIndex m_rows;
   public:
-    DenseStorage() : m_data(0), m_rows(0) {}
-    DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {}
-    DenseStorage(DenseIndex size, DenseIndex nbRows, DenseIndex) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(nbRows)
+    inline explicit DenseStorage() : m_data(0), m_rows(0) {}
+    inline DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {}
+    inline DenseStorage(DenseIndex size, DenseIndex rows, DenseIndex) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(rows)
     { EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN }
-#ifdef EIGEN_HAVE_RVALUE_REFERENCES
-    DenseStorage(DenseStorage&& other)
-      : m_data(std::move(other.m_data))
-      , m_rows(std::move(other.m_rows))
-    {
-      other.m_data = nullptr;
-    }
-    DenseStorage& operator=(DenseStorage&& other)
-    {
-      using std::swap;
-      swap(m_data, other.m_data);
-      swap(m_rows, other.m_rows);
-      return *this;
-    }
-#endif
-    ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Cols*m_rows); }
-    void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
-    DenseIndex rows(void) const {return m_rows;}
-    static DenseIndex cols(void) {return _Cols;}
-    void conservativeResize(DenseIndex size, DenseIndex nbRows, DenseIndex)
+    inline ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Cols*m_rows); }
+    inline void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
+    inline DenseIndex rows(void) const {return m_rows;}
+    static inline DenseIndex cols(void) {return _Cols;}
+    inline void conservativeResize(DenseIndex size, DenseIndex rows, DenseIndex)
     {
       m_data = internal::conditional_aligned_realloc_new_auto<T,(_Options&DontAlign)==0>(m_data, size, m_rows*_Cols);
-      m_rows = nbRows;
+      m_rows = rows;
     }
-    EIGEN_STRONG_INLINE void resize(DenseIndex size, DenseIndex nbRows, DenseIndex)
+    EIGEN_STRONG_INLINE void resize(DenseIndex size, DenseIndex rows, DenseIndex)
     {
       if(size != m_rows*_Cols)
       {
@@ -420,13 +303,10 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
           m_data = 0;
         EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
       }
-      m_rows = nbRows;
+      m_rows = rows;
     }
-    const T *data() const { return m_data; }
-    T *data() { return m_data; }
-  private:
-    DenseStorage(const DenseStorage&);
-    DenseStorage& operator=(const DenseStorage&);
+    inline const T *data() const { return m_data; }
+    inline T *data() { return m_data; }
 };
 
 } // end namespace Eigen

Eigen/src/Core/Diagonal.h

@@ -41,12 +41,12 @@ struct traits<Diagonal<MatrixType,DiagIndex> >
   typedef typename remove_reference<MatrixTypeNested>::type _MatrixTypeNested;
   typedef typename MatrixType::StorageKind StorageKind;
   enum {
-    RowsAtCompileTime = (int(DiagIndex) == DynamicIndex || int(MatrixType::SizeAtCompileTime) == Dynamic) ? Dynamic
-                      : (EIGEN_PLAIN_ENUM_MIN(MatrixType::RowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0),
-                                              MatrixType::ColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))),
+    RowsAtCompileTime = (int(DiagIndex) == Dynamic || int(MatrixType::SizeAtCompileTime) == Dynamic) ? Dynamic
+    : (EIGEN_PLAIN_ENUM_MIN(MatrixType::RowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0),
+                            MatrixType::ColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))),
     ColsAtCompileTime = 1,
     MaxRowsAtCompileTime = int(MatrixType::MaxSizeAtCompileTime) == Dynamic ? Dynamic
-                         : DiagIndex == DynamicIndex ? EIGEN_SIZE_MIN_PREFER_FIXED(MatrixType::MaxRowsAtCompileTime,
+                         : DiagIndex == Dynamic ? EIGEN_SIZE_MIN_PREFER_FIXED(MatrixType::MaxRowsAtCompileTime,
                                                                               MatrixType::MaxColsAtCompileTime)
                          : (EIGEN_PLAIN_ENUM_MIN(MatrixType::MaxRowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0),
                                                  MatrixType::MaxColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))),
@@ -61,21 +61,20 @@ struct traits<Diagonal<MatrixType,DiagIndex> >
 };
 }
 
-template<typename MatrixType, int _DiagIndex> class Diagonal
-   : public internal::dense_xpr_base< Diagonal<MatrixType,_DiagIndex> >::type
+template<typename MatrixType, int DiagIndex> class Diagonal
+   : public internal::dense_xpr_base< Diagonal<MatrixType,DiagIndex> >::type
 {
   public:
 
-    enum { DiagIndex = _DiagIndex };
     typedef typename internal::dense_xpr_base<Diagonal>::type Base;
     EIGEN_DENSE_PUBLIC_INTERFACE(Diagonal)
 
-    inline Diagonal(MatrixType& matrix, Index a_index = DiagIndex) : m_matrix(matrix), m_index(a_index) {}
+    inline Diagonal(MatrixType& matrix, Index index = DiagIndex) : m_matrix(matrix), m_index(index) {}
 
     EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Diagonal)
 
     inline Index rows() const
-    { return m_index.value()<0 ? (std::min<Index>)(m_matrix.cols(),m_matrix.rows()+m_index.value()) : (std::min<Index>)(m_matrix.rows(),m_matrix.cols()-m_index.value()); }
+    { return m_index.value()<0 ? (std::min)(m_matrix.cols(),m_matrix.rows()+m_index.value()) : (std::min)(m_matrix.rows(),m_matrix.cols()-m_index.value()); }
 
     inline Index cols() const { return 1; }
 
@@ -114,20 +113,20 @@ template<typename MatrixType, int _DiagIndex> class Diagonal
       return m_matrix.coeff(row+rowOffset(), row+colOffset());
     }
 
-    inline Scalar& coeffRef(Index idx)
+    inline Scalar& coeffRef(Index index)
     {
       EIGEN_STATIC_ASSERT_LVALUE(MatrixType)
-      return m_matrix.const_cast_derived().coeffRef(idx+rowOffset(), idx+colOffset());
+      return m_matrix.const_cast_derived().coeffRef(index+rowOffset(), index+colOffset());
     }
 
-    inline const Scalar& coeffRef(Index idx) const
+    inline const Scalar& coeffRef(Index index) const
     {
-      return m_matrix.const_cast_derived().coeffRef(idx+rowOffset(), idx+colOffset());
+      return m_matrix.const_cast_derived().coeffRef(index+rowOffset(), index+colOffset());
     }
 
-    inline CoeffReturnType coeff(Index idx) const
+    inline CoeffReturnType coeff(Index index) const
     {
-      return m_matrix.coeff(idx+rowOffset(), idx+colOffset());
+      return m_matrix.coeff(index+rowOffset(), index+colOffset());
     }
 
     const typename internal::remove_all<typename MatrixType::Nested>::type& 
@@ -143,7 +142,7 @@ template<typename MatrixType, int _DiagIndex> class Diagonal
 
   protected:
     typename MatrixType::Nested m_matrix;
-    const internal::variable_if_dynamicindex<Index, DiagIndex> m_index;
+    const internal::variable_if_dynamic<Index, DiagIndex> m_index;
 
   private:
     // some compilers may fail to optimize std::max etc in case of compile-time constants...
@@ -172,7 +171,7 @@ MatrixBase<Derived>::diagonal()
 
 /** This is the const version of diagonal(). */
 template<typename Derived>
-inline typename MatrixBase<Derived>::ConstDiagonalReturnType
+inline const typename MatrixBase<Derived>::ConstDiagonalReturnType
 MatrixBase<Derived>::diagonal() const
 {
   return ConstDiagonalReturnType(derived());
@@ -190,18 +189,18 @@ MatrixBase<Derived>::diagonal() const
   *
   * \sa MatrixBase::diagonal(), class Diagonal */
 template<typename Derived>
-inline typename MatrixBase<Derived>::DiagonalDynamicIndexReturnType
+inline typename MatrixBase<Derived>::template DiagonalIndexReturnType<Dynamic>::Type
 MatrixBase<Derived>::diagonal(Index index)
 {
-  return DiagonalDynamicIndexReturnType(derived(), index);
+  return typename DiagonalIndexReturnType<Dynamic>::Type(derived(), index);
 }
 
 /** This is the const version of diagonal(Index). */
 template<typename Derived>
-inline typename MatrixBase<Derived>::ConstDiagonalDynamicIndexReturnType
+inline typename MatrixBase<Derived>::template ConstDiagonalIndexReturnType<Dynamic>::Type
 MatrixBase<Derived>::diagonal(Index index) const
 {
-  return ConstDiagonalDynamicIndexReturnType(derived(), index);
+  return typename ConstDiagonalIndexReturnType<Dynamic>::Type(derived(), index);
 }
 
 /** \returns an expression of the \a DiagIndex-th sub or super diagonal of the matrix \c *this

Eigen/src/Core/DiagonalMatrix.h

@@ -44,10 +44,10 @@ class DiagonalBase : public EigenBase<Derived>
     template<typename DenseDerived>
     void evalTo(MatrixBase<DenseDerived> &other) const;
     template<typename DenseDerived>
-    inline void addTo(MatrixBase<DenseDerived> &other) const
+    void addTo(MatrixBase<DenseDerived> &other) const
     { other.diagonal() += diagonal(); }
     template<typename DenseDerived>
-    inline void subTo(MatrixBase<DenseDerived> &other) const
+    void subTo(MatrixBase<DenseDerived> &other) const
     { other.diagonal() -= diagonal(); }
 
     inline const DiagonalVectorType& diagonal() const { return derived().diagonal(); }
@@ -56,14 +56,9 @@ class DiagonalBase : public EigenBase<Derived>
     inline Index rows() const { return diagonal().size(); }
     inline Index cols() const { return diagonal().size(); }
 
-    /** \returns the diagonal matrix product of \c *this by the matrix \a matrix.
-      */
     template<typename MatrixDerived>
     const DiagonalProduct<MatrixDerived, Derived, OnTheLeft>
-    operator*(const MatrixBase<MatrixDerived> &matrix) const
-    {
-      return DiagonalProduct<MatrixDerived, Derived, OnTheLeft>(matrix.derived(), derived());
-    }
+    operator*(const MatrixBase<MatrixDerived> &matrix) const;
 
     inline const DiagonalWrapper<const CwiseUnaryOp<internal::scalar_inverse_op<Scalar>, const DiagonalVectorType> >
     inverse() const
@@ -98,7 +93,7 @@ class DiagonalBase : public EigenBase<Derived>
 
 template<typename Derived>
 template<typename DenseDerived>
-inline void DiagonalBase<Derived>::evalTo(MatrixBase<DenseDerived> &other) const
+void DiagonalBase<Derived>::evalTo(MatrixBase<DenseDerived> &other) const
 {
   other.setZero();
   other.diagonal() = diagonal();
@@ -255,7 +250,7 @@ class DiagonalWrapper
     #endif
 
     /** Constructor from expression of diagonal coefficients to wrap. */
-    inline DiagonalWrapper(DiagonalVectorType& a_diagonal) : m_diagonal(a_diagonal) {}
+    inline DiagonalWrapper(DiagonalVectorType& diagonal) : m_diagonal(diagonal) {}
 
     /** \returns a const reference to the wrapped expression of diagonal coefficients. */
     const DiagonalVectorType& diagonal() const { return m_diagonal; }
@@ -289,14 +284,13 @@ MatrixBase<Derived>::asDiagonal() const
   * \sa asDiagonal()
   */
 template<typename Derived>
-bool MatrixBase<Derived>::isDiagonal(const RealScalar& prec) const
+bool MatrixBase<Derived>::isDiagonal(RealScalar prec) const
 {
-  using std::abs;
   if(cols() != rows()) return false;
   RealScalar maxAbsOnDiagonal = static_cast<RealScalar>(-1);
   for(Index j = 0; j < cols(); ++j)
   {
-    RealScalar absOnDiagonal = abs(coeff(j,j));
+    RealScalar absOnDiagonal = internal::abs(coeff(j,j));
     if(absOnDiagonal > maxAbsOnDiagonal) maxAbsOnDiagonal = absOnDiagonal;
   }
   for(Index j = 0; j < cols(); ++j)

Eigen/src/Core/DiagonalProduct.h

@@ -26,17 +26,15 @@ struct traits<DiagonalProduct<MatrixType, DiagonalType, ProductOrder> >
     MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime,
 
     _StorageOrder = MatrixType::Flags & RowMajorBit ? RowMajor : ColMajor,
-    _ScalarAccessOnDiag =  !((int(_StorageOrder) == ColMajor && int(ProductOrder) == OnTheLeft)
-                          ||(int(_StorageOrder) == RowMajor && int(ProductOrder) == OnTheRight)),
+    _PacketOnDiag = !((int(_StorageOrder) == RowMajor && int(ProductOrder) == OnTheLeft)
+                    ||(int(_StorageOrder) == ColMajor && int(ProductOrder) == OnTheRight)),
     _SameTypes = is_same<typename MatrixType::Scalar, typename DiagonalType::Scalar>::value,
     // FIXME currently we need same types, but in the future the next rule should be the one
-    //_Vectorizable = bool(int(MatrixType::Flags)&PacketAccessBit) && ((!_PacketOnDiag) || (_SameTypes && bool(int(DiagonalType::DiagonalVectorType::Flags)&PacketAccessBit))),
-    _Vectorizable = bool(int(MatrixType::Flags)&PacketAccessBit) && _SameTypes && (_ScalarAccessOnDiag || (bool(int(DiagonalType::DiagonalVectorType::Flags)&PacketAccessBit))),
-    _LinearAccessMask = (RowsAtCompileTime==1 || ColsAtCompileTime==1) ? LinearAccessBit : 0,
+    //_Vectorizable = bool(int(MatrixType::Flags)&PacketAccessBit) && ((!_PacketOnDiag) || (_SameTypes && bool(int(DiagonalType::Flags)&PacketAccessBit))),
+    _Vectorizable = bool(int(MatrixType::Flags)&PacketAccessBit) && _SameTypes && ((!_PacketOnDiag) || (bool(int(DiagonalType::Flags)&PacketAccessBit))),
 
-    Flags = ((HereditaryBits|_LinearAccessMask|AlignedBit) & (unsigned int)(MatrixType::Flags)) | (_Vectorizable ? PacketAccessBit : 0),//(int(MatrixType::Flags)&int(DiagonalType::DiagonalVectorType::Flags)&AlignedBit),
-    Cost0 = EIGEN_ADD_COST(NumTraits<Scalar>::MulCost, MatrixType::CoeffReadCost),
-    CoeffReadCost = EIGEN_ADD_COST(Cost0,DiagonalType::DiagonalVectorType::CoeffReadCost)
+    Flags = (HereditaryBits & (unsigned int)(MatrixType::Flags)) | (_Vectorizable ? PacketAccessBit : 0),
+    CoeffReadCost = NumTraits<Scalar>::MulCost + MatrixType::CoeffReadCost + DiagonalType::DiagonalVectorType::CoeffReadCost
   };
 };
 }
@@ -56,21 +54,13 @@ class DiagonalProduct : internal::no_assignment_operator,
       eigen_assert(diagonal.diagonal().size() == (ProductOrder == OnTheLeft ? matrix.rows() : matrix.cols()));
     }
 
-    EIGEN_STRONG_INLINE Index rows() const { return m_matrix.rows(); }
-    EIGEN_STRONG_INLINE Index cols() const { return m_matrix.cols(); }
+    inline Index rows() const { return m_matrix.rows(); }
+    inline Index cols() const { return m_matrix.cols(); }
 
-    EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const
+    const Scalar coeff(Index row, Index col) const
     {
       return m_diagonal.diagonal().coeff(ProductOrder == OnTheLeft ? row : col) * m_matrix.coeff(row, col);
     }
-    
-    EIGEN_STRONG_INLINE const Scalar coeff(Index idx) const
-    {
-      enum {
-        StorageOrder = int(MatrixType::Flags) & RowMajorBit ? RowMajor : ColMajor
-      };
-      return coeff(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx);
-    }
 
     template<int LoadMode>
     EIGEN_STRONG_INLINE PacketScalar packet(Index row, Index col) const
@@ -79,19 +69,11 @@ class DiagonalProduct : internal::no_assignment_operator,
         StorageOrder = Flags & RowMajorBit ? RowMajor : ColMajor
       };
       const Index indexInDiagonalVector = ProductOrder == OnTheLeft ? row : col;
+
       return packet_impl<LoadMode>(row,col,indexInDiagonalVector,typename internal::conditional<
         ((int(StorageOrder) == RowMajor && int(ProductOrder) == OnTheLeft)
        ||(int(StorageOrder) == ColMajor && int(ProductOrder) == OnTheRight)), internal::true_type, internal::false_type>::type());
     }
-    
-    template<int LoadMode>
-    EIGEN_STRONG_INLINE PacketScalar packet(Index idx) const
-    {
-      enum {
-        StorageOrder = int(MatrixType::Flags) & RowMajorBit ? RowMajor : ColMajor
-      };
-      return packet<LoadMode>(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx);
-    }
 
   protected:
     template<int LoadMode>
@@ -106,7 +88,7 @@ class DiagonalProduct : internal::no_assignment_operator,
     {
       enum {
         InnerSize = (MatrixType::Flags & RowMajorBit) ? MatrixType::ColsAtCompileTime : MatrixType::RowsAtCompileTime,
-        DiagonalVectorPacketLoadMode = (LoadMode == Aligned && (((InnerSize%16) == 0) || (int(DiagonalType::DiagonalVectorType::Flags)&AlignedBit)==AlignedBit) ? Aligned : Unaligned)
+        DiagonalVectorPacketLoadMode = (LoadMode == Aligned && ((InnerSize%16) == 0)) ? Aligned : Unaligned
       };
       return internal::pmul(m_matrix.template packet<LoadMode>(row, col),
                      m_diagonal.diagonal().template packet<DiagonalVectorPacketLoadMode>(id));
@@ -121,9 +103,19 @@ class DiagonalProduct : internal::no_assignment_operator,
 template<typename Derived>
 template<typename DiagonalDerived>
 inline const DiagonalProduct<Derived, DiagonalDerived, OnTheRight>
-MatrixBase<Derived>::operator*(const DiagonalBase<DiagonalDerived> &a_diagonal) const
+MatrixBase<Derived>::operator*(const DiagonalBase<DiagonalDerived> &diagonal) const
 {
-  return DiagonalProduct<Derived, DiagonalDerived, OnTheRight>(derived(), a_diagonal.derived());
+  return DiagonalProduct<Derived, DiagonalDerived, OnTheRight>(derived(), diagonal.derived());
+}
+
+/** \returns the diagonal matrix product of \c *this by the matrix \a matrix.
+  */
+template<typename DiagonalDerived>
+template<typename MatrixDerived>
+inline const DiagonalProduct<MatrixDerived, DiagonalDerived, OnTheLeft>
+DiagonalBase<DiagonalDerived>::operator*(const MatrixBase<MatrixDerived> &matrix) const
+{
+  return DiagonalProduct<MatrixDerived, DiagonalDerived, OnTheLeft>(matrix.derived(), derived());
 }
 
 } // end namespace Eigen

Eigen/src/Core/Dot.h

@@ -59,7 +59,7 @@ struct dot_nocheck<T, U, true>
   */
 template<typename Derived>
 template<typename OtherDerived>
-inline typename internal::scalar_product_traits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
+typename internal::scalar_product_traits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
 MatrixBase<Derived>::dot(const MatrixBase<OtherDerived>& other) const
 {
   EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
@@ -112,7 +112,7 @@ MatrixBase<Derived>::eigen2_dot(const MatrixBase<OtherDerived>& other) const
 template<typename Derived>
 EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::squaredNorm() const
 {
-  return numext::real((*this).cwiseAbs2().sum());
+  return internal::real((*this).cwiseAbs2().sum());
 }
 
 /** \returns, for vectors, the \em l2 norm of \c *this, and for matrices the Frobenius norm.
@@ -124,8 +124,7 @@ EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scala
 template<typename Derived>
 inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
 {
-  using std::sqrt;
-  return sqrt(squaredNorm());
+  return internal::sqrt(squaredNorm());
 }
 
 /** \returns an expression of the quotient of *this by its own norm.
@@ -166,7 +165,6 @@ struct lpNorm_selector
   typedef typename NumTraits<typename traits<Derived>::Scalar>::Real RealScalar;
   static inline RealScalar run(const MatrixBase<Derived>& m)
   {
-    using std::pow;
     return pow(m.cwiseAbs().array().pow(p).sum(), RealScalar(1)/p);
   }
 };
@@ -225,11 +223,11 @@ MatrixBase<Derived>::lpNorm() const
 template<typename Derived>
 template<typename OtherDerived>
 bool MatrixBase<Derived>::isOrthogonal
-(const MatrixBase<OtherDerived>& other, const RealScalar& prec) const
+(const MatrixBase<OtherDerived>& other, RealScalar prec) const
 {
   typename internal::nested<Derived,2>::type nested(derived());
   typename internal::nested<OtherDerived,2>::type otherNested(other.derived());
-  return numext::abs2(nested.dot(otherNested)) <= prec * prec * nested.squaredNorm() * otherNested.squaredNorm();
+  return internal::abs2(nested.dot(otherNested)) <= prec * prec * nested.squaredNorm() * otherNested.squaredNorm();
 }
 
 /** \returns true if *this is approximately an unitary matrix,
@@ -244,7 +242,7 @@ bool MatrixBase<Derived>::isOrthogonal
   * Output: \verbinclude MatrixBase_isUnitary.out
   */
 template<typename Derived>
-bool MatrixBase<Derived>::isUnitary(const RealScalar& prec) const
+bool MatrixBase<Derived>::isUnitary(RealScalar prec) const
 {
   typename Derived::Nested nested(derived());
   for(Index i = 0; i < cols(); ++i)

Eigen/src/Core/EigenBase.h

@@ -126,6 +126,35 @@ Derived& DenseBase<Derived>::operator-=(const EigenBase<OtherDerived> &other)
   return derived();
 }
 
+/** replaces \c *this by \c *this * \a other.
+  *
+  * \returns a reference to \c *this
+  */
+template<typename Derived>
+template<typename OtherDerived>
+inline Derived&
+MatrixBase<Derived>::operator*=(const EigenBase<OtherDerived> &other)
+{
+  other.derived().applyThisOnTheRight(derived());
+  return derived();
+}
+
+/** replaces \c *this by \c *this * \a other. It is equivalent to MatrixBase::operator*=() */
+template<typename Derived>
+template<typename OtherDerived>
+inline void MatrixBase<Derived>::applyOnTheRight(const EigenBase<OtherDerived> &other)
+{
+  other.derived().applyThisOnTheRight(derived());
+}
+
+/** replaces \c *this by \c *this * \a other. */
+template<typename Derived>
+template<typename OtherDerived>
+inline void MatrixBase<Derived>::applyOnTheLeft(const EigenBase<OtherDerived> &other)
+{
+  other.derived().applyThisOnTheLeft(derived());
+}
+
 } // end namespace Eigen
 
 #endif // EIGEN_EIGENBASE_H

Eigen/src/Core/Functors.h

@@ -154,7 +154,6 @@ template<typename Scalar> struct scalar_hypot_op {
   {
     using std::max;
     using std::min;
-    using std::sqrt;
     Scalar p = (max)(_x, _y);
     Scalar q = (min)(_x, _y);
     Scalar qp = q/p;
@@ -171,7 +170,7 @@ struct functor_traits<scalar_hypot_op<Scalar> > {
   */
 template<typename Scalar, typename OtherScalar> struct scalar_binary_pow_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_binary_pow_op)
-  inline Scalar operator() (const Scalar& a, const OtherScalar& b) const { return numext::pow(a, b); }
+  inline Scalar operator() (const Scalar& a, const OtherScalar& b) const { return internal::pow(a, b); }
 };
 template<typename Scalar, typename OtherScalar>
 struct functor_traits<scalar_binary_pow_op<Scalar,OtherScalar> > {
@@ -205,28 +204,21 @@ struct functor_traits<scalar_difference_op<Scalar> > {
   *
   * \sa class CwiseBinaryOp, Cwise::operator/()
   */
-template<typename LhsScalar,typename RhsScalar> struct scalar_quotient_op {
-  enum {
-    // TODO vectorize mixed product
-    Vectorizable = is_same<LhsScalar,RhsScalar>::value && packet_traits<LhsScalar>::HasDiv && packet_traits<RhsScalar>::HasDiv
-  };
-  typedef typename scalar_product_traits<LhsScalar,RhsScalar>::ReturnType result_type;
+template<typename Scalar> struct scalar_quotient_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_quotient_op)
-  EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a / b; }
+  EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a, const Scalar& b) const { return a / b; }
   template<typename Packet>
   EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const
   { return internal::pdiv(a,b); }
 };
-template<typename LhsScalar,typename RhsScalar>
-struct functor_traits<scalar_quotient_op<LhsScalar,RhsScalar> > {
+template<typename Scalar>
+struct functor_traits<scalar_quotient_op<Scalar> > {
   enum {
-    Cost = (NumTraits<LhsScalar>::MulCost + NumTraits<RhsScalar>::MulCost), // rough estimate!
-    PacketAccess = scalar_quotient_op<LhsScalar,RhsScalar>::Vectorizable
+    Cost = 2 * NumTraits<Scalar>::MulCost,
+    PacketAccess = packet_traits<Scalar>::HasDiv
   };
 };
 
-
-
 /** \internal
   * \brief Template functor to compute the and of two booleans
   *
@@ -259,47 +251,6 @@ template<> struct functor_traits<scalar_boolean_or_op> {
   };
 };
 
-/** \internal
-  * \brief Template functors for comparison of two scalars
-  * \todo Implement packet-comparisons
-  */
-template<typename Scalar, ComparisonName cmp> struct scalar_cmp_op;
-
-template<typename Scalar, ComparisonName cmp>
-struct functor_traits<scalar_cmp_op<Scalar, cmp> > {
-  enum {
-    Cost = NumTraits<Scalar>::AddCost,
-    PacketAccess = false
-  };
-};
-
-template<ComparisonName Cmp, typename Scalar>
-struct result_of<scalar_cmp_op<Scalar, Cmp>(Scalar,Scalar)> {
-  typedef bool type;
-};
-
-
-template<typename Scalar> struct scalar_cmp_op<Scalar, cmp_EQ> {
-  EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op)
-  EIGEN_STRONG_INLINE bool operator()(const Scalar& a, const Scalar& b) const {return a==b;}
-};
-template<typename Scalar> struct scalar_cmp_op<Scalar, cmp_LT> {
-  EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op)
-  EIGEN_STRONG_INLINE bool operator()(const Scalar& a, const Scalar& b) const {return a<b;}
-};
-template<typename Scalar> struct scalar_cmp_op<Scalar, cmp_LE> {
-  EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op)
-  EIGEN_STRONG_INLINE bool operator()(const Scalar& a, const Scalar& b) const {return a<=b;}
-};
-template<typename Scalar> struct scalar_cmp_op<Scalar, cmp_UNORD> {
-  EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op)
-  EIGEN_STRONG_INLINE bool operator()(const Scalar& a, const Scalar& b) const {return !(a<=b || b<=a);}
-};
-template<typename Scalar> struct scalar_cmp_op<Scalar, cmp_NEQ> {
-  EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op)
-  EIGEN_STRONG_INLINE bool operator()(const Scalar& a, const Scalar& b) const {return a!=b;}
-};
-
 // unary functors:
 
 /** \internal
@@ -329,7 +280,7 @@ struct functor_traits<scalar_opposite_op<Scalar> >
 template<typename Scalar> struct scalar_abs_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_abs_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { using std::abs; return abs(a); }
+  EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return internal::abs(a); }
   template<typename Packet>
   EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const
   { return internal::pabs(a); }
@@ -351,7 +302,7 @@ struct functor_traits<scalar_abs_op<Scalar> >
 template<typename Scalar> struct scalar_abs2_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_abs2_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return numext::abs2(a); }
+  EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return internal::abs2(a); }
   template<typename Packet>
   EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const
   { return internal::pmul(a,a); }
@@ -367,7 +318,7 @@ struct functor_traits<scalar_abs2_op<Scalar> >
   */
 template<typename Scalar> struct scalar_conjugate_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_conjugate_op)
-  EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { using numext::conj; return conj(a); }
+  EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return internal::conj(a); }
   template<typename Packet>
   EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const { return internal::pconj(a); }
 };
@@ -404,7 +355,7 @@ template<typename Scalar>
 struct scalar_real_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_real_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return numext::real(a); }
+  EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return internal::real(a); }
 };
 template<typename Scalar>
 struct functor_traits<scalar_real_op<Scalar> >
@@ -419,7 +370,7 @@ template<typename Scalar>
 struct scalar_imag_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_imag_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return numext::imag(a); }
+  EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return internal::imag(a); }
 };
 template<typename Scalar>
 struct functor_traits<scalar_imag_op<Scalar> >
@@ -434,7 +385,7 @@ template<typename Scalar>
 struct scalar_real_ref_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_real_ref_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return numext::real_ref(*const_cast<Scalar*>(&a)); }
+  EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return internal::real_ref(*const_cast<Scalar*>(&a)); }
 };
 template<typename Scalar>
 struct functor_traits<scalar_real_ref_op<Scalar> >
@@ -449,7 +400,7 @@ template<typename Scalar>
 struct scalar_imag_ref_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_imag_ref_op)
   typedef typename NumTraits<Scalar>::Real result_type;
-  EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return numext::imag_ref(*const_cast<Scalar*>(&a)); }
+  EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return internal::imag_ref(*const_cast<Scalar*>(&a)); }
 };
 template<typename Scalar>
 struct functor_traits<scalar_imag_ref_op<Scalar> >
@@ -463,7 +414,7 @@ struct functor_traits<scalar_imag_ref_op<Scalar> >
   */
 template<typename Scalar> struct scalar_exp_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_exp_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::exp; return exp(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::exp(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::pexp(a); }
 };
@@ -479,7 +430,7 @@ struct functor_traits<scalar_exp_op<Scalar> >
   */
 template<typename Scalar> struct scalar_log_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_log_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::log; return log(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::log(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::plog(a); }
 };
@@ -592,7 +543,7 @@ struct linspaced_op_impl<Scalar,false>
 {
   typedef typename packet_traits<Scalar>::type Packet;
 
-  linspaced_op_impl(const Scalar& low, const Scalar& step) :
+  linspaced_op_impl(Scalar low, Scalar step) :
   m_low(low), m_step(step),
   m_packetStep(pset1<Packet>(packet_traits<Scalar>::size*step)),
   m_base(padd(pset1<Packet>(low), pmul(pset1<Packet>(step),plset<Scalar>(-packet_traits<Scalar>::size)))) {}
@@ -601,7 +552,7 @@ struct linspaced_op_impl<Scalar,false>
   EIGEN_STRONG_INLINE const Scalar operator() (Index i) const 
   { 
     m_base = padd(m_base, pset1<Packet>(m_step));
-    return m_low+Scalar(i)*m_step; 
+    return m_low+i*m_step; 
   }
 
   template<typename Index>
@@ -621,7 +572,7 @@ struct linspaced_op_impl<Scalar,true>
 {
   typedef typename packet_traits<Scalar>::type Packet;
 
-  linspaced_op_impl(const Scalar& low, const Scalar& step) :
+  linspaced_op_impl(Scalar low, Scalar step) :
   m_low(low), m_step(step),
   m_lowPacket(pset1<Packet>(m_low)), m_stepPacket(pset1<Packet>(m_step)), m_interPacket(plset<Scalar>(0)) {}
 
@@ -630,7 +581,7 @@ struct linspaced_op_impl<Scalar,true>
 
   template<typename Index>
   EIGEN_STRONG_INLINE const Packet packetOp(Index i) const
-  { return internal::padd(m_lowPacket, pmul(m_stepPacket, padd(pset1<Packet>(Scalar(i)),m_interPacket))); }
+  { return internal::padd(m_lowPacket, pmul(m_stepPacket, padd(pset1<Packet>(i),m_interPacket))); }
 
   const Scalar m_low;
   const Scalar m_step;
@@ -650,7 +601,7 @@ template <typename Scalar, bool RandomAccess> struct functor_traits< linspaced_o
 template <typename Scalar, bool RandomAccess> struct linspaced_op
 {
   typedef typename packet_traits<Scalar>::type Packet;
-  linspaced_op(const Scalar& low, const Scalar& high, DenseIndex num_steps) : impl((num_steps==1 ? high : low), (num_steps==1 ? Scalar() : (high-low)/Scalar(num_steps-1))) {}
+  linspaced_op(Scalar low, Scalar high, int num_steps) : impl((num_steps==1 ? high : low), (num_steps==1 ? Scalar() : (high-low)/(num_steps-1))) {}
 
   template<typename Index>
   EIGEN_STRONG_INLINE const Scalar operator() (Index i) const { return impl(i); }
@@ -689,14 +640,12 @@ template <typename Scalar, bool RandomAccess> struct linspaced_op
 template<typename Functor> struct functor_has_linear_access { enum { ret = 1 }; };
 template<typename Scalar> struct functor_has_linear_access<scalar_identity_op<Scalar> > { enum { ret = 0 }; };
 
-// In Eigen, any binary op (Product, CwiseBinaryOp) require the Lhs and Rhs to have the same scalar type, except for multiplication
-// where the mixing of different types is handled by scalar_product_traits
-// In particular, real * complex<real> is allowed.
+// in CwiseBinaryOp, we require the Lhs and Rhs to have the same scalar type, except for multiplication
+// where we only require them to have the same _real_ scalar type so one may multiply, say, float by complex<float>.
 // FIXME move this to functor_traits adding a functor_default
-template<typename Functor> struct functor_is_product_like { enum { ret = 0 }; };
-template<typename LhsScalar,typename RhsScalar> struct functor_is_product_like<scalar_product_op<LhsScalar,RhsScalar> > { enum { ret = 1 }; };
-template<typename LhsScalar,typename RhsScalar> struct functor_is_product_like<scalar_conj_product_op<LhsScalar,RhsScalar> > { enum { ret = 1 }; };
-template<typename LhsScalar,typename RhsScalar> struct functor_is_product_like<scalar_quotient_op<LhsScalar,RhsScalar> > { enum { ret = 1 }; };
+template<typename Functor> struct functor_allows_mixing_real_and_complex { enum { ret = 0 }; };
+template<typename LhsScalar,typename RhsScalar> struct functor_allows_mixing_real_and_complex<scalar_product_op<LhsScalar,RhsScalar> > { enum { ret = 1 }; };
+template<typename LhsScalar,typename RhsScalar> struct functor_allows_mixing_real_and_complex<scalar_conj_product_op<LhsScalar,RhsScalar> > { enum { ret = 1 }; };
 
 
 /** \internal
@@ -725,7 +674,7 @@ struct functor_traits<scalar_add_op<Scalar> >
   */
 template<typename Scalar> struct scalar_sqrt_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_sqrt_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::sqrt; return sqrt(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::sqrt(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::psqrt(a); }
 };
@@ -743,7 +692,7 @@ struct functor_traits<scalar_sqrt_op<Scalar> >
   */
 template<typename Scalar> struct scalar_cos_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_cos_op)
-  inline Scalar operator() (const Scalar& a) const { using std::cos; return cos(a); }
+  inline Scalar operator() (const Scalar& a) const { return internal::cos(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::pcos(a); }
 };
@@ -762,7 +711,7 @@ struct functor_traits<scalar_cos_op<Scalar> >
   */
 template<typename Scalar> struct scalar_sin_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_sin_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::sin; return sin(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::sin(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::psin(a); }
 };
@@ -782,7 +731,7 @@ struct functor_traits<scalar_sin_op<Scalar> >
   */
 template<typename Scalar> struct scalar_tan_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_tan_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::tan; return tan(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::tan(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::ptan(a); }
 };
@@ -801,7 +750,7 @@ struct functor_traits<scalar_tan_op<Scalar> >
   */
 template<typename Scalar> struct scalar_acos_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_acos_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::acos; return acos(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::acos(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::pacos(a); }
 };
@@ -820,7 +769,7 @@ struct functor_traits<scalar_acos_op<Scalar> >
   */
 template<typename Scalar> struct scalar_asin_op {
   EIGEN_EMPTY_STRUCT_CTOR(scalar_asin_op)
-  inline const Scalar operator() (const Scalar& a) const { using std::asin; return asin(a); }
+  inline const Scalar operator() (const Scalar& a) const { return internal::asin(a); }
   typedef typename packet_traits<Scalar>::type Packet;
   inline Packet packetOp(const Packet& a) const { return internal::pasin(a); }
 };
@@ -842,7 +791,7 @@ struct scalar_pow_op {
   // FIXME default copy constructors seems bugged with std::complex<>
   inline scalar_pow_op(const scalar_pow_op& other) : m_exponent(other.m_exponent) { }
   inline scalar_pow_op(const Scalar& exponent) : m_exponent(exponent) {}
-  inline Scalar operator() (const Scalar& a) const { return numext::pow(a, m_exponent); }
+  inline Scalar operator() (const Scalar& a) const { return internal::pow(a, m_exponent); }
   const Scalar m_exponent;
 };
 template<typename Scalar>
@@ -969,8 +918,6 @@ template<typename T>
 struct functor_traits<std::not_equal_to<T> >
 { enum { Cost = 1, PacketAccess = false }; };
 
-#if(__cplusplus < 201103L)
-// std::binder* are deprecated since c++11 and will be removed in c++17
 template<typename T>
 struct functor_traits<std::binder2nd<T> >
 { enum { Cost = functor_traits<T>::Cost, PacketAccess = false }; };
@@ -978,7 +925,6 @@ struct functor_traits<std::binder2nd<T> >
 template<typename T>
 struct functor_traits<std::binder1st<T> >
 { enum { Cost = functor_traits<T>::Cost, PacketAccess = false }; };
-#endif
 
 template<typename T>
 struct functor_traits<std::unary_negate<T> >

Eigen/src/Core/Fuzzy.h

@@ -19,7 +19,7 @@ namespace internal
 template<typename Derived, typename OtherDerived, bool is_integer = NumTraits<typename Derived::Scalar>::IsInteger>
 struct isApprox_selector
 {
-  static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec)
+  static bool run(const Derived& x, const OtherDerived& y, typename Derived::RealScalar prec)
   {
     using std::min;
     typename internal::nested<Derived,2>::type nested(x);
@@ -31,7 +31,7 @@ struct isApprox_selector
 template<typename Derived, typename OtherDerived>
 struct isApprox_selector<Derived, OtherDerived, true>
 {
-  static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar&)
+  static bool run(const Derived& x, const OtherDerived& y, typename Derived::RealScalar)
   {
     return x.matrix() == y.matrix();
   }
@@ -40,16 +40,16 @@ struct isApprox_selector<Derived, OtherDerived, true>
 template<typename Derived, typename OtherDerived, bool is_integer = NumTraits<typename Derived::Scalar>::IsInteger>
 struct isMuchSmallerThan_object_selector
 {
-  static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec)
+  static bool run(const Derived& x, const OtherDerived& y, typename Derived::RealScalar prec)
   {
-    return x.cwiseAbs2().sum() <= numext::abs2(prec) * y.cwiseAbs2().sum();
+    return x.cwiseAbs2().sum() <= abs2(prec) * y.cwiseAbs2().sum();
   }
 };
 
 template<typename Derived, typename OtherDerived>
 struct isMuchSmallerThan_object_selector<Derived, OtherDerived, true>
 {
-  static bool run(const Derived& x, const OtherDerived&, const typename Derived::RealScalar&)
+  static bool run(const Derived& x, const OtherDerived&, typename Derived::RealScalar)
   {
     return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix();
   }
@@ -58,16 +58,16 @@ struct isMuchSmallerThan_object_selector<Derived, OtherDerived, true>
 template<typename Derived, bool is_integer = NumTraits<typename Derived::Scalar>::IsInteger>
 struct isMuchSmallerThan_scalar_selector
 {
-  static bool run(const Derived& x, const typename Derived::RealScalar& y, const typename Derived::RealScalar& prec)
+  static bool run(const Derived& x, const typename Derived::RealScalar& y, typename Derived::RealScalar prec)
   {
-    return x.cwiseAbs2().sum() <= numext::abs2(prec * y);
+    return x.cwiseAbs2().sum() <= abs2(prec * y);
   }
 };
 
 template<typename Derived>
 struct isMuchSmallerThan_scalar_selector<Derived, true>
 {
-  static bool run(const Derived& x, const typename Derived::RealScalar&, const typename Derived::RealScalar&)
+  static bool run(const Derived& x, const typename Derived::RealScalar&, typename Derived::RealScalar)
   {
     return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix();
   }
@@ -97,7 +97,7 @@ template<typename Derived>
 template<typename OtherDerived>
 bool DenseBase<Derived>::isApprox(
   const DenseBase<OtherDerived>& other,
-  const RealScalar& prec
+  RealScalar prec
 ) const
 {
   return internal::isApprox_selector<Derived, OtherDerived>::run(derived(), other.derived(), prec);
@@ -119,7 +119,7 @@ bool DenseBase<Derived>::isApprox(
 template<typename Derived>
 bool DenseBase<Derived>::isMuchSmallerThan(
   const typename NumTraits<Scalar>::Real& other,
-  const RealScalar& prec
+  RealScalar prec
 ) const
 {
   return internal::isMuchSmallerThan_scalar_selector<Derived>::run(derived(), other, prec);
@@ -139,7 +139,7 @@ template<typename Derived>
 template<typename OtherDerived>
 bool DenseBase<Derived>::isMuchSmallerThan(
   const DenseBase<OtherDerived>& other,
-  const RealScalar& prec
+  RealScalar prec
 ) const
 {
   return internal::isMuchSmallerThan_object_selector<Derived, OtherDerived>::run(derived(), other.derived(), prec);

Eigen/src/Core/GeneralProduct.h

@@ -205,6 +205,9 @@ class GeneralProduct<Lhs, Rhs, InnerProduct>
   public:
     GeneralProduct(const Lhs& lhs, const Rhs& rhs)
     {
+      EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::RealScalar, typename Rhs::RealScalar>::value),
+        YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
+
       Base::coeffRef(0,0) = (lhs.transpose().cwiseProduct(rhs)).sum();
     }
 
@@ -219,29 +222,7 @@ class GeneralProduct<Lhs, Rhs, InnerProduct>
 ***********************************************************************/
 
 namespace internal {
-
-// Column major
-template<typename ProductType, typename Dest, typename Func>
-EIGEN_DONT_INLINE void outer_product_selector_run(const ProductType& prod, Dest& dest, const Func& func, const false_type&)
-{
-  typedef typename Dest::Index Index;
-  // FIXME make sure lhs is sequentially stored
-  // FIXME not very good if rhs is real and lhs complex while alpha is real too
-  const Index cols = dest.cols();
-  for (Index j=0; j<cols; ++j)
-    func(dest.col(j), prod.rhs().coeff(0,j) * prod.lhs());
-}
-
-// Row major
-template<typename ProductType, typename Dest, typename Func>
-EIGEN_DONT_INLINE void outer_product_selector_run(const ProductType& prod, Dest& dest, const Func& func, const true_type&) {
-  typedef typename Dest::Index Index;
-  // FIXME make sure rhs is sequentially stored
-  // FIXME not very good if lhs is real and rhs complex while alpha is real too
-  const Index rows = dest.rows();
-  for (Index i=0; i<rows; ++i)
-    func(dest.row(i), prod.lhs().coeff(i,0) * prod.rhs());
-}
+template<int StorageOrder> struct outer_product_selector;
 
 template<typename Lhs, typename Rhs>
 struct traits<GeneralProduct<Lhs,Rhs,OuterProduct> >
@@ -254,47 +235,49 @@ template<typename Lhs, typename Rhs>
 class GeneralProduct<Lhs, Rhs, OuterProduct>
   : public ProductBase<GeneralProduct<Lhs,Rhs,OuterProduct>, Lhs, Rhs>
 {
-    template<typename T> struct is_row_major : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {};
-    
   public:
     EIGEN_PRODUCT_PUBLIC_INTERFACE(GeneralProduct)
 
     GeneralProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs)
     {
-    }
-    
-    struct set  { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived()  = src; } };
-    struct add  { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() += src; } };
-    struct sub  { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() -= src; } };
-    struct adds {
-      Scalar m_scale;
-      adds(const Scalar& s) : m_scale(s) {}
-      template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const {
-        dst.const_cast_derived() += m_scale * src;
-      }
-    };
-    
-    template<typename Dest>
-    inline void evalTo(Dest& dest) const {
-      internal::outer_product_selector_run(*this, dest, set(), is_row_major<Dest>());
-    }
-    
-    template<typename Dest>
-    inline void addTo(Dest& dest) const {
-      internal::outer_product_selector_run(*this, dest, add(), is_row_major<Dest>());
+      EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::RealScalar, typename Rhs::RealScalar>::value),
+        YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
     }
 
-    template<typename Dest>
-    inline void subTo(Dest& dest) const {
-      internal::outer_product_selector_run(*this, dest, sub(), is_row_major<Dest>());
-    }
-
-    template<typename Dest> void scaleAndAddTo(Dest& dest, const Scalar& alpha) const
+    template<typename Dest> void scaleAndAddTo(Dest& dest, Scalar alpha) const
     {
-      internal::outer_product_selector_run(*this, dest, adds(alpha), is_row_major<Dest>());
+      internal::outer_product_selector<(int(Dest::Flags)&RowMajorBit) ? RowMajor : ColMajor>::run(*this, dest, alpha);
     }
 };
 
+namespace internal {
+
+template<> struct outer_product_selector<ColMajor> {
+  template<typename ProductType, typename Dest>
+  static EIGEN_DONT_INLINE void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha) {
+    typedef typename Dest::Index Index;
+    // FIXME make sure lhs is sequentially stored
+    // FIXME not very good if rhs is real and lhs complex while alpha is real too
+    const Index cols = dest.cols();
+    for (Index j=0; j<cols; ++j)
+      dest.col(j) += (alpha * prod.rhs().coeff(j)) * prod.lhs();
+  }
+};
+
+template<> struct outer_product_selector<RowMajor> {
+  template<typename ProductType, typename Dest>
+  static EIGEN_DONT_INLINE void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha) {
+    typedef typename Dest::Index Index;
+    // FIXME make sure rhs is sequentially stored
+    // FIXME not very good if lhs is real and rhs complex while alpha is real too
+    const Index rows = dest.rows();
+    for (Index i=0; i<rows; ++i)
+      dest.row(i) += (alpha * prod.lhs().coeff(i)) * prod.rhs();
+  }
+};
+
+} // end namespace internal
+
 /***********************************************************************
 *  Implementation of General Matrix Vector Product
 ***********************************************************************/
@@ -328,7 +311,7 @@ class GeneralProduct<Lhs, Rhs, GemvProduct>
     typedef typename Lhs::Scalar LhsScalar;
     typedef typename Rhs::Scalar RhsScalar;
 
-    GeneralProduct(const Lhs& a_lhs, const Rhs& a_rhs) : Base(a_lhs,a_rhs)
+    GeneralProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs)
     {
 //       EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::Scalar, typename Rhs::Scalar>::value),
 //         YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
@@ -337,7 +320,7 @@ class GeneralProduct<Lhs, Rhs, GemvProduct>
     enum { Side = Lhs::IsVectorAtCompileTime ? OnTheLeft : OnTheRight };
     typedef typename internal::conditional<int(Side)==OnTheRight,_LhsNested,_RhsNested>::type MatrixType;
 
-    template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
+    template<typename Dest> void scaleAndAddTo(Dest& dst, Scalar alpha) const
     {
       eigen_assert(m_lhs.rows() == dst.rows() && m_rhs.cols() == dst.cols());
       internal::gemv_selector<Side,(int(MatrixType::Flags)&RowMajorBit) ? RowMajor : ColMajor,
@@ -352,7 +335,7 @@ template<int StorageOrder, bool BlasCompatible>
 struct gemv_selector<OnTheLeft,StorageOrder,BlasCompatible>
 {
   template<typename ProductType, typename Dest>
-  static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
+  static void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha)
   {
     Transpose<Dest> destT(dest);
     enum { OtherStorageOrder = StorageOrder == RowMajor ? ColMajor : RowMajor };
@@ -401,7 +384,7 @@ struct gemv_static_vector_if<Scalar,Size,MaxSize,true>
 template<> struct gemv_selector<OnTheRight,ColMajor,true>
 {
   template<typename ProductType, typename Dest>
-  static inline void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
+  static inline void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha)
   {
     typedef typename ProductType::Index Index;
     typedef typename ProductType::LhsScalar   LhsScalar;
@@ -420,20 +403,17 @@ template<> struct gemv_selector<OnTheRight,ColMajor,true>
     ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs())
                                   * RhsBlasTraits::extractScalarFactor(prod.rhs());
 
-    // make sure Dest is a compile-time vector type (bug 1166)
-    typedef typename conditional<Dest::IsVectorAtCompileTime, Dest, typename Dest::ColXpr>::type ActualDest;
-
     enum {
       // FIXME find a way to allow an inner stride on the result if packet_traits<Scalar>::size==1
       // on, the other hand it is good for the cache to pack the vector anyways...
-      EvalToDestAtCompileTime = (ActualDest::InnerStrideAtCompileTime==1),
+      EvalToDestAtCompileTime = Dest::InnerStrideAtCompileTime==1,
       ComplexByReal = (NumTraits<LhsScalar>::IsComplex) && (!NumTraits<RhsScalar>::IsComplex),
-      MightCannotUseDest = (ActualDest::InnerStrideAtCompileTime!=1) || ComplexByReal
+      MightCannotUseDest = (Dest::InnerStrideAtCompileTime!=1) || ComplexByReal
     };
 
-    gemv_static_vector_if<ResScalar,ActualDest::SizeAtCompileTime,ActualDest::MaxSizeAtCompileTime,MightCannotUseDest> static_dest;
+    gemv_static_vector_if<ResScalar,Dest::SizeAtCompileTime,Dest::MaxSizeAtCompileTime,MightCannotUseDest> static_dest;
 
-    bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0));
+    bool alphaIsCompatible = (!ComplexByReal) || (imag(actualAlpha)==RealScalar(0));
     bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible;
     
     RhsScalar compatibleAlpha = get_factor<ResScalar,RhsScalar>::run(actualAlpha);
@@ -477,7 +457,7 @@ template<> struct gemv_selector<OnTheRight,ColMajor,true>
 template<> struct gemv_selector<OnTheRight,RowMajor,true>
 {
   template<typename ProductType, typename Dest>
-  static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
+  static void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha)
   {
     typedef typename ProductType::LhsScalar LhsScalar;
     typedef typename ProductType::RhsScalar RhsScalar;
@@ -520,7 +500,7 @@ template<> struct gemv_selector<OnTheRight,RowMajor,true>
         actualLhs.rows(), actualLhs.cols(),
         actualLhs.data(), actualLhs.outerStride(),
         actualRhsPtr, 1,
-        dest.data(), dest.col(0).innerStride(), //NOTE  if dest is not a vector at compile-time, then dest.innerStride() might be wrong. (bug 1166)
+        dest.data(), dest.innerStride(),
         actualAlpha);
   }
 };
@@ -528,7 +508,7 @@ template<> struct gemv_selector<OnTheRight,RowMajor,true>
 template<> struct gemv_selector<OnTheRight,ColMajor,false>
 {
   template<typename ProductType, typename Dest>
-  static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
+  static void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha)
   {
     typedef typename Dest::Index Index;
     // TODO makes sure dest is sequentially stored in memory, otherwise use a temp
@@ -541,7 +521,7 @@ template<> struct gemv_selector<OnTheRight,ColMajor,false>
 template<> struct gemv_selector<OnTheRight,RowMajor,false>
 {
   template<typename ProductType, typename Dest>
-  static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
+  static void run(const ProductType& prod, Dest& dest, typename ProductType::Scalar alpha)
   {
     typedef typename Dest::Index Index;
     // TODO makes sure rhs is sequentially stored in memory, otherwise use a temp

Eigen/src/Core/GenericPacketMath.h

@@ -106,7 +106,7 @@ pnegate(const Packet& a) { return -a; }
 
 /** \internal \returns conj(a) (coeff-wise) */
 template<typename Packet> inline Packet
-pconj(const Packet& a) { return numext::conj(a); }
+pconj(const Packet& a) { return conj(a); }
 
 /** \internal \returns a * b (coeff-wise) */
 template<typename Packet> inline Packet
@@ -130,7 +130,7 @@ pmax(const Packet& a,
 
 /** \internal \returns the absolute value of \a a */
 template<typename Packet> inline Packet
-pabs(const Packet& a) { using std::abs; return abs(a); }
+pabs(const Packet& a) { return abs(a); }
 
 /** \internal \returns the bitwise and of \a a and \a b */
 template<typename Packet> inline Packet
@@ -156,11 +156,7 @@ pload(const typename unpacket_traits<Packet>::type* from) { return *from; }
 template<typename Packet> inline Packet
 ploadu(const typename unpacket_traits<Packet>::type* from) { return *from; }
 
-/** \internal \returns a packet with elements of \a *from duplicated.
-  * For instance, for a packet of 8 elements, 4 scalar will be read from \a *from and
-  * duplicated to form: {from[0],from[0],from[1],from[1],,from[2],from[2],,from[3],from[3]}
-  * Currently, this function is only used for scalar * complex products.
- */
+/** \internal \returns a packet with elements of \a *from duplicated, e.g.: (from[0],from[0],from[1],from[1]) */
 template<typename Packet> inline Packet
 ploaddup(const typename unpacket_traits<Packet>::type* from) { return *from; }
 
@@ -183,8 +179,8 @@ template<typename Scalar, typename Packet> inline void pstoreu(Scalar* to, const
 /** \internal tries to do cache prefetching of \a addr */
 template<typename Scalar> inline void prefetch(const Scalar* addr)
 {
-#if (!EIGEN_COMP_MSVC) && (EIGEN_COMP_GNUC || EIGEN_COMP_CLANG || EIGEN_COMP_ICC)
-  __builtin_prefetch(addr);
+#if !defined(_MSC_VER)
+__builtin_prefetch(addr);
 #endif
 }
 
@@ -219,12 +215,7 @@ template<typename Packet> inline Packet preverse(const Packet& a)
 
 /** \internal \returns \a a with real and imaginary part flipped (for complex type only) */
 template<typename Packet> inline Packet pcplxflip(const Packet& a)
-{
-  // FIXME: uncomment the following in case we drop the internal imag and real functions.
-//   using std::imag;
-//   using std::real;
-  return Packet(imag(a),real(a));
-}
+{ return Packet(imag(a),real(a)); }
 
 /**************************
 * Special math functions
@@ -232,35 +223,35 @@ template<typename Packet> inline Packet pcplxflip(const Packet& a)
 
 /** \internal \returns the sine of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet psin(const Packet& a) { using std::sin; return sin(a); }
+Packet psin(const Packet& a) { return sin(a); }
 
 /** \internal \returns the cosine of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet pcos(const Packet& a) { using std::cos; return cos(a); }
+Packet pcos(const Packet& a) { return cos(a); }
 
 /** \internal \returns the tan of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet ptan(const Packet& a) { using std::tan; return tan(a); }
+Packet ptan(const Packet& a) { return tan(a); }
 
 /** \internal \returns the arc sine of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet pasin(const Packet& a) { using std::asin; return asin(a); }
+Packet pasin(const Packet& a) { return asin(a); }
 
 /** \internal \returns the arc cosine of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet pacos(const Packet& a) { using std::acos; return acos(a); }
+Packet pacos(const Packet& a) { return acos(a); }
 
 /** \internal \returns the exp of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet pexp(const Packet& a) { using std::exp; return exp(a); }
+Packet pexp(const Packet& a) { return exp(a); }
 
 /** \internal \returns the log of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet plog(const Packet& a) { using std::log; return log(a); }
+Packet plog(const Packet& a) { return log(a); }
 
 /** \internal \returns the square-root of \a a (coeff-wise) */
 template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
-Packet psqrt(const Packet& a) { using std::sqrt; return sqrt(a); }
+Packet psqrt(const Packet& a) { return sqrt(a); }
 
 /***************************************************************************
 * The following functions might not have to be overwritten for vectorized types
@@ -311,21 +302,8 @@ struct palign_impl
   static inline void run(PacketType&, const PacketType&) {}
 };
 
-/** \internal update \a first using the concatenation of the packet_size minus \a Offset last elements
-  * of \a first and \a Offset first elements of \a second.
-  * 
-  * This function is currently only used to optimize matrix-vector products on unligned matrices.
-  * It takes 2 packets that represent a contiguous memory array, and returns a packet starting
-  * at the position \a Offset. For instance, for packets of 4 elements, we have:
-  *  Input:
-  *  - first = {f0,f1,f2,f3}
-  *  - second = {s0,s1,s2,s3}
-  * Output: 
-  *   - if Offset==0 then {f0,f1,f2,f3}
-  *   - if Offset==1 then {f1,f2,f3,s0}
-  *   - if Offset==2 then {f2,f3,s0,s1}
-  *   - if Offset==3 then {f3,s0,s1,s3}
-  */
+/** \internal update \a first using the concatenation of the \a Offset last elements
+  * of \a first and packet_size minus \a Offset first elements of \a second */
 template<int Offset,typename PacketType>
 inline void palign(PacketType& first, const PacketType& second)
 {

Eigen/src/Core/GlobalFunctions.h

@@ -1,7 +1,7 @@
 // This file is part of Eigen, a lightweight C++ template library
 // for linear algebra.
 //
-// Copyright (C) 2010-2012 Gael Guennebaud <gael.guennebaud@inria.fr>
+// Copyright (C) 2010 Gael Guennebaud <gael.guennebaud@inria.fr>
 // Copyright (C) 2010 Benoit Jacob <jacob.benoit.1@gmail.com>
 //
 // This Source Code Form is subject to the terms of the Mozilla
@@ -11,7 +11,7 @@
 #ifndef EIGEN_GLOBAL_FUNCTIONS_H
 #define EIGEN_GLOBAL_FUNCTIONS_H
 
-#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR) \
+#define EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(NAME,FUNCTOR) \
   template<typename Derived> \
   inline const Eigen::CwiseUnaryOp<Eigen::internal::FUNCTOR<typename Derived::Scalar>, const Derived> \
   NAME(const Eigen::ArrayBase<Derived>& x) { \
@@ -35,21 +35,20 @@
   };
 
 
-namespace Eigen
+namespace std
 {
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(real,scalar_real_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(imag,scalar_imag_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(conj,scalar_conjugate_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sin,scalar_sin_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cos,scalar_cos_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(asin,scalar_asin_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(acos,scalar_acos_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tan,scalar_tan_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(exp,scalar_exp_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log,scalar_log_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs,scalar_abs_op)
-  EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sqrt,scalar_sqrt_op)
-  
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(real,scalar_real_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(imag,scalar_imag_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(sin,scalar_sin_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(cos,scalar_cos_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(asin,scalar_asin_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(acos,scalar_acos_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(tan,scalar_tan_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(exp,scalar_exp_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(log,scalar_log_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(abs,scalar_abs_op)
+  EIGEN_ARRAY_DECLARE_GLOBAL_STD_UNARY(sqrt,scalar_sqrt_op)
+
   template<typename Derived>
   inline const Eigen::CwiseUnaryOp<Eigen::internal::scalar_pow_op<typename Derived::Scalar>, const Derived>
   pow(const Eigen::ArrayBase<Derived>& x, const typename Derived::Scalar& exponent) {
@@ -65,13 +64,16 @@ namespace Eigen
       exponents.derived()
     );
   }
-  
+}
+
+namespace Eigen
+{
   /**
   * \brief Component-wise division of a scalar by array elements.
   **/
   template <typename Derived>
   inline const Eigen::CwiseUnaryOp<Eigen::internal::scalar_inverse_mult_op<typename Derived::Scalar>, const Derived>
-    operator/(const typename Derived::Scalar& s, const Eigen::ArrayBase<Derived>& a)
+    operator/(typename Derived::Scalar s, const Eigen::ArrayBase<Derived>& a)
   {
     return Eigen::CwiseUnaryOp<Eigen::internal::scalar_inverse_mult_op<typename Derived::Scalar>, const Derived>(
       a.derived(),
@@ -83,10 +85,19 @@ namespace Eigen
   {
     EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(real,scalar_real_op)
     EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(imag,scalar_imag_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(sin,scalar_sin_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(cos,scalar_cos_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(asin,scalar_asin_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(acos,scalar_acos_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(tan,scalar_tan_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(exp,scalar_exp_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(log,scalar_log_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(abs,scalar_abs_op)
     EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(abs2,scalar_abs2_op)
+    EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(sqrt,scalar_sqrt_op)
   }
 }
 
-// TODO: cleanly disable those functions that are not supported on Array (numext::real_ref, internal::random, internal::isApprox...)
+// TODO: cleanly disable those functions that are not supported on Array (internal::real_ref, internal::random, internal::isApprox...)
 
 #endif // EIGEN_GLOBAL_FUNCTIONS_H

Eigen/src/Core/IO.h

@@ -55,8 +55,9 @@ struct IOFormat
     const std::string& _rowSeparator = "\n", const std::string& _rowPrefix="", const std::string& _rowSuffix="",
     const std::string& _matPrefix="", const std::string& _matSuffix="")
   : matPrefix(_matPrefix), matSuffix(_matSuffix), rowPrefix(_rowPrefix), rowSuffix(_rowSuffix), rowSeparator(_rowSeparator),
-    rowSpacer(""), coeffSeparator(_coeffSeparator), precision(_precision), flags(_flags)
+    coeffSeparator(_coeffSeparator), precision(_precision), flags(_flags)
   {
+    rowSpacer = "";
     int i = int(matSuffix.length())-1;
     while (i>=0 && matSuffix[i]!='\n')
     {
@@ -128,7 +129,6 @@ struct significant_decimals_default_impl
   static inline int run()
   {
     using std::ceil;
-    using std::log;
     return cast<RealScalar,int>(ceil(-log(NumTraits<RealScalar>::epsilon())/log(RealScalar(10))));
   }
 };
@@ -185,22 +185,21 @@ std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat&
     explicit_precision = fmt.precision;
   }
 
-  std::streamsize old_precision = 0;
-  if(explicit_precision) old_precision = s.precision(explicit_precision);
-
   bool align_cols = !(fmt.flags & DontAlignCols);
   if(align_cols)
   {
     // compute the largest width
-    for(Index j = 0; j < m.cols(); ++j)
+    for(Index j = 1; j < m.cols(); ++j)
       for(Index i = 0; i < m.rows(); ++i)
       {
         std::stringstream sstr;
-        sstr.copyfmt(s);
+        if(explicit_precision) sstr.precision(explicit_precision);
         sstr << m.coeff(i,j);
         width = std::max<Index>(width, Index(sstr.str().length()));
       }
   }
+  std::streamsize old_precision = 0;
+  if(explicit_precision) old_precision = s.precision(explicit_precision);
   s << fmt.matPrefix;
   for(Index i = 0; i < m.rows(); ++i)
   {

Eigen/src/Core/Map.h

@@ -133,36 +133,36 @@ template<typename PlainObjectType, int MapOptions, typename StrideType> class Ma
 
     /** Constructor in the fixed-size case.
       *
-      * \param dataPtr pointer to the array to map
-      * \param a_stride optional Stride object, passing the strides.
+      * \param data pointer to the array to map
+      * \param stride optional Stride object, passing the strides.
       */
-    inline Map(PointerArgType dataPtr, const StrideType& a_stride = StrideType())
-      : Base(cast_to_pointer_type(dataPtr)), m_stride(a_stride)
+    inline Map(PointerArgType data, const StrideType& stride = StrideType())
+      : Base(cast_to_pointer_type(data)), m_stride(stride)
     {
       PlainObjectType::Base::_check_template_params();
     }
 
     /** Constructor in the dynamic-size vector case.
       *
-      * \param dataPtr pointer to the array to map
-      * \param a_size the size of the vector expression
-      * \param a_stride optional Stride object, passing the strides.
+      * \param data pointer to the array to map
+      * \param size the size of the vector expression
+      * \param stride optional Stride object, passing the strides.
       */
-    inline Map(PointerArgType dataPtr, Index a_size, const StrideType& a_stride = StrideType())
-      : Base(cast_to_pointer_type(dataPtr), a_size), m_stride(a_stride)
+    inline Map(PointerArgType data, Index size, const StrideType& stride = StrideType())
+      : Base(cast_to_pointer_type(data), size), m_stride(stride)
     {
       PlainObjectType::Base::_check_template_params();
     }
 
     /** Constructor in the dynamic-size matrix case.
       *
-      * \param dataPtr pointer to the array to map
-      * \param nbRows the number of rows of the matrix expression
-      * \param nbCols the number of columns of the matrix expression
-      * \param a_stride optional Stride object, passing the strides.
+      * \param data pointer to the array to map
+      * \param rows the number of rows of the matrix expression
+      * \param cols the number of columns of the matrix expression
+      * \param stride optional Stride object, passing the strides.
       */
-    inline Map(PointerArgType dataPtr, Index nbRows, Index nbCols, const StrideType& a_stride = StrideType())
-      : Base(cast_to_pointer_type(dataPtr), nbRows, nbCols), m_stride(a_stride)
+    inline Map(PointerArgType data, Index rows, Index cols, const StrideType& stride = StrideType())
+      : Base(cast_to_pointer_type(data), rows, cols), m_stride(stride)
     {
       PlainObjectType::Base::_check_template_params();
     }

Eigen/src/Core/MapBase.h

@@ -87,9 +87,9 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
       */
     inline const Scalar* data() const { return m_data; }
 
-    inline const Scalar& coeff(Index rowId, Index colId) const
+    inline const Scalar& coeff(Index row, Index col) const
     {
-      return m_data[colId * colStride() + rowId * rowStride()];
+      return m_data[col * colStride() + row * rowStride()];
     }
 
     inline const Scalar& coeff(Index index) const
@@ -98,9 +98,9 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
       return m_data[index * innerStride()];
     }
 
-    inline const Scalar& coeffRef(Index rowId, Index colId) const
+    inline const Scalar& coeffRef(Index row, Index col) const
     {
-      return this->m_data[colId * colStride() + rowId * rowStride()];
+      return this->m_data[col * colStride() + row * rowStride()];
     }
 
     inline const Scalar& coeffRef(Index index) const
@@ -110,10 +110,10 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
     }
 
     template<int LoadMode>
-    inline PacketScalar packet(Index rowId, Index colId) const
+    inline PacketScalar packet(Index row, Index col) const
     {
       return internal::ploadt<PacketScalar, LoadMode>
-               (m_data + (colId * colStride() + rowId * rowStride()));
+               (m_data + (col * colStride() + row * rowStride()));
     }
 
     template<int LoadMode>
@@ -123,36 +123,32 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
       return internal::ploadt<PacketScalar, LoadMode>(m_data + index * innerStride());
     }
 
-    explicit inline MapBase(PointerType dataPtr) : m_data(dataPtr), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime)
+    inline MapBase(PointerType data) : m_data(data), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime)
     {
       EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived)
       checkSanity();
     }
 
-    inline MapBase(PointerType dataPtr, Index vecSize)
-            : m_data(dataPtr),
-              m_rows(RowsAtCompileTime == Dynamic ? vecSize : Index(RowsAtCompileTime)),
-              m_cols(ColsAtCompileTime == Dynamic ? vecSize : Index(ColsAtCompileTime))
+    inline MapBase(PointerType data, Index size)
+            : m_data(data),
+              m_rows(RowsAtCompileTime == Dynamic ? size : Index(RowsAtCompileTime)),
+              m_cols(ColsAtCompileTime == Dynamic ? size : Index(ColsAtCompileTime))
     {
       EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
-      eigen_assert(vecSize >= 0);
-      eigen_assert(dataPtr == 0 || SizeAtCompileTime == Dynamic || SizeAtCompileTime == vecSize);
+      eigen_assert(size >= 0);
+      eigen_assert(data == 0 || SizeAtCompileTime == Dynamic || SizeAtCompileTime == size);
       checkSanity();
     }
 
-    inline MapBase(PointerType dataPtr, Index nbRows, Index nbCols)
-            : m_data(dataPtr), m_rows(nbRows), m_cols(nbCols)
+    inline MapBase(PointerType data, Index rows, Index cols)
+            : m_data(data), m_rows(rows), m_cols(cols)
     {
-      eigen_assert( (dataPtr == 0)
-              || (   nbRows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == nbRows)
-                  && nbCols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == nbCols)));
+      eigen_assert( (data == 0)
+              || (   rows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows)
+                  && cols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols)));
       checkSanity();
     }
 
-    #ifdef EIGEN_MAPBASE_PLUGIN
-    #include EIGEN_MAPBASE_PLUGIN
-    #endif
-
   protected:
 
     void checkSanity() const
@@ -161,7 +157,7 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
                                         internal::inner_stride_at_compile_time<Derived>::ret==1),
                           PACKET_ACCESS_REQUIRES_TO_HAVE_INNER_STRIDE_FIXED_TO_1);
       eigen_assert(EIGEN_IMPLIES(internal::traits<Derived>::Flags&AlignedBit, (size_t(m_data) % 16) == 0)
-                   && "input pointer is not aligned on a 16 byte boundary");
+                   && "data is not aligned");
     }
 
     PointerType m_data;
@@ -172,7 +168,6 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
 template<typename Derived> class MapBase<Derived, WriteAccessors>
   : public MapBase<Derived, ReadOnlyAccessors>
 {
-    typedef MapBase<Derived, ReadOnlyAccessors> ReadOnlyMapBase;
   public:
 
     typedef MapBase<Derived, ReadOnlyAccessors> Base;
@@ -215,37 +210,33 @@ template<typename Derived> class MapBase<Derived, WriteAccessors>
     }
 
     template<int StoreMode>
-    inline void writePacket(Index row, Index col, const PacketScalar& val)
+    inline void writePacket(Index row, Index col, const PacketScalar& x)
     {
       internal::pstoret<Scalar, PacketScalar, StoreMode>
-               (this->m_data + (col * colStride() + row * rowStride()), val);
+               (this->m_data + (col * colStride() + row * rowStride()), x);
     }
 
     template<int StoreMode>
-    inline void writePacket(Index index, const PacketScalar& val)
+    inline void writePacket(Index index, const PacketScalar& x)
     {
       EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived)
       internal::pstoret<Scalar, PacketScalar, StoreMode>
-                (this->m_data + index * innerStride(), val);
+                (this->m_data + index * innerStride(), x);
     }
 
-    explicit inline MapBase(PointerType dataPtr) : Base(dataPtr) {}
-    inline MapBase(PointerType dataPtr, Index vecSize) : Base(dataPtr, vecSize) {}
-    inline MapBase(PointerType dataPtr, Index nbRows, Index nbCols) : Base(dataPtr, nbRows, nbCols) {}
+    explicit inline MapBase(PointerType data) : Base(data) {}
+    inline MapBase(PointerType data, Index size) : Base(data, size) {}
+    inline MapBase(PointerType data, Index rows, Index cols) : Base(data, rows, cols) {}
 
     Derived& operator=(const MapBase& other)
     {
-      ReadOnlyMapBase::Base::operator=(other);
+      Base::Base::operator=(other);
       return derived();
     }
 
-    // In theory we could simply refer to Base:Base::operator=, but MSVC does not like Base::Base,
-    // see bugs 821 and 920.
-    using ReadOnlyMapBase::Base::operator=;
+    using Base::Base::operator=;
 };
 
-#undef EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS
-
 } // end namespace Eigen
 
 #endif // EIGEN_MAPBASE_H

Eigen/src/Core/MathFunctions.h

@@ -51,15 +51,16 @@ struct global_math_functions_filtering_base
   typedef typename T::Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl type;
 };
 
-#define EIGEN_MATHFUNC_IMPL(func, scalar) Eigen::internal::func##_impl<typename Eigen::internal::global_math_functions_filtering_base<scalar>::type>
-#define EIGEN_MATHFUNC_RETVAL(func, scalar) typename Eigen::internal::func##_retval<typename Eigen::internal::global_math_functions_filtering_base<scalar>::type>::type
+#define EIGEN_MATHFUNC_IMPL(func, scalar) func##_impl<typename global_math_functions_filtering_base<scalar>::type>
+#define EIGEN_MATHFUNC_RETVAL(func, scalar) typename func##_retval<typename global_math_functions_filtering_base<scalar>::type>::type
+
 
 /****************************************************************************
 * Implementation of real                                                 *
 ****************************************************************************/
 
-template<typename Scalar, bool IsComplex = NumTraits<Scalar>::IsComplex>
-struct real_default_impl
+template<typename Scalar>
+struct real_impl
 {
   typedef typename NumTraits<Scalar>::Real RealScalar;
   static inline RealScalar run(const Scalar& x)
@@ -68,32 +69,34 @@ struct real_default_impl
   }
 };
 
-template<typename Scalar>
-struct real_default_impl<Scalar,true>
+template<typename RealScalar>
+struct real_impl<std::complex<RealScalar> >
 {
-  typedef typename NumTraits<Scalar>::Real RealScalar;
-  static inline RealScalar run(const Scalar& x)
+  static inline RealScalar run(const std::complex<RealScalar>& x)
   {
     using std::real;
     return real(x);
   }
 };
 
-template<typename Scalar> struct real_impl : real_default_impl<Scalar> {};
-
 template<typename Scalar>
 struct real_retval
 {
   typedef typename NumTraits<Scalar>::Real type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(real, Scalar) real(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(real, Scalar)::run(x);
+}
 
 /****************************************************************************
 * Implementation of imag                                                 *
 ****************************************************************************/
 
-template<typename Scalar, bool IsComplex = NumTraits<Scalar>::IsComplex>
-struct imag_default_impl
+template<typename Scalar>
+struct imag_impl
 {
   typedef typename NumTraits<Scalar>::Real RealScalar;
   static inline RealScalar run(const Scalar&)
@@ -102,25 +105,28 @@ struct imag_default_impl
   }
 };
 
-template<typename Scalar>
-struct imag_default_impl<Scalar,true>
+template<typename RealScalar>
+struct imag_impl<std::complex<RealScalar> >
 {
-  typedef typename NumTraits<Scalar>::Real RealScalar;
-  static inline RealScalar run(const Scalar& x)
+  static inline RealScalar run(const std::complex<RealScalar>& x)
   {
     using std::imag;
     return imag(x);
   }
 };
 
-template<typename Scalar> struct imag_impl : imag_default_impl<Scalar> {};
-
 template<typename Scalar>
 struct imag_retval
 {
   typedef typename NumTraits<Scalar>::Real type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(imag, Scalar) imag(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(imag, Scalar)::run(x);
+}
+
 /****************************************************************************
 * Implementation of real_ref                                             *
 ****************************************************************************/
@@ -145,6 +151,18 @@ struct real_ref_retval
   typedef typename NumTraits<Scalar>::Real & type;
 };
 
+template<typename Scalar>
+inline typename add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) >::type real_ref(const Scalar& x)
+{
+  return real_ref_impl<Scalar>::run(x);
+}
+
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) real_ref(Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(real_ref, Scalar)::run(x);
+}
+
 /****************************************************************************
 * Implementation of imag_ref                                             *
 ****************************************************************************/
@@ -185,11 +203,23 @@ struct imag_ref_retval
   typedef typename NumTraits<Scalar>::Real & type;
 };
 
+template<typename Scalar>
+inline typename add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) >::type imag_ref(const Scalar& x)
+{
+  return imag_ref_impl<Scalar>::run(x);
+}
+
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) imag_ref(Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(imag_ref, Scalar)::run(x);
+}
+
 /****************************************************************************
 * Implementation of conj                                                 *
 ****************************************************************************/
 
-template<typename Scalar, bool IsComplex = NumTraits<Scalar>::IsComplex>
+template<typename Scalar>
 struct conj_impl
 {
   static inline Scalar run(const Scalar& x)
@@ -198,10 +228,10 @@ struct conj_impl
   }
 };
 
-template<typename Scalar>
-struct conj_impl<Scalar,true>
+template<typename RealScalar>
+struct conj_impl<std::complex<RealScalar> >
 {
-  static inline Scalar run(const Scalar& x)
+  static inline std::complex<RealScalar> run(const std::complex<RealScalar>& x)
   {
     using std::conj;
     return conj(x);
@@ -214,12 +244,45 @@ struct conj_retval
   typedef Scalar type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(conj, Scalar) conj(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(conj, Scalar)::run(x);
+}
+
+/****************************************************************************
+* Implementation of abs                                                  *
+****************************************************************************/
+
+template<typename Scalar>
+struct abs_impl
+{
+  typedef typename NumTraits<Scalar>::Real RealScalar;
+  static inline RealScalar run(const Scalar& x)
+  {
+    using std::abs;
+    return abs(x);
+  }
+};
+
+template<typename Scalar>
+struct abs_retval
+{
+  typedef typename NumTraits<Scalar>::Real type;
+};
+
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(abs, Scalar) abs(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(abs, Scalar)::run(x);
+}
+
 /****************************************************************************
 * Implementation of abs2                                                 *
 ****************************************************************************/
 
-template<typename Scalar,bool IsComplex>
-struct abs2_impl_default
+template<typename Scalar>
+struct abs2_impl
 {
   typedef typename NumTraits<Scalar>::Real RealScalar;
   static inline RealScalar run(const Scalar& x)
@@ -228,32 +291,27 @@ struct abs2_impl_default
   }
 };
 
-template<typename Scalar>
-struct abs2_impl_default<Scalar, true> // IsComplex
+template<typename RealScalar>
+struct abs2_impl<std::complex<RealScalar> >
 {
-  typedef typename NumTraits<Scalar>::Real RealScalar;
-  static inline RealScalar run(const Scalar& x)
+  static inline RealScalar run(const std::complex<RealScalar>& x)
   {
     return real(x)*real(x) + imag(x)*imag(x);
   }
 };
 
-template<typename Scalar>
-struct abs2_impl
-{
-  typedef typename NumTraits<Scalar>::Real RealScalar;
-  static inline RealScalar run(const Scalar& x)
-  {
-    return abs2_impl_default<Scalar,NumTraits<Scalar>::IsComplex>::run(x);
-  }
-};
-
 template<typename Scalar>
 struct abs2_retval
 {
   typedef typename NumTraits<Scalar>::Real type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(abs2, Scalar) abs2(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(abs2, Scalar)::run(x);
+}
+
 /****************************************************************************
 * Implementation of norm1                                                *
 ****************************************************************************/
@@ -264,7 +322,6 @@ struct norm1_default_impl
   typedef typename NumTraits<Scalar>::Real RealScalar;
   static inline RealScalar run(const Scalar& x)
   {
-    using std::abs;
     return abs(real(x)) + abs(imag(x));
   }
 };
@@ -274,7 +331,6 @@ struct norm1_default_impl<Scalar, false>
 {
   static inline Scalar run(const Scalar& x)
   {
-    using std::abs;
     return abs(x);
   }
 };
@@ -288,6 +344,12 @@ struct norm1_retval
   typedef typename NumTraits<Scalar>::Real type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(norm1, Scalar) norm1(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(norm1, Scalar)::run(x);
+}
+
 /****************************************************************************
 * Implementation of hypot                                                *
 ****************************************************************************/
@@ -300,12 +362,9 @@ struct hypot_impl
   {
     using std::max;
     using std::min;
-    using std::abs;
-    using std::sqrt;
     RealScalar _x = abs(x);
     RealScalar _y = abs(y);
     RealScalar p = (max)(_x, _y);
-    if(p==RealScalar(0)) return RealScalar(0);
     RealScalar q = (min)(_x, _y);
     RealScalar qp = q/p;
     return p * sqrt(RealScalar(1) + qp*qp);
@@ -318,6 +377,12 @@ struct hypot_retval
   typedef typename NumTraits<Scalar>::Real type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(hypot, Scalar) hypot(const Scalar& x, const Scalar& y)
+{
+  return EIGEN_MATHFUNC_IMPL(hypot, Scalar)::run(x, y);
+}
+
 /****************************************************************************
 * Implementation of cast                                                 *
 ****************************************************************************/
@@ -340,29 +405,97 @@ inline NewType cast(const OldType& x)
 }
 
 /****************************************************************************
-* Implementation of atanh2                                                *
+* Implementation of sqrt                                                 *
 ****************************************************************************/
 
 template<typename Scalar, bool IsInteger>
-struct atanh2_default_impl
+struct sqrt_default_impl
 {
-  typedef Scalar retval;
-  typedef typename NumTraits<Scalar>::Real RealScalar;
-  static inline Scalar run(const Scalar& x, const Scalar& y)
+  static inline Scalar run(const Scalar& x)
   {
-    using std::abs;
-    using std::log;
     using std::sqrt;
-    Scalar z = x / y;
-    if (y == Scalar(0) || abs(z) > sqrt(NumTraits<RealScalar>::epsilon()))
-      return RealScalar(0.5) * log((y + x) / (y - x));
-    else
-      return z + z*z*z / RealScalar(3);
+    return sqrt(x);
   }
 };
 
 template<typename Scalar>
-struct atanh2_default_impl<Scalar, true>
+struct sqrt_default_impl<Scalar, true>
+{
+  static inline Scalar run(const Scalar&)
+  {
+#ifdef EIGEN2_SUPPORT
+    eigen_assert(!NumTraits<Scalar>::IsInteger);
+#else
+    EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar)
+#endif
+    return Scalar(0);
+  }
+};
+
+template<typename Scalar>
+struct sqrt_impl : sqrt_default_impl<Scalar, NumTraits<Scalar>::IsInteger> {};
+
+template<typename Scalar>
+struct sqrt_retval
+{
+  typedef Scalar type;
+};
+
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(sqrt, Scalar) sqrt(const Scalar& x)
+{
+  return EIGEN_MATHFUNC_IMPL(sqrt, Scalar)::run(x);
+}
+
+/****************************************************************************
+* Implementation of standard unary real functions (exp, log, sin, cos, ...  *
+****************************************************************************/
+
+// This macro instanciate all the necessary template mechanism which is common to all unary real functions.
+#define EIGEN_MATHFUNC_STANDARD_REAL_UNARY(NAME) \
+  template<typename Scalar, bool IsInteger> struct NAME##_default_impl {            \
+    static inline Scalar run(const Scalar& x) { using std::NAME; return NAME(x); }  \
+  };                                                                                \
+  template<typename Scalar> struct NAME##_default_impl<Scalar, true> {              \
+    static inline Scalar run(const Scalar&) {                                       \
+      EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar)                                       \
+      return Scalar(0);                                                             \
+    }                                                                               \
+  };                                                                                \
+  template<typename Scalar> struct NAME##_impl                                      \
+    : NAME##_default_impl<Scalar, NumTraits<Scalar>::IsInteger>                     \
+  {};                                                                               \
+  template<typename Scalar> struct NAME##_retval { typedef Scalar type; };          \
+  template<typename Scalar>                                                         \
+  inline EIGEN_MATHFUNC_RETVAL(NAME, Scalar) NAME(const Scalar& x) {                \
+    return EIGEN_MATHFUNC_IMPL(NAME, Scalar)::run(x);                               \
+  }
+
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(exp)
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(log)
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(sin)
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(cos)
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(tan)
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(asin)
+EIGEN_MATHFUNC_STANDARD_REAL_UNARY(acos)
+
+/****************************************************************************
+* Implementation of atan2                                                *
+****************************************************************************/
+
+template<typename Scalar, bool IsInteger>
+struct atan2_default_impl
+{
+  typedef Scalar retval;
+  static inline Scalar run(const Scalar& x, const Scalar& y)
+  {
+    using std::atan2;
+    return atan2(x, y);
+  }
+};
+
+template<typename Scalar>
+struct atan2_default_impl<Scalar, true>
 {
   static inline Scalar run(const Scalar&, const Scalar&)
   {
@@ -372,14 +505,20 @@ struct atanh2_default_impl<Scalar, true>
 };
 
 template<typename Scalar>
-struct atanh2_impl : atanh2_default_impl<Scalar, NumTraits<Scalar>::IsInteger> {};
+struct atan2_impl : atan2_default_impl<Scalar, NumTraits<Scalar>::IsInteger> {};
 
 template<typename Scalar>
-struct atanh2_retval
+struct atan2_retval
 {
   typedef Scalar type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(atan2, Scalar) atan2(const Scalar& x, const Scalar& y)
+{
+  return EIGEN_MATHFUNC_IMPL(atan2, Scalar)::run(x, y);
+}
+
 /****************************************************************************
 * Implementation of pow                                                  *
 ****************************************************************************/
@@ -423,6 +562,12 @@ struct pow_retval
   typedef Scalar type;
 };
 
+template<typename Scalar>
+inline EIGEN_MATHFUNC_RETVAL(pow, Scalar) pow(const Scalar& x, const Scalar& y)
+{
+  return EIGEN_MATHFUNC_IMPL(pow, Scalar)::run(x, y);
+}
+
 /****************************************************************************
 * Implementation of random                                               *
 ****************************************************************************/
@@ -507,24 +652,11 @@ struct floor_log2<n, lower, upper, floor_log2_bogus>
 template<typename Scalar>
 struct random_default_impl<Scalar, false, true>
 {
+  typedef typename NumTraits<Scalar>::NonInteger NonInteger;
+
   static inline Scalar run(const Scalar& x, const Scalar& y)
   {
-    typedef typename conditional<NumTraits<Scalar>::IsSigned,std::ptrdiff_t,std::size_t>::type ScalarX;
-    if(y<x)
-      return x;
-    // the following difference might overflow on a 32 bits system,
-    // but since y>=x the result converted to an unsigned long is still correct.
-    std::size_t range = ScalarX(y)-ScalarX(x);
-    std::size_t offset = 0;
-    // rejection sampling
-    std::size_t divisor = 1;
-    std::size_t multiplier = 1;
-    if(range<RAND_MAX) divisor = (std::size_t(RAND_MAX)+1)/(range+1);
-    else               multiplier = 1 + range/(std::size_t(RAND_MAX)+1);
-    do {
-      offset = (std::size_t(std::rand()) * multiplier) / divisor;
-    } while (offset > range);
-    return Scalar(ScalarX(x) + offset);
+    return x + Scalar((NonInteger(y)-x+1) * std::rand() / (RAND_MAX + NonInteger(1)));
   }
 
   static inline Scalar run()
@@ -534,10 +666,11 @@ struct random_default_impl<Scalar, false, true>
 #else
     enum { rand_bits = floor_log2<(unsigned int)(RAND_MAX)+1>::value,
            scalar_bits = sizeof(Scalar) * CHAR_BIT,
-           shift = EIGEN_PLAIN_ENUM_MAX(0, int(rand_bits) - int(scalar_bits)),
-           offset = NumTraits<Scalar>::IsSigned ? (1 << (EIGEN_PLAIN_ENUM_MIN(rand_bits,scalar_bits)-1)) : 0
+           shift = EIGEN_PLAIN_ENUM_MAX(0, int(rand_bits) - int(scalar_bits))
     };
-    return Scalar((std::rand() >> shift) - offset);
+    Scalar x = Scalar(std::rand() >> shift);
+    Scalar offset = NumTraits<Scalar>::IsSigned ? Scalar(1 << (rand_bits-1)) : Scalar(0);
+    return x - offset;
 #endif
   }
 };
@@ -569,97 +702,6 @@ inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random()
   return EIGEN_MATHFUNC_IMPL(random, Scalar)::run();
 }
 
-} // end namespace internal
-
-/****************************************************************************
-* Generic math function                                                    *
-****************************************************************************/
-
-namespace numext {
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(real, Scalar) real(const Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(real, Scalar)::run(x);
-}  
-
-template<typename Scalar>
-inline typename internal::add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) >::type real_ref(const Scalar& x)
-{
-  return internal::real_ref_impl<Scalar>::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) real_ref(Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(real_ref, Scalar)::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(imag, Scalar) imag(const Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(imag, Scalar)::run(x);
-}
-
-template<typename Scalar>
-inline typename internal::add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) >::type imag_ref(const Scalar& x)
-{
-  return internal::imag_ref_impl<Scalar>::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) imag_ref(Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(imag_ref, Scalar)::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(conj, Scalar) conj(const Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(conj, Scalar)::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(abs2, Scalar) abs2(const Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(abs2, Scalar)::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(norm1, Scalar) norm1(const Scalar& x)
-{
-  return EIGEN_MATHFUNC_IMPL(norm1, Scalar)::run(x);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(hypot, Scalar) hypot(const Scalar& x, const Scalar& y)
-{
-  return EIGEN_MATHFUNC_IMPL(hypot, Scalar)::run(x, y);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(atanh2, Scalar) atanh2(const Scalar& x, const Scalar& y)
-{
-  return EIGEN_MATHFUNC_IMPL(atanh2, Scalar)::run(x, y);
-}
-
-template<typename Scalar>
-inline EIGEN_MATHFUNC_RETVAL(pow, Scalar) pow(const Scalar& x, const Scalar& y)
-{
-  return EIGEN_MATHFUNC_IMPL(pow, Scalar)::run(x, y);
-}
-
-// std::isfinite is non standard, so let's define our own version,
-// even though it is not very efficient.
-template<typename T> bool (isfinite)(const T& x)
-{
-  return x<NumTraits<T>::highest() && x>NumTraits<T>::lowest();
-}
-
-} // end namespace numext
-
-namespace internal {
-
 /****************************************************************************
 * Implementation of fuzzy comparisons                                       *
 ****************************************************************************/
@@ -676,13 +718,11 @@ struct scalar_fuzzy_default_impl<Scalar, false, false>
   template<typename OtherScalar>
   static inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, const RealScalar& prec)
   {
-    using std::abs;
     return abs(x) <= abs(y) * prec;
   }
   static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar& prec)
   {
     using std::min;
-    using std::abs;
     return abs(x - y) <= (min)(abs(x), abs(y)) * prec;
   }
   static inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, const RealScalar& prec)
@@ -717,12 +757,12 @@ struct scalar_fuzzy_default_impl<Scalar, true, false>
   template<typename OtherScalar>
   static inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, const RealScalar& prec)
   {
-    return numext::abs2(x) <= numext::abs2(y) * prec * prec;
+    return abs2(x) <= abs2(y) * prec * prec;
   }
   static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar& prec)
   {
     using std::min;
-    return numext::abs2(x - y) <= (min)(numext::abs2(x), numext::abs2(y)) * prec * prec;
+    return abs2(x - y) <= (min)(abs2(x), abs2(y)) * prec * prec;
   }
 };
 
@@ -731,21 +771,21 @@ struct scalar_fuzzy_impl : scalar_fuzzy_default_impl<Scalar, NumTraits<Scalar>::
 
 template<typename Scalar, typename OtherScalar>
 inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y,
-                              const typename NumTraits<Scalar>::Real &precision = NumTraits<Scalar>::dummy_precision())
+                                   typename NumTraits<Scalar>::Real precision = NumTraits<Scalar>::dummy_precision())
 {
   return scalar_fuzzy_impl<Scalar>::template isMuchSmallerThan<OtherScalar>(x, y, precision);
 }
 
 template<typename Scalar>
 inline bool isApprox(const Scalar& x, const Scalar& y,
-                     const typename NumTraits<Scalar>::Real &precision = NumTraits<Scalar>::dummy_precision())
+                          typename NumTraits<Scalar>::Real precision = NumTraits<Scalar>::dummy_precision())
 {
   return scalar_fuzzy_impl<Scalar>::isApprox(x, y, precision);
 }
 
 template<typename Scalar>
 inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y,
-                               const typename NumTraits<Scalar>::Real &precision = NumTraits<Scalar>::dummy_precision())
+                                    typename NumTraits<Scalar>::Real precision = NumTraits<Scalar>::dummy_precision())
 {
   return scalar_fuzzy_impl<Scalar>::isApproxOrLessThan(x, y, precision);
 }
@@ -784,7 +824,17 @@ template<> struct scalar_fuzzy_impl<bool>
   
 };
 
-  
+/****************************************************************************
+* Special functions                                                          *
+****************************************************************************/
+
+// std::isfinite is non standard, so let's define our own version,
+// even though it is not very efficient.
+template<typename T> bool (isfinite)(const T& x)
+{
+  return x<NumTraits<T>::highest() && x>NumTraits<T>::lowest();
+}
+
 } // end namespace internal
 
 } // end namespace Eigen

Eigen/src/Core/Matrix.h

@@ -200,31 +200,16 @@ class Matrix
       *
       * \sa resize(Index,Index)
       */
-    EIGEN_STRONG_INLINE Matrix() : Base()
+    EIGEN_STRONG_INLINE explicit Matrix() : Base()
     {
       Base::_check_template_params();
-      EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+      EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 
     // FIXME is it still needed
     Matrix(internal::constructor_without_unaligned_array_assert)
       : Base(internal::constructor_without_unaligned_array_assert())
-    { Base::_check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED }
-
-#ifdef EIGEN_HAVE_RVALUE_REFERENCES
-    Matrix(Matrix&& other)
-      : Base(std::move(other))
-    {
-      Base::_check_template_params();
-      if (RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic)
-        Base::_set_noalias(other);
-    }
-    Matrix& operator=(Matrix&& other)
-    {
-      other.swap(*this);
-      return *this;
-    }
-#endif
+    { Base::_check_template_params(); EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED }
 
     /** \brief Constructs a vector or row-vector with given dimension. \only_for_vectors
       *
@@ -239,7 +224,7 @@ class Matrix
       EIGEN_STATIC_ASSERT_VECTOR_ONLY(Matrix)
       eigen_assert(dim >= 0);
       eigen_assert(SizeAtCompileTime == Dynamic || SizeAtCompileTime == dim);
-      EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+      EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 
     #ifndef EIGEN_PARSED_BY_DOXYGEN
@@ -319,7 +304,7 @@ class Matrix
       : Base(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
     {
       Base::_check_template_params();
-      Base::_resize_to_match(other);
+      Base::resize(other.rows(), other.cols());
       // FIXME/CHECK: isn't *this = other.derived() more efficient. it allows to
       //              go for pure _set() implementations, right?
       *this = other;

Eigen/src/Core/MatrixBase.h

@@ -159,11 +159,10 @@ template<typename Derived> class MatrixBase
     template<typename OtherDerived>
     Derived& operator=(const ReturnByValue<OtherDerived>& other);
 
+#ifndef EIGEN_PARSED_BY_DOXYGEN
     template<typename ProductDerived, typename Lhs, typename Rhs>
     Derived& lazyAssign(const ProductBase<ProductDerived, Lhs,Rhs>& other);
-
-    template<typename MatrixPower, typename Lhs, typename Rhs>
-    Derived& lazyAssign(const MatrixPowerProduct<MatrixPower, Lhs,Rhs>& other);
+#endif // not EIGEN_PARSED_BY_DOXYGEN
 
     template<typename OtherDerived>
     Derived& operator+=(const MatrixBase<OtherDerived>& other);
@@ -213,20 +212,24 @@ template<typename Derived> class MatrixBase
 
     typedef Diagonal<Derived> DiagonalReturnType;
     DiagonalReturnType diagonal();
-    typedef typename internal::add_const<Diagonal<const Derived> >::type ConstDiagonalReturnType;
-    ConstDiagonalReturnType diagonal() const;
+    typedef const Diagonal<const Derived> ConstDiagonalReturnType;
+    const ConstDiagonalReturnType diagonal() const;
 
     template<int Index> struct DiagonalIndexReturnType { typedef Diagonal<Derived,Index> Type; };
     template<int Index> struct ConstDiagonalIndexReturnType { typedef const Diagonal<const Derived,Index> Type; };
 
     template<int Index> typename DiagonalIndexReturnType<Index>::Type diagonal();
     template<int Index> typename ConstDiagonalIndexReturnType<Index>::Type diagonal() const;
-    
-    typedef Diagonal<Derived,DynamicIndex> DiagonalDynamicIndexReturnType;
-    typedef typename internal::add_const<Diagonal<const Derived,DynamicIndex> >::type ConstDiagonalDynamicIndexReturnType;
 
-    DiagonalDynamicIndexReturnType diagonal(Index index);
-    ConstDiagonalDynamicIndexReturnType diagonal(Index index) const;
+    // Note: The "MatrixBase::" prefixes are added to help MSVC9 to match these declarations with the later implementations.
+    // On the other hand they confuse MSVC8...
+    #if (defined _MSC_VER) && (_MSC_VER >= 1500) // 2008 or later
+    typename MatrixBase::template DiagonalIndexReturnType<Dynamic>::Type diagonal(Index index);
+    typename MatrixBase::template ConstDiagonalIndexReturnType<Dynamic>::Type diagonal(Index index) const;
+    #else
+    typename DiagonalIndexReturnType<Dynamic>::Type diagonal(Index index);
+    typename ConstDiagonalIndexReturnType<Dynamic>::Type diagonal(Index index) const;
+    #endif
 
     #ifdef EIGEN2_SUPPORT
     template<unsigned int Mode> typename internal::eigen2_part_return_type<Derived, Mode>::type part();
@@ -252,7 +255,7 @@ template<typename Derived> class MatrixBase
     template<unsigned int UpLo> typename ConstSelfAdjointViewReturnType<UpLo>::Type selfadjointView() const;
 
     const SparseView<Derived> sparseView(const Scalar& m_reference = Scalar(0),
-                                         const typename NumTraits<Scalar>::Real& m_epsilon = NumTraits<Scalar>::dummy_precision()) const;
+                                         typename NumTraits<Scalar>::Real m_epsilon = NumTraits<Scalar>::dummy_precision()) const;
     static const IdentityReturnType Identity();
     static const IdentityReturnType Identity(Index rows, Index cols);
     static const BasisReturnType Unit(Index size, Index i);
@@ -268,16 +271,16 @@ template<typename Derived> class MatrixBase
     Derived& setIdentity();
     Derived& setIdentity(Index rows, Index cols);
 
-    bool isIdentity(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    bool isDiagonal(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isIdentity(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isDiagonal(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
 
-    bool isUpperTriangular(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    bool isLowerTriangular(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isUpperTriangular(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isLowerTriangular(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
 
     template<typename OtherDerived>
     bool isOrthogonal(const MatrixBase<OtherDerived>& other,
-                      const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
-    bool isUnitary(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
+                      RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
+    bool isUnitary(RealScalar prec = NumTraits<Scalar>::dummy_precision()) const;
 
     /** \returns true if each coefficients of \c *this and \a other are all exactly equal.
       * \warning When using floating point scalar values you probably should rather use a
@@ -311,7 +314,7 @@ template<typename Derived> class MatrixBase
     MatrixBase<Derived>& matrix() { return *this; }
     const MatrixBase<Derived>& matrix() const { return *this; }
 
-    /** \returns an \link Eigen::ArrayBase Array \endlink expression of this matrix
+    /** \returns an \link ArrayBase Array \endlink expression of this matrix
       * \sa ArrayBase::matrix() */
     ArrayWrapper<Derived> array() { return derived(); }
     const ArrayWrapper<const Derived> array() const { return derived(); }
@@ -440,15 +443,6 @@ template<typename Derived> class MatrixBase
     template<typename OtherScalar>
     void applyOnTheRight(Index p, Index q, const JacobiRotation<OtherScalar>& j);
 
-///////// SparseCore module /////////
-
-    template<typename OtherDerived>
-    EIGEN_STRONG_INLINE const typename SparseMatrixBase<OtherDerived>::template CwiseProductDenseReturnType<Derived>::Type
-    cwiseProduct(const SparseMatrixBase<OtherDerived> &other) const
-    {
-      return other.cwiseProduct(derived());
-    }
-
 ///////// MatrixFunctions module /////////
 
     typedef typename internal::stem_function<Scalar>::type StemFunction;
@@ -460,7 +454,6 @@ template<typename Derived> class MatrixBase
     const MatrixFunctionReturnValue<Derived> sin() const;
     const MatrixSquareRootReturnValue<Derived> sqrt() const;
     const MatrixLogarithmReturnValue<Derived> log() const;
-    const MatrixPowerReturnValue<Derived> pow(const RealScalar& p) const;
 
 #ifdef EIGEN2_SUPPORT
     template<typename ProductDerived, typename Lhs, typename Rhs>
@@ -513,51 +506,6 @@ template<typename Derived> class MatrixBase
     {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;}
 };
 
-
-/***************************************************************************
-* Implementation of matrix base methods
-***************************************************************************/
-
-/** replaces \c *this by \c *this * \a other.
-  *
-  * \returns a reference to \c *this
-  *
-  * Example: \include MatrixBase_applyOnTheRight.cpp
-  * Output: \verbinclude MatrixBase_applyOnTheRight.out
-  */
-template<typename Derived>
-template<typename OtherDerived>
-inline Derived&
-MatrixBase<Derived>::operator*=(const EigenBase<OtherDerived> &other)
-{
-  other.derived().applyThisOnTheRight(derived());
-  return derived();
-}
-
-/** replaces \c *this by \c *this * \a other. It is equivalent to MatrixBase::operator*=().
-  *
-  * Example: \include MatrixBase_applyOnTheRight.cpp
-  * Output: \verbinclude MatrixBase_applyOnTheRight.out
-  */
-template<typename Derived>
-template<typename OtherDerived>
-inline void MatrixBase<Derived>::applyOnTheRight(const EigenBase<OtherDerived> &other)
-{
-  other.derived().applyThisOnTheRight(derived());
-}
-
-/** replaces \c *this by \a other * \c *this.
-  *
-  * Example: \include MatrixBase_applyOnTheLeft.cpp
-  * Output: \verbinclude MatrixBase_applyOnTheLeft.out
-  */
-template<typename Derived>
-template<typename OtherDerived>
-inline void MatrixBase<Derived>::applyOnTheLeft(const EigenBase<OtherDerived> &other)
-{
-  other.derived().applyThisOnTheLeft(derived());
-}
-
 } // end namespace Eigen
 
 #endif // EIGEN_MATRIXBASE_H

Eigen/src/Core/NoAlias.h

@@ -80,17 +80,8 @@ class NoAlias
     template<typename Lhs, typename Rhs, int NestingFlags>
     EIGEN_STRONG_INLINE ExpressionType& operator-=(const CoeffBasedProduct<Lhs,Rhs,NestingFlags>& other)
     { return m_expression.derived() -= CoeffBasedProduct<Lhs,Rhs,NestByRefBit>(other.lhs(), other.rhs()); }
-    
-    template<typename OtherDerived>
-    ExpressionType& operator=(const ReturnByValue<OtherDerived>& func)
-    { return m_expression = func; }
 #endif
 
-    ExpressionType& expression() const
-    {
-      return m_expression;
-    }
-
   protected:
     ExpressionType& m_expression;
 };

Eigen/src/Core/NumTraits.h

@@ -140,9 +140,6 @@ struct NumTraits<Array<Scalar, Rows, Cols, Options, MaxRows, MaxCols> >
     AddCost  = ArrayType::SizeAtCompileTime==Dynamic ? Dynamic : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::AddCost,
     MulCost  = ArrayType::SizeAtCompileTime==Dynamic ? Dynamic : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::MulCost
   };
-  
-  static inline RealScalar epsilon() { return NumTraits<RealScalar>::epsilon(); }
-  static inline RealScalar dummy_precision() { return NumTraits<RealScalar>::dummy_precision(); }
 };
 
 } // end namespace Eigen

Eigen/src/Core/PermutationMatrix.h

@@ -139,9 +139,9 @@ class PermutationBase : public EigenBase<Derived>
 
     /** Resizes to given size.
       */
-    inline void resize(Index newSize)
+    inline void resize(Index size)
     {
-      indices().resize(newSize);
+      indices().resize(size);
     }
 
     /** Sets *this to be the identity permutation matrix */
@@ -153,9 +153,9 @@ class PermutationBase : public EigenBase<Derived>
 
     /** Sets *this to be the identity permutation matrix of given size.
       */
-    void setIdentity(Index newSize)
+    void setIdentity(Index size)
     {
-      resize(newSize);
+      resize(size);
       setIdentity();
     }
 
@@ -250,35 +250,6 @@ class PermutationBase : public EigenBase<Derived>
     template<typename Other> friend
     inline PlainPermutationType operator*(const Transpose<PermutationBase<Other> >& other, const PermutationBase& perm)
     { return PlainPermutationType(internal::PermPermProduct, other.eval(), perm); }
-    
-    /** \returns the determinant of the permutation matrix, which is either 1 or -1 depending on the parity of the permutation.
-      *
-      * This function is O(\c n) procedure allocating a buffer of \c n booleans.
-      */
-    Index determinant() const
-    {
-      Index res = 1;
-      Index n = size();
-      Matrix<bool,RowsAtCompileTime,1,0,MaxRowsAtCompileTime> mask(n);
-      mask.fill(false);
-      Index r = 0;
-      while(r < n)
-      {
-        // search for the next seed
-        while(r<n && mask[r]) r++;
-        if(r>=n)
-          break;
-        // we got one, let's follow it until we are back to the seed
-        Index k0 = r++;
-        mask.coeffRef(k0) = true;
-        for(Index k=indices().coeff(k0); k!=k0; k=indices().coeff(k))
-        {
-          mask.coeffRef(k) = true;
-          res = -res;
-        }
-      }
-      return res;
-    }
 
   protected:
 
@@ -346,7 +317,7 @@ class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompile
       * array's size.
       */
     template<typename Other>
-    explicit inline PermutationMatrix(const MatrixBase<Other>& a_indices) : m_indices(a_indices)
+    explicit inline PermutationMatrix(const MatrixBase<Other>& indices) : m_indices(indices)
     {}
 
     /** Convert the Transpositions \a tr to a permutation matrix */
@@ -435,12 +406,12 @@ class Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType>,
     typedef typename IndicesType::Scalar Index;
     #endif
 
-    inline Map(const Index* indicesPtr)
-      : m_indices(indicesPtr)
+    inline Map(const Index* indices)
+      : m_indices(indices)
     {}
 
-    inline Map(const Index* indicesPtr, Index size)
-      : m_indices(indicesPtr,size)
+    inline Map(const Index* indices, Index size)
+      : m_indices(indices,size)
     {}
 
     /** Copies the other permutation into *this */
@@ -519,8 +490,8 @@ class PermutationWrapper : public PermutationBase<PermutationWrapper<_IndicesTyp
     typedef typename Traits::IndicesType IndicesType;
     #endif
 
-    inline PermutationWrapper(const IndicesType& a_indices)
-      : m_indices(a_indices)
+    inline PermutationWrapper(const IndicesType& indices)
+      : m_indices(indices)
     {}
 
     /** const version of indices(). */
@@ -570,30 +541,24 @@ struct permut_matrix_product_retval
  : public ReturnByValue<permut_matrix_product_retval<PermutationType, MatrixType, Side, Transposed> >
 {
     typedef typename remove_all<typename MatrixType::Nested>::type MatrixTypeNestedCleaned;
-    typedef typename MatrixType::Index Index;
 
     permut_matrix_product_retval(const PermutationType& perm, const MatrixType& matrix)
       : m_permutation(perm), m_matrix(matrix)
     {}
 
-    inline Index rows() const { return m_matrix.rows(); }
-    inline Index cols() const { return m_matrix.cols(); }
+    inline int rows() const { return m_matrix.rows(); }
+    inline int cols() const { return m_matrix.cols(); }
 
     template<typename Dest> inline void evalTo(Dest& dst) const
     {
-      const Index n = Side==OnTheLeft ? rows() : cols();
-      // FIXME we need an is_same for expression that is not sensitive to constness. For instance
-      // is_same_xpr<Block<const Matrix>, Block<Matrix> >::value should be true.
-      const typename Dest::Scalar *dst_data = internal::extract_data(dst);
-      if(    is_same<MatrixTypeNestedCleaned,Dest>::value
-          && blas_traits<MatrixTypeNestedCleaned>::HasUsableDirectAccess
-          && blas_traits<Dest>::HasUsableDirectAccess
-          && dst_data!=0 && dst_data == extract_data(m_matrix))
+      const int n = Side==OnTheLeft ? rows() : cols();
+
+      if(is_same<MatrixTypeNestedCleaned,Dest>::value && extract_data(dst) == extract_data(m_matrix))
       {
         // apply the permutation inplace
         Matrix<bool,PermutationType::RowsAtCompileTime,1,0,PermutationType::MaxRowsAtCompileTime> mask(m_permutation.size());
         mask.fill(false);
-        Index r = 0;
+        int r = 0;
         while(r < m_permutation.size())
         {
           // search for the next seed
@@ -601,10 +566,10 @@ struct permut_matrix_product_retval
           if(r>=m_permutation.size())
             break;
           // we got one, let's follow it until we are back to the seed
-          Index k0 = r++;
-          Index kPrev = k0;
+          int k0 = r++;
+          int kPrev = k0;
           mask.coeffRef(k0) = true;
-          for(Index k=m_permutation.indices().coeff(k0); k!=k0; k=m_permutation.indices().coeff(k))
+          for(int k=m_permutation.indices().coeff(k0); k!=k0; k=m_permutation.indices().coeff(k))
           {
                   Block<Dest, Side==OnTheLeft ? 1 : Dest::RowsAtCompileTime, Side==OnTheRight ? 1 : Dest::ColsAtCompileTime>(dst, k)
             .swap(Block<Dest, Side==OnTheLeft ? 1 : Dest::RowsAtCompileTime, Side==OnTheRight ? 1 : Dest::ColsAtCompileTime>

Eigen/src/Core/PlainObjectBase.h

@@ -11,46 +11,30 @@
 #ifndef EIGEN_DENSESTORAGEBASE_H
 #define EIGEN_DENSESTORAGEBASE_H
 
-#if defined(EIGEN_INITIALIZE_MATRICES_BY_ZERO)
-# define EIGEN_INITIALIZE_COEFFS
-# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED for(int i=0;i<base().size();++i) coeffRef(i)=Scalar(0);
-#elif defined(EIGEN_INITIALIZE_MATRICES_BY_NAN)
-# define EIGEN_INITIALIZE_COEFFS
-# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED for(int i=0;i<base().size();++i) coeffRef(i)=std::numeric_limits<Scalar>::quiet_NaN();
+#ifdef EIGEN_INITIALIZE_MATRICES_BY_ZERO
+# define EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED for(int i=0;i<base().size();++i) coeffRef(i)=Scalar(0);
 #else
-# undef EIGEN_INITIALIZE_COEFFS
-# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+# define EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
 #endif
 
 namespace Eigen {
 
 namespace internal {
 
-template<int MaxSizeAtCompileTime> struct check_rows_cols_for_overflow {
-  template<typename Index>
-  static EIGEN_ALWAYS_INLINE void run(Index, Index)
-  {
-  }
-};
+template<typename Index>
+EIGEN_ALWAYS_INLINE void check_rows_cols_for_overflow(Index rows, Index cols)
+{
+  // http://hg.mozilla.org/mozilla-central/file/6c8a909977d3/xpcom/ds/CheckedInt.h#l242
+  // we assume Index is signed
+  Index max_index = (size_t(1) << (8 * sizeof(Index) - 1)) - 1; // assume Index is signed
+  bool error = (rows < 0  || cols < 0)  ? true
+             : (rows == 0 || cols == 0) ? false
+                                        : (rows > max_index / cols);
+  if (error)
+    throw_std_bad_alloc();
+}
 
-template<> struct check_rows_cols_for_overflow<Dynamic> {
-  template<typename Index>
-  static EIGEN_ALWAYS_INLINE void run(Index rows, Index cols)
-  {
-    // http://hg.mozilla.org/mozilla-central/file/6c8a909977d3/xpcom/ds/CheckedInt.h#l242
-    // we assume Index is signed
-    Index max_index = (size_t(1) << (8 * sizeof(Index) - 1)) - 1; // assume Index is signed
-    bool error = (rows == 0 || cols == 0) ? false
-               : (rows > max_index / cols);
-    if (error)
-      throw_std_bad_alloc();
-  }
-};
-
-template <typename Derived,
-          typename OtherDerived = Derived,
-          bool IsVector = bool(Derived::IsVectorAtCompileTime) && bool(OtherDerived::IsVectorAtCompileTime)>
-struct conservative_resize_like_impl;
+template <typename Derived, typename OtherDerived = Derived, bool IsVector = bool(Derived::IsVectorAtCompileTime)> struct conservative_resize_like_impl;
 
 template<typename MatrixTypeA, typename MatrixTypeB, bool SwapPointers> struct matrix_swap_impl;
 
@@ -135,12 +119,12 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
     EIGEN_STRONG_INLINE Index rows() const { return m_storage.rows(); }
     EIGEN_STRONG_INLINE Index cols() const { return m_storage.cols(); }
 
-    EIGEN_STRONG_INLINE const Scalar& coeff(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE const Scalar& coeff(Index row, Index col) const
     {
       if(Flags & RowMajorBit)
-        return m_storage.data()[colId + rowId * m_storage.cols()];
+        return m_storage.data()[col + row * m_storage.cols()];
       else // column-major
-        return m_storage.data()[rowId + colId * m_storage.rows()];
+        return m_storage.data()[row + col * m_storage.rows()];
     }
 
     EIGEN_STRONG_INLINE const Scalar& coeff(Index index) const
@@ -148,12 +132,12 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       return m_storage.data()[index];
     }
 
-    EIGEN_STRONG_INLINE Scalar& coeffRef(Index rowId, Index colId)
+    EIGEN_STRONG_INLINE Scalar& coeffRef(Index row, Index col)
     {
       if(Flags & RowMajorBit)
-        return m_storage.data()[colId + rowId * m_storage.cols()];
+        return m_storage.data()[col + row * m_storage.cols()];
       else // column-major
-        return m_storage.data()[rowId + colId * m_storage.rows()];
+        return m_storage.data()[row + col * m_storage.rows()];
     }
 
     EIGEN_STRONG_INLINE Scalar& coeffRef(Index index)
@@ -161,12 +145,12 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       return m_storage.data()[index];
     }
 
-    EIGEN_STRONG_INLINE const Scalar& coeffRef(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE const Scalar& coeffRef(Index row, Index col) const
     {
       if(Flags & RowMajorBit)
-        return m_storage.data()[colId + rowId * m_storage.cols()];
+        return m_storage.data()[col + row * m_storage.cols()];
       else // column-major
-        return m_storage.data()[rowId + colId * m_storage.rows()];
+        return m_storage.data()[row + col * m_storage.rows()];
     }
 
     EIGEN_STRONG_INLINE const Scalar& coeffRef(Index index) const
@@ -176,12 +160,12 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
 
     /** \internal */
     template<int LoadMode>
-    EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
+    EIGEN_STRONG_INLINE PacketScalar packet(Index row, Index col) const
     {
       return internal::ploadt<PacketScalar, LoadMode>
                (m_storage.data() + (Flags & RowMajorBit
-                                   ? colId + rowId * m_storage.cols()
-                                   : rowId + colId * m_storage.rows()));
+                                   ? col + row * m_storage.cols()
+                                   : row + col * m_storage.rows()));
     }
 
     /** \internal */
@@ -193,19 +177,19 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
 
     /** \internal */
     template<int StoreMode>
-    EIGEN_STRONG_INLINE void writePacket(Index rowId, Index colId, const PacketScalar& val)
+    EIGEN_STRONG_INLINE void writePacket(Index row, Index col, const PacketScalar& x)
     {
       internal::pstoret<Scalar, PacketScalar, StoreMode>
               (m_storage.data() + (Flags & RowMajorBit
-                                   ? colId + rowId * m_storage.cols()
-                                   : rowId + colId * m_storage.rows()), val);
+                                   ? col + row * m_storage.cols()
+                                   : row + col * m_storage.rows()), x);
     }
 
     /** \internal */
     template<int StoreMode>
-    EIGEN_STRONG_INLINE void writePacket(Index index, const PacketScalar& val)
+    EIGEN_STRONG_INLINE void writePacket(Index index, const PacketScalar& x)
     {
-      internal::pstoret<Scalar, PacketScalar, StoreMode>(m_storage.data() + index, val);
+      internal::pstoret<Scalar, PacketScalar, StoreMode>(m_storage.data() + index, x);
     }
 
     /** \returns a const pointer to the data array of this matrix */
@@ -232,22 +216,17 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       *
       * \sa resize(Index) for vectors, resize(NoChange_t, Index), resize(Index, NoChange_t)
       */
-    EIGEN_STRONG_INLINE void resize(Index nbRows, Index nbCols)
+    EIGEN_STRONG_INLINE void resize(Index rows, Index cols)
     {
-      eigen_assert(   EIGEN_IMPLIES(RowsAtCompileTime!=Dynamic,nbRows==RowsAtCompileTime)
-                   && EIGEN_IMPLIES(ColsAtCompileTime!=Dynamic,nbCols==ColsAtCompileTime)
-                   && EIGEN_IMPLIES(RowsAtCompileTime==Dynamic && MaxRowsAtCompileTime!=Dynamic,nbRows<=MaxRowsAtCompileTime)
-                   && EIGEN_IMPLIES(ColsAtCompileTime==Dynamic && MaxColsAtCompileTime!=Dynamic,nbCols<=MaxColsAtCompileTime)
-                   && nbRows>=0 && nbCols>=0 && "Invalid sizes when resizing a matrix or array.");
-      internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(nbRows, nbCols);
-      #ifdef EIGEN_INITIALIZE_COEFFS
-        Index size = nbRows*nbCols;
+      #ifdef EIGEN_INITIALIZE_MATRICES_BY_ZERO
+        internal::check_rows_cols_for_overflow(rows, cols);
+        Index size = rows*cols;
         bool size_changed = size != this->size();
-        m_storage.resize(size, nbRows, nbCols);
-        if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+        m_storage.resize(size, rows, cols);
+        if(size_changed) EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
       #else
-        internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(nbRows, nbCols);
-        m_storage.resize(nbRows*nbCols, nbRows, nbCols);
+        internal::check_rows_cols_for_overflow(rows, cols);
+        m_storage.resize(rows*cols, rows, cols);
       #endif
     }
 
@@ -265,16 +244,16 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
     inline void resize(Index size)
     {
       EIGEN_STATIC_ASSERT_VECTOR_ONLY(PlainObjectBase)
-      eigen_assert(((SizeAtCompileTime == Dynamic && (MaxSizeAtCompileTime==Dynamic || size<=MaxSizeAtCompileTime)) || SizeAtCompileTime == size) && size>=0);
-      #ifdef EIGEN_INITIALIZE_COEFFS
+      eigen_assert(SizeAtCompileTime == Dynamic || SizeAtCompileTime == size);
+      #ifdef EIGEN_INITIALIZE_MATRICES_BY_ZERO
         bool size_changed = size != this->size();
       #endif
       if(RowsAtCompileTime == 1)
         m_storage.resize(size, 1, size);
       else
         m_storage.resize(size, size, 1);
-      #ifdef EIGEN_INITIALIZE_COEFFS
-        if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+      #ifdef EIGEN_INITIALIZE_MATRICES_BY_ZERO
+        if(size_changed) EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
       #endif
     }
 
@@ -286,9 +265,9 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       *
       * \sa resize(Index,Index)
       */
-    inline void resize(NoChange_t, Index nbCols)
+    inline void resize(NoChange_t, Index cols)
     {
-      resize(rows(), nbCols);
+      resize(rows(), cols);
     }
 
     /** Resizes the matrix, changing only the number of rows. For the parameter of type NoChange_t, just pass the special value \c NoChange
@@ -299,9 +278,9 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       *
       * \sa resize(Index,Index)
       */
-    inline void resize(Index nbRows, NoChange_t)
+    inline void resize(Index rows, NoChange_t)
     {
-      resize(nbRows, cols());
+      resize(rows, cols());
     }
 
     /** Resizes \c *this to have the same dimensions as \a other.
@@ -315,8 +294,8 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
     EIGEN_STRONG_INLINE void resizeLike(const EigenBase<OtherDerived>& _other)
     {
       const OtherDerived& other = _other.derived();
-      internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(Index(other.rows()), Index(other.cols()));
-      const Index othersize = Index(other.rows())*Index(other.cols());
+      internal::check_rows_cols_for_overflow(other.rows(), other.cols());
+      const Index othersize = other.rows()*other.cols();
       if(RowsAtCompileTime == 1)
       {
         eigen_assert(other.rows() == 1 || other.cols() == 1);
@@ -339,9 +318,9 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       * Matrices are resized relative to the top-left element. In case values need to be 
       * appended to the matrix they will be uninitialized.
       */
-    EIGEN_STRONG_INLINE void conservativeResize(Index nbRows, Index nbCols)
+    EIGEN_STRONG_INLINE void conservativeResize(Index rows, Index cols)
     {
-      internal::conservative_resize_like_impl<Derived>::run(*this, nbRows, nbCols);
+      internal::conservative_resize_like_impl<Derived>::run(*this, rows, cols);
     }
 
     /** Resizes the matrix to \a rows x \a cols while leaving old values untouched.
@@ -351,10 +330,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       *
       * In case the matrix is growing, new rows will be uninitialized.
       */
-    EIGEN_STRONG_INLINE void conservativeResize(Index nbRows, NoChange_t)
+    EIGEN_STRONG_INLINE void conservativeResize(Index rows, NoChange_t)
     {
       // Note: see the comment in conservativeResize(Index,Index)
-      conservativeResize(nbRows, cols());
+      conservativeResize(rows, cols());
     }
 
     /** Resizes the matrix to \a rows x \a cols while leaving old values untouched.
@@ -364,10 +343,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       *
       * In case the matrix is growing, new columns will be uninitialized.
       */
-    EIGEN_STRONG_INLINE void conservativeResize(NoChange_t, Index nbCols)
+    EIGEN_STRONG_INLINE void conservativeResize(NoChange_t, Index cols)
     {
       // Note: see the comment in conservativeResize(Index,Index)
-      conservativeResize(rows(), nbCols);
+      conservativeResize(rows(), cols);
     }
 
     /** Resizes the vector to \a size while retaining old values.
@@ -421,10 +400,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
       return Base::operator=(func);
     }
 
-    EIGEN_STRONG_INLINE PlainObjectBase() : m_storage()
+    EIGEN_STRONG_INLINE explicit PlainObjectBase() : m_storage()
     {
 //       _check_template_params();
-//       EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+//       EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 
 #ifndef EIGEN_PARSED_BY_DOXYGEN
@@ -433,45 +412,15 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
     PlainObjectBase(internal::constructor_without_unaligned_array_assert)
       : m_storage(internal::constructor_without_unaligned_array_assert())
     {
-//       _check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+//       _check_template_params(); EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 #endif
 
-#ifdef EIGEN_HAVE_RVALUE_REFERENCES
-    PlainObjectBase(PlainObjectBase&& other)
-      : m_storage( std::move(other.m_storage) )
-    {
-    }
-
-    PlainObjectBase& operator=(PlainObjectBase&& other)
-    {
-      using std::swap;
-      swap(m_storage, other.m_storage);
-      return *this;
-    }
-#endif
-
-    /** Copy constructor */
-    EIGEN_STRONG_INLINE PlainObjectBase(const PlainObjectBase& other)
-      : m_storage()
-    {
-      _check_template_params();
-      lazyAssign(other);
-    }
-
-    template<typename OtherDerived>
-    EIGEN_STRONG_INLINE PlainObjectBase(const DenseBase<OtherDerived> &other)
-      : m_storage()
-    {
-      _check_template_params();
-      lazyAssign(other);
-    }
-
-    EIGEN_STRONG_INLINE PlainObjectBase(Index a_size, Index nbRows, Index nbCols)
-      : m_storage(a_size, nbRows, nbCols)
+    EIGEN_STRONG_INLINE PlainObjectBase(Index size, Index rows, Index cols)
+      : m_storage(size, rows, cols)
     {
 //       _check_template_params();
-//       EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
+//       EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
 
     /** \copydoc MatrixBase::operator=(const EigenBase<OtherDerived>&)
@@ -487,10 +436,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
     /** \sa MatrixBase::operator=(const EigenBase<OtherDerived>&) */
     template<typename OtherDerived>
     EIGEN_STRONG_INLINE PlainObjectBase(const EigenBase<OtherDerived> &other)
-      : m_storage(Index(other.derived().rows()) * Index(other.derived().cols()), other.derived().rows(), other.derived().cols())
+      : m_storage(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
     {
       _check_template_params();
-      internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(other.derived().rows(), other.derived().cols());
+      internal::check_rows_cols_for_overflow(other.derived().rows(), other.derived().cols());
       Base::operator=(other.derived());
     }
 
@@ -603,8 +552,6 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
                  : (rows() == other.rows() && cols() == other.cols())))
         && "Size mismatch. Automatic resizing is disabled because EIGEN_NO_AUTOMATIC_RESIZING is defined");
       EIGEN_ONLY_USED_FOR_DEBUG(other);
-      if(this->size()==0)
-        resizeLike(other);
       #else
       resizeLike(other);
       #endif
@@ -654,19 +601,23 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
     }
 
     template<typename T0, typename T1>
-    EIGEN_STRONG_INLINE void _init2(Index nbRows, Index nbCols, typename internal::enable_if<Base::SizeAtCompileTime!=2,T0>::type* = 0)
+    EIGEN_STRONG_INLINE void _init2(Index rows, Index cols, typename internal::enable_if<Base::SizeAtCompileTime!=2,T0>::type* = 0)
     {
       EIGEN_STATIC_ASSERT(bool(NumTraits<T0>::IsInteger) &&
                           bool(NumTraits<T1>::IsInteger),
                           FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED)
-      resize(nbRows,nbCols);
+      eigen_assert(rows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows)
+             && cols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols));
+      internal::check_rows_cols_for_overflow(rows, cols);      
+      m_storage.resize(rows*cols,rows,cols);
+      EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
     }
     template<typename T0, typename T1>
-    EIGEN_STRONG_INLINE void _init2(const Scalar& val0, const Scalar& val1, typename internal::enable_if<Base::SizeAtCompileTime==2,T0>::type* = 0)
+    EIGEN_STRONG_INLINE void _init2(const Scalar& x, const Scalar& y, typename internal::enable_if<Base::SizeAtCompileTime==2,T0>::type* = 0)
     {
       EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2)
-      m_storage.data()[0] = val0;
-      m_storage.data()[1] = val1;
+      m_storage.data()[0] = x;
+      m_storage.data()[1] = y;
     }
 
     template<typename MatrixTypeA, typename MatrixTypeB, bool SwapPointers>
@@ -703,10 +654,8 @@ private:
     enum { ThisConstantIsPrivateInPlainObjectBase };
 };
 
-namespace internal {
-
 template <typename Derived, typename OtherDerived, bool IsVector>
-struct conservative_resize_like_impl
+struct internal::conservative_resize_like_impl
 {
   typedef typename Derived::Index Index;
   static void run(DenseBase<Derived>& _this, Index rows, Index cols)
@@ -717,7 +666,7 @@ struct conservative_resize_like_impl
     if ( ( Derived::IsRowMajor && _this.cols() == cols) || // row-major and we change only the number of rows
          (!Derived::IsRowMajor && _this.rows() == rows) )  // column-major and we change only the number of columns
     {
-      internal::check_rows_cols_for_overflow<Derived::MaxSizeAtCompileTime>::run(rows, cols);
+      internal::check_rows_cols_for_overflow(rows, cols);
       _this.derived().m_storage.conservativeResize(rows*cols,rows,cols);
     }
     else
@@ -766,14 +715,11 @@ struct conservative_resize_like_impl
   }
 };
 
-// Here, the specialization for vectors inherits from the general matrix case
-// to allow calling .conservativeResize(rows,cols) on vectors.
+namespace internal {
+
 template <typename Derived, typename OtherDerived>
 struct conservative_resize_like_impl<Derived,OtherDerived,true>
-  : conservative_resize_like_impl<Derived,OtherDerived,false>
 {
-  using conservative_resize_like_impl<Derived,OtherDerived,false>::run;
-  
   typedef typename Derived::Index Index;
   static void run(DenseBase<Derived>& _this, Index size)
   {

Eigen/src/Core/Product.h

@@ -0,0 +1,98 @@
+// This file is part of Eigen, a lightweight C++ template library
+// for linear algebra.
+//
+// Copyright (C) 2008-2011 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/.
+
+#ifndef EIGEN_PRODUCT_H
+#define EIGEN_PRODUCT_H
+
+template<typename Lhs, typename Rhs> class Product;
+template<typename Lhs, typename Rhs, typename StorageKind> class ProductImpl;
+
+/** \class Product
+  * \ingroup Core_Module
+  *
+  * \brief Expression of the product of two arbitrary matrices or vectors
+  *
+  * \param Lhs the type of the left-hand side expression
+  * \param Rhs the type of the right-hand side expression
+  *
+  * This class represents an expression of the product of two arbitrary matrices.
+  *
+  */
+
+namespace internal {
+template<typename Lhs, typename Rhs>
+struct traits<Product<Lhs, Rhs> >
+{
+  typedef MatrixXpr XprKind;
+  typedef typename remove_all<Lhs>::type LhsCleaned;
+  typedef typename remove_all<Rhs>::type RhsCleaned;
+  typedef typename scalar_product_traits<typename traits<LhsCleaned>::Scalar, typename traits<RhsCleaned>::Scalar>::ReturnType Scalar;
+  typedef typename promote_storage_type<typename traits<LhsCleaned>::StorageKind,
+                                        typename traits<RhsCleaned>::StorageKind>::ret StorageKind;
+  typedef typename promote_index_type<typename traits<LhsCleaned>::Index,
+                                      typename traits<RhsCleaned>::Index>::type Index;
+  enum {
+    RowsAtCompileTime = LhsCleaned::RowsAtCompileTime,
+    ColsAtCompileTime = RhsCleaned::ColsAtCompileTime,
+    MaxRowsAtCompileTime = LhsCleaned::MaxRowsAtCompileTime,
+    MaxColsAtCompileTime = RhsCleaned::MaxColsAtCompileTime,
+    Flags = (MaxRowsAtCompileTime==1 ? RowMajorBit : 0), // TODO should be no storage order
+    CoeffReadCost = 0 // TODO CoeffReadCost should not be part of the expression traits
+  };
+};
+} // end namespace internal
+
+
+template<typename Lhs, typename Rhs>
+class Product : public ProductImpl<Lhs,Rhs,typename internal::promote_storage_type<typename internal::traits<Lhs>::StorageKind,
+                                                                            typename internal::traits<Rhs>::StorageKind>::ret>
+{
+  public:
+    
+    typedef typename ProductImpl<
+        Lhs, Rhs,
+        typename internal::promote_storage_type<typename Lhs::StorageKind,
+                                                typename Rhs::StorageKind>::ret>::Base Base;
+    EIGEN_GENERIC_PUBLIC_INTERFACE(Product)
+
+    typedef typename Lhs::Nested LhsNested;
+    typedef typename Rhs::Nested RhsNested;
+    typedef typename internal::remove_all<LhsNested>::type LhsNestedCleaned;
+    typedef typename internal::remove_all<RhsNested>::type RhsNestedCleaned;
+
+    Product(const Lhs& lhs, const Rhs& rhs) : m_lhs(lhs), m_rhs(rhs)
+    {
+      eigen_assert(lhs.cols() == rhs.rows()
+        && "invalid matrix product"
+        && "if you wanted a coeff-wise or a dot product use the respective explicit functions");
+    }
+
+    inline Index rows() const { return m_lhs.rows(); }
+    inline Index cols() const { return m_rhs.cols(); }
+
+    const LhsNestedCleaned& lhs() const { return m_lhs; }
+    const RhsNestedCleaned& rhs() const { return m_rhs; }
+
+  protected:
+
+    const LhsNested m_lhs;
+    const RhsNested m_rhs;
+};
+
+template<typename Lhs, typename Rhs>
+class ProductImpl<Lhs,Rhs,Dense> : public internal::dense_xpr_base<Product<Lhs,Rhs> >::type
+{
+    typedef Product<Lhs, Rhs> Derived;
+  public:
+
+    typedef typename internal::dense_xpr_base<Product<Lhs, Rhs> >::type Base;
+    EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
+};
+
+#endif // EIGEN_PRODUCT_H

Eigen/src/Core/ProductBase.h

@@ -85,19 +85,12 @@ class ProductBase : public MatrixBase<Derived>
 
   public:
 
-#ifndef EIGEN_NO_MALLOC
-    typedef typename Base::PlainObject BasePlainObject;
-    typedef Matrix<Scalar,RowsAtCompileTime==1?1:Dynamic,ColsAtCompileTime==1?1:Dynamic,BasePlainObject::Options> DynPlainObject;
-    typedef typename internal::conditional<(BasePlainObject::SizeAtCompileTime==Dynamic) || (BasePlainObject::SizeAtCompileTime*int(sizeof(Scalar)) < int(EIGEN_STACK_ALLOCATION_LIMIT)),
-                                           BasePlainObject, DynPlainObject>::type PlainObject;
-#else
     typedef typename Base::PlainObject PlainObject;
-#endif
 
-    ProductBase(const Lhs& a_lhs, const Rhs& a_rhs)
-      : m_lhs(a_lhs), m_rhs(a_rhs)
+    ProductBase(const Lhs& lhs, const Rhs& rhs)
+      : m_lhs(lhs), m_rhs(rhs)
     {
-      eigen_assert(a_lhs.cols() == a_rhs.rows()
+      eigen_assert(lhs.cols() == rhs.rows()
         && "invalid matrix product"
         && "if you wanted a coeff-wise or a dot product use the respective explicit functions");
     }
@@ -115,7 +108,7 @@ class ProductBase : public MatrixBase<Derived>
     inline void subTo(Dest& dst) const { scaleAndAddTo(dst,Scalar(-1)); }
 
     template<typename Dest>
-    inline void scaleAndAddTo(Dest& dst, const Scalar& alpha) const { derived().scaleAndAddTo(dst,alpha); }
+    inline void scaleAndAddTo(Dest& dst,Scalar alpha) const { derived().scaleAndAddTo(dst,alpha); }
 
     const _LhsNested& lhs() const { return m_lhs; }
     const _RhsNested& rhs() const { return m_rhs; }
@@ -187,12 +180,7 @@ namespace internal {
 template<typename Lhs, typename Rhs, int Mode, int N, typename PlainObject>
 struct nested<GeneralProduct<Lhs,Rhs,Mode>, N, PlainObject>
 {
-  typedef typename GeneralProduct<Lhs,Rhs,Mode>::PlainObject const& type;
-};
-template<typename Lhs, typename Rhs, int Mode, int N, typename PlainObject>
-struct nested<const GeneralProduct<Lhs,Rhs,Mode>, N, PlainObject>
-{
-  typedef typename GeneralProduct<Lhs,Rhs,Mode>::PlainObject const& type;
+  typedef PlainObject const& type;
 };
 }
 
@@ -207,25 +195,25 @@ class ScaledProduct;
 // Also note that here we accept any compatible scalar types
 template<typename Derived,typename Lhs,typename Rhs>
 const ScaledProduct<Derived>
-operator*(const ProductBase<Derived,Lhs,Rhs>& prod, const typename Derived::Scalar& x)
+operator*(const ProductBase<Derived,Lhs,Rhs>& prod, typename Derived::Scalar x)
 { return ScaledProduct<Derived>(prod.derived(), x); }
 
 template<typename Derived,typename Lhs,typename Rhs>
 typename internal::enable_if<!internal::is_same<typename Derived::Scalar,typename Derived::RealScalar>::value,
                       const ScaledProduct<Derived> >::type
-operator*(const ProductBase<Derived,Lhs,Rhs>& prod, const typename Derived::RealScalar& x)
+operator*(const ProductBase<Derived,Lhs,Rhs>& prod, typename Derived::RealScalar x)
 { return ScaledProduct<Derived>(prod.derived(), x); }
 
 
 template<typename Derived,typename Lhs,typename Rhs>
 const ScaledProduct<Derived>
-operator*(const typename Derived::Scalar& x,const ProductBase<Derived,Lhs,Rhs>& prod)
+operator*(typename Derived::Scalar x,const ProductBase<Derived,Lhs,Rhs>& prod)
 { return ScaledProduct<Derived>(prod.derived(), x); }
 
 template<typename Derived,typename Lhs,typename Rhs>
 typename internal::enable_if<!internal::is_same<typename Derived::Scalar,typename Derived::RealScalar>::value,
                       const ScaledProduct<Derived> >::type
-operator*(const typename Derived::RealScalar& x,const ProductBase<Derived,Lhs,Rhs>& prod)
+operator*(typename Derived::RealScalar x,const ProductBase<Derived,Lhs,Rhs>& prod)
 { return ScaledProduct<Derived>(prod.derived(), x); }
 
 namespace internal {
@@ -253,7 +241,7 @@ class ScaledProduct
     typedef typename Base::PlainObject PlainObject;
 //     EIGEN_PRODUCT_PUBLIC_INTERFACE(ScaledProduct)
 
-    ScaledProduct(const NestedProduct& prod, const Scalar& x)
+    ScaledProduct(const NestedProduct& prod, Scalar x)
     : Base(prod.lhs(),prod.rhs()), m_prod(prod), m_alpha(x) {}
 
     template<typename Dest>
@@ -266,7 +254,7 @@ class ScaledProduct
     inline void subTo(Dest& dst) const { scaleAndAddTo(dst, Scalar(-1)); }
 
     template<typename Dest>
-    inline void scaleAndAddTo(Dest& dst, const Scalar& a_alpha) const { m_prod.derived().scaleAndAddTo(dst,a_alpha * m_alpha); }
+    inline void scaleAndAddTo(Dest& dst,Scalar alpha) const { m_prod.derived().scaleAndAddTo(dst,alpha * m_alpha); }
 
     const Scalar& alpha() const { return m_alpha; }
     

Eigen/src/Core/Random.h

@@ -112,7 +112,7 @@ inline Derived& DenseBase<Derived>::setRandom()
   return *this = Random(rows(), cols());
 }
 
-/** Resizes to the given \a newSize, and sets all coefficients in this expression to random values.
+/** Resizes to the given \a size, and sets all coefficients in this expression to random values.
   *
   * \only_for_vectors
   *
@@ -123,16 +123,16 @@ inline Derived& DenseBase<Derived>::setRandom()
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setRandom(Index newSize)
+PlainObjectBase<Derived>::setRandom(Index size)
 {
-  resize(newSize);
+  resize(size);
   return setRandom();
 }
 
 /** Resizes to the given size, and sets all coefficients in this expression to random values.
   *
-  * \param nbRows the new number of rows
-  * \param nbCols the new number of columns
+  * \param rows the new number of rows
+  * \param cols the new number of columns
   *
   * Example: \include Matrix_setRandom_int_int.cpp
   * Output: \verbinclude Matrix_setRandom_int_int.out
@@ -141,9 +141,9 @@ PlainObjectBase<Derived>::setRandom(Index newSize)
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE Derived&
-PlainObjectBase<Derived>::setRandom(Index nbRows, Index nbCols)
+PlainObjectBase<Derived>::setRandom(Index rows, Index cols)
 {
-  resize(nbRows, nbCols);
+  resize(rows, cols);
   return setRandom();
 }
 

Eigen/src/Core/Redux.h

@@ -247,9 +247,8 @@ struct redux_impl<Func, Derived, LinearVectorizedTraversal, NoUnrolling>
   }
 };
 
-// NOTE: for SliceVectorizedTraversal we simply bypass unrolling
-template<typename Func, typename Derived, int Unrolling>
-struct redux_impl<Func, Derived, SliceVectorizedTraversal, Unrolling>
+template<typename Func, typename Derived>
+struct redux_impl<Func, Derived, SliceVectorizedTraversal, NoUnrolling>
 {
   typedef typename Derived::Scalar Scalar;
   typedef typename packet_traits<Scalar>::type PacketScalar;
@@ -331,8 +330,7 @@ DenseBase<Derived>::redux(const Func& func) const
             ::run(derived(), func);
 }
 
-/** \returns the minimum of all coefficients of \c *this.
-  * \warning the result is undefined if \c *this contains NaN.
+/** \returns the minimum of all coefficients of *this
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
@@ -341,8 +339,7 @@ DenseBase<Derived>::minCoeff() const
   return this->redux(Eigen::internal::scalar_min_op<Scalar>());
 }
 
-/** \returns the maximum of all coefficients of \c *this.
-  * \warning the result is undefined if \c *this contains NaN.
+/** \returns the maximum of all coefficients of *this
   */
 template<typename Derived>
 EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar

Eigen/src/Core/Ref.h

@@ -1,278 +0,0 @@
-// This file is part of Eigen, a lightweight C++ template library
-// for linear algebra.
-//
-// Copyright (C) 2012 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/.
-
-#ifndef EIGEN_REF_H
-#define EIGEN_REF_H
-
-namespace Eigen { 
-
-template<typename Derived> class RefBase;
-template<typename PlainObjectType, int Options = 0,
-         typename StrideType = typename internal::conditional<PlainObjectType::IsVectorAtCompileTime,InnerStride<1>,OuterStride<> >::type > class Ref;
-
-/** \class Ref
-  * \ingroup Core_Module
-  *
-  * \brief A matrix or vector expression mapping an existing expressions
-  *
-  * \tparam PlainObjectType the equivalent matrix type of the mapped data
-  * \tparam Options specifies whether the pointer is \c #Aligned, or \c #Unaligned.
-  *                The default is \c #Unaligned.
-  * \tparam StrideType optionally specifies strides. By default, Ref implies a contiguous storage along the inner dimension (inner stride==1),
-  *                   but accept a variable outer stride (leading dimension).
-  *                   This can be overridden by specifying strides.
-  *                   The type passed here must be a specialization of the Stride template, see examples below.
-  *
-  * This class permits to write non template functions taking Eigen's object as parameters while limiting the number of copies.
-  * A Ref<> object can represent either a const expression or a l-value:
-  * \code
-  * // in-out argument:
-  * void foo1(Ref<VectorXf> x);
-  *
-  * // read-only const argument:
-  * void foo2(const Ref<const VectorXf>& x);
-  * \endcode
-  *
-  * In the in-out case, the input argument must satisfies the constraints of the actual Ref<> type, otherwise a compilation issue will be triggered.
-  * By default, a Ref<VectorXf> can reference any dense vector expression of float having a contiguous memory layout.
-  * Likewise, a Ref<MatrixXf> can reference any column major dense matrix expression of float whose column's elements are contiguously stored with
-  * the possibility to have a constant space inbetween each column, i.e.: the inner stride mmust be equal to 1, but the outer-stride (or leading dimension),
-  * can be greater than the number of rows.
-  *
-  * In the const case, if the input expression does not match the above requirement, then it is evaluated into a temporary before being passed to the function.
-  * Here are some examples:
-  * \code
-  * MatrixXf A;
-  * VectorXf a;
-  * foo1(a.head());             // OK
-  * foo1(A.col());              // OK
-  * foo1(A.row());              // compilation error because here innerstride!=1
-  * foo2(A.row());              // The row is copied into a contiguous temporary
-  * foo2(2*a);                  // The expression is evaluated into a temporary
-  * foo2(A.col().segment(2,4)); // No temporary
-  * \endcode
-  *
-  * The range of inputs that can be referenced without temporary can be enlarged using the last two template parameter.
-  * Here is an example accepting an innerstride!=1:
-  * \code
-  * // in-out argument:
-  * void foo3(Ref<VectorXf,0,InnerStride<> > x);
-  * foo3(A.row());              // OK
-  * \endcode
-  * The downside here is that the function foo3 might be significantly slower than foo1 because it won't be able to exploit vectorization, and will involved more
-  * expensive address computations even if the input is contiguously stored in memory. To overcome this issue, one might propose to overloads internally calling a
-  * template function, e.g.:
-  * \code
-  * // in the .h:
-  * void foo(const Ref<MatrixXf>& A);
-  * void foo(const Ref<MatrixXf,0,Stride<> >& A);
-  *
-  * // in the .cpp:
-  * template<typename TypeOfA> void foo_impl(const TypeOfA& A) {
-  *     ... // crazy code goes here
-  * }
-  * void foo(const Ref<MatrixXf>& A) { foo_impl(A); }
-  * void foo(const Ref<MatrixXf,0,Stride<> >& A) { foo_impl(A); }
-  * \endcode
-  *
-  *
-  * \sa PlainObjectBase::Map(), \ref TopicStorageOrders
-  */
-
-namespace internal {
-
-template<typename _PlainObjectType, int _Options, typename _StrideType>
-struct traits<Ref<_PlainObjectType, _Options, _StrideType> >
-  : public traits<Map<_PlainObjectType, _Options, _StrideType> >
-{
-  typedef _PlainObjectType PlainObjectType;
-  typedef _StrideType StrideType;
-  enum {
-    Options = _Options,
-    Flags = traits<Map<_PlainObjectType, _Options, _StrideType> >::Flags | NestByRefBit
-  };
-
-  template<typename Derived> struct match {
-    enum {
-      HasDirectAccess = internal::has_direct_access<Derived>::ret,
-      StorageOrderMatch = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)),
-      InnerStrideMatch = int(StrideType::InnerStrideAtCompileTime)==int(Dynamic)
-                      || int(StrideType::InnerStrideAtCompileTime)==int(Derived::InnerStrideAtCompileTime)
-                      || (int(StrideType::InnerStrideAtCompileTime)==0 && int(Derived::InnerStrideAtCompileTime)==1),
-      OuterStrideMatch = Derived::IsVectorAtCompileTime
-                      || int(StrideType::OuterStrideAtCompileTime)==int(Dynamic) || int(StrideType::OuterStrideAtCompileTime)==int(Derived::OuterStrideAtCompileTime),
-      AlignmentMatch = (_Options!=Aligned) || ((PlainObjectType::Flags&AlignedBit)==0) || ((traits<Derived>::Flags&AlignedBit)==AlignedBit),
-      ScalarTypeMatch = internal::is_same<typename PlainObjectType::Scalar, typename Derived::Scalar>::value,
-      MatchAtCompileTime = HasDirectAccess && StorageOrderMatch && InnerStrideMatch && OuterStrideMatch && AlignmentMatch && ScalarTypeMatch
-    };
-    typedef typename internal::conditional<MatchAtCompileTime,internal::true_type,internal::false_type>::type type;
-  };
-  
-};
-
-template<typename Derived>
-struct traits<RefBase<Derived> > : public traits<Derived> {};
-
-}
-
-template<typename Derived> class RefBase
- : public MapBase<Derived>
-{
-  typedef typename internal::traits<Derived>::PlainObjectType PlainObjectType;
-  typedef typename internal::traits<Derived>::StrideType StrideType;
-
-public:
-
-  typedef MapBase<Derived> Base;
-  EIGEN_DENSE_PUBLIC_INTERFACE(RefBase)
-
-  inline Index innerStride() const
-  {
-    return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1;
-  }
-
-  inline Index outerStride() const
-  {
-    return StrideType::OuterStrideAtCompileTime != 0 ? m_stride.outer()
-         : IsVectorAtCompileTime ? this->size()
-         : int(Flags)&RowMajorBit ? this->cols()
-         : this->rows();
-  }
-
-  RefBase()
-    : Base(0,RowsAtCompileTime==Dynamic?0:RowsAtCompileTime,ColsAtCompileTime==Dynamic?0:ColsAtCompileTime),
-      // Stride<> does not allow default ctor for Dynamic strides, so let' initialize it with dummy values:
-      m_stride(StrideType::OuterStrideAtCompileTime==Dynamic?0:StrideType::OuterStrideAtCompileTime,
-               StrideType::InnerStrideAtCompileTime==Dynamic?0:StrideType::InnerStrideAtCompileTime)
-  {}
-  
-  EIGEN_INHERIT_ASSIGNMENT_OPERATORS(RefBase)
-
-protected:
-
-  typedef Stride<StrideType::OuterStrideAtCompileTime,StrideType::InnerStrideAtCompileTime> StrideBase;
-
-  template<typename Expression>
-  void construct(Expression& expr)
-  {
-    if(PlainObjectType::RowsAtCompileTime==1)
-    {
-      eigen_assert(expr.rows()==1 || expr.cols()==1);
-      ::new (static_cast<Base*>(this)) Base(expr.data(), 1, expr.size());
-    }
-    else if(PlainObjectType::ColsAtCompileTime==1)
-    {
-      eigen_assert(expr.rows()==1 || expr.cols()==1);
-      ::new (static_cast<Base*>(this)) Base(expr.data(), expr.size(), 1);
-    }
-    else
-      ::new (static_cast<Base*>(this)) Base(expr.data(), expr.rows(), expr.cols());
-    
-    if(Expression::IsVectorAtCompileTime && (!PlainObjectType::IsVectorAtCompileTime) && ((Expression::Flags&RowMajorBit)!=(PlainObjectType::Flags&RowMajorBit)))
-      ::new (&m_stride) StrideBase(expr.innerStride(), StrideType::InnerStrideAtCompileTime==0?0:1);
-    else
-      ::new (&m_stride) StrideBase(StrideType::OuterStrideAtCompileTime==0?0:expr.outerStride(),
-                                   StrideType::InnerStrideAtCompileTime==0?0:expr.innerStride());    
-  }
-
-  StrideBase m_stride;
-};
-
-
-template<typename PlainObjectType, int Options, typename StrideType> class Ref
-  : public RefBase<Ref<PlainObjectType, Options, StrideType> >
-{
-  private:
-    typedef internal::traits<Ref> Traits;
-    template<typename Derived>
-    inline Ref(const PlainObjectBase<Derived>& expr,
-               typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0);
-  public:
-
-    typedef RefBase<Ref> Base;
-    EIGEN_DENSE_PUBLIC_INTERFACE(Ref)
-
-
-    #ifndef EIGEN_PARSED_BY_DOXYGEN
-    template<typename Derived>
-    inline Ref(PlainObjectBase<Derived>& expr,
-               typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0)
-    {
-      EIGEN_STATIC_ASSERT(static_cast<bool>(Traits::template match<Derived>::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH);
-      Base::construct(expr.derived());
-    }
-    template<typename Derived>
-    inline Ref(const DenseBase<Derived>& expr,
-               typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0)
-    #else
-    template<typename Derived>
-    inline Ref(DenseBase<Derived>& expr)
-    #endif
-    {
-      EIGEN_STATIC_ASSERT(static_cast<bool>(internal::is_lvalue<Derived>::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY);
-      EIGEN_STATIC_ASSERT(static_cast<bool>(Traits::template match<Derived>::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH);
-      enum { THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY = Derived::ThisConstantIsPrivateInPlainObjectBase};
-      Base::construct(expr.const_cast_derived());
-    }
-
-    EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Ref)
-
-};
-
-// this is the const ref version
-template<typename TPlainObjectType, int Options, typename StrideType> class Ref<const TPlainObjectType, Options, StrideType>
-  : public RefBase<Ref<const TPlainObjectType, Options, StrideType> >
-{
-    typedef internal::traits<Ref> Traits;
-  public:
-
-    typedef RefBase<Ref> Base;
-    EIGEN_DENSE_PUBLIC_INTERFACE(Ref)
-
-    template<typename Derived>
-    inline Ref(const DenseBase<Derived>& expr,
-               typename internal::enable_if<bool(Traits::template match<Derived>::ScalarTypeMatch),Derived>::type* = 0)
-    {
-//      std::cout << match_helper<Derived>::HasDirectAccess << "," << match_helper<Derived>::OuterStrideMatch << "," << match_helper<Derived>::InnerStrideMatch << "\n";
-//      std::cout << int(StrideType::OuterStrideAtCompileTime) << " - " << int(Derived::OuterStrideAtCompileTime) << "\n";
-//      std::cout << int(StrideType::InnerStrideAtCompileTime) << " - " << int(Derived::InnerStrideAtCompileTime) << "\n";
-      construct(expr.derived(), typename Traits::template match<Derived>::type());
-    }
-    
-    inline Ref(const Ref& other) : Base(other) {
-      // copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy
-    }
-
-    template<typename OtherRef>
-    inline Ref(const RefBase<OtherRef>& other) {
-      construct(other.derived(), typename Traits::template match<OtherRef>::type());
-    }
-
-  protected:
-
-    template<typename Expression>
-    void construct(const Expression& expr,internal::true_type)
-    {
-      Base::construct(expr);
-    }
-
-    template<typename Expression>
-    void construct(const Expression& expr, internal::false_type)
-    {
-      m_object.lazyAssign(expr);
-      Base::construct(m_object);
-    }
-
-  protected:
-    TPlainObjectType m_object;
-};
-
-} // end namespace Eigen
-
-#endif // EIGEN_REF_H

Eigen/src/Core/Replicate.h

@@ -70,8 +70,8 @@ template<typename MatrixType,int RowFactor,int ColFactor> class Replicate
     EIGEN_DENSE_PUBLIC_INTERFACE(Replicate)
 
     template<typename OriginalMatrixType>
-    inline explicit Replicate(const OriginalMatrixType& a_matrix)
-      : m_matrix(a_matrix), m_rowFactor(RowFactor), m_colFactor(ColFactor)
+    inline explicit Replicate(const OriginalMatrixType& matrix)
+      : m_matrix(matrix), m_rowFactor(RowFactor), m_colFactor(ColFactor)
     {
       EIGEN_STATIC_ASSERT((internal::is_same<typename internal::remove_const<MatrixType>::type,OriginalMatrixType>::value),
                           THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE)
@@ -79,8 +79,8 @@ template<typename MatrixType,int RowFactor,int ColFactor> class Replicate
     }
 
     template<typename OriginalMatrixType>
-    inline Replicate(const OriginalMatrixType& a_matrix, Index rowFactor, Index colFactor)
-      : m_matrix(a_matrix), m_rowFactor(rowFactor), m_colFactor(colFactor)
+    inline Replicate(const OriginalMatrixType& matrix, Index rowFactor, Index colFactor)
+      : m_matrix(matrix), m_rowFactor(rowFactor), m_colFactor(colFactor)
     {
       EIGEN_STATIC_ASSERT((internal::is_same<typename internal::remove_const<MatrixType>::type,OriginalMatrixType>::value),
                           THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE)
@@ -89,27 +89,27 @@ template<typename MatrixType,int RowFactor,int ColFactor> class Replicate
     inline Index rows() const { return m_matrix.rows() * m_rowFactor.value(); }
     inline Index cols() const { return m_matrix.cols() * m_colFactor.value(); }
 
-    inline Scalar coeff(Index rowId, Index colId) const
+    inline Scalar coeff(Index row, Index col) const
     {
       // try to avoid using modulo; this is a pure optimization strategy
       const Index actual_row  = internal::traits<MatrixType>::RowsAtCompileTime==1 ? 0
-                            : RowFactor==1 ? rowId
-                            : rowId%m_matrix.rows();
+                            : RowFactor==1 ? row
+                            : row%m_matrix.rows();
       const Index actual_col  = internal::traits<MatrixType>::ColsAtCompileTime==1 ? 0
-                            : ColFactor==1 ? colId
-                            : colId%m_matrix.cols();
+                            : ColFactor==1 ? col
+                            : col%m_matrix.cols();
 
       return m_matrix.coeff(actual_row, actual_col);
     }
     template<int LoadMode>
-    inline PacketScalar packet(Index rowId, Index colId) const
+    inline PacketScalar packet(Index row, Index col) const
     {
       const Index actual_row  = internal::traits<MatrixType>::RowsAtCompileTime==1 ? 0
-                            : RowFactor==1 ? rowId
-                            : rowId%m_matrix.rows();
+                            : RowFactor==1 ? row
+                            : row%m_matrix.rows();
       const Index actual_col  = internal::traits<MatrixType>::ColsAtCompileTime==1 ? 0
-                            : ColFactor==1 ? colId
-                            : colId%m_matrix.cols();
+                            : ColFactor==1 ? col
+                            : col%m_matrix.cols();
 
       return m_matrix.template packet<LoadMode>(actual_row, actual_col);
     }
@@ -135,7 +135,7 @@ template<typename MatrixType,int RowFactor,int ColFactor> class Replicate
   */
 template<typename Derived>
 template<int RowFactor, int ColFactor>
-const Replicate<Derived,RowFactor,ColFactor>
+inline const Replicate<Derived,RowFactor,ColFactor>
 DenseBase<Derived>::replicate() const
 {
   return Replicate<Derived,RowFactor,ColFactor>(derived());
@@ -150,7 +150,7 @@ DenseBase<Derived>::replicate() const
   * \sa VectorwiseOp::replicate(), DenseBase::replicate<int,int>(), class Replicate
   */
 template<typename Derived>
-const typename DenseBase<Derived>::ReplicateReturnType
+inline const Replicate<Derived,Dynamic,Dynamic>
 DenseBase<Derived>::replicate(Index rowFactor,Index colFactor) const
 {
   return Replicate<Derived,Dynamic,Dynamic>(derived(),rowFactor,colFactor);

Eigen/src/Core/ReturnByValue.h

@@ -48,7 +48,7 @@ struct nested<ReturnByValue<Derived>, n, PlainObject>
 } // end namespace internal
 
 template<typename Derived> class ReturnByValue
-  : internal::no_assignment_operator, public internal::dense_xpr_base< ReturnByValue<Derived> >::type
+  : public internal::dense_xpr_base< ReturnByValue<Derived> >::type
 {
   public:
     typedef typename internal::traits<Derived>::ReturnType ReturnType;
@@ -72,8 +72,6 @@ template<typename Derived> class ReturnByValue
     const Unusable& coeff(Index,Index) const { return *reinterpret_cast<const Unusable*>(this); }
     Unusable& coeffRef(Index) { return *reinterpret_cast<Unusable*>(this); }
     Unusable& coeffRef(Index,Index) { return *reinterpret_cast<Unusable*>(this); }
-    template<int LoadMode>  Unusable& packet(Index) const;
-    template<int LoadMode>  Unusable& packet(Index, Index) const;
 #endif
 };
 
@@ -85,15 +83,6 @@ Derived& DenseBase<Derived>::operator=(const ReturnByValue<OtherDerived>& other)
   return derived();
 }
 
-template<typename Derived>
-template<typename OtherDerived>
-Derived& DenseBase<Derived>::lazyAssign(const ReturnByValue<OtherDerived>& other)
-{
-  other.evalTo(derived());
-  return derived();
-}
-
-
 } // end namespace Eigen
 
 #endif // EIGEN_RETURNBYVALUE_H

Eigen/src/Core/Reverse.h

@@ -76,23 +76,9 @@ template<typename MatrixType, int Direction> class Reverse
     EIGEN_DENSE_PUBLIC_INTERFACE(Reverse)
     using Base::IsRowMajor;
 
-    // The following two operators are provided to worarkound
-    // a MSVC 2013 issue. In theory, we could simply do:
-    //   using Base::operator(); 
-    // to make const version of operator() visible.
-    // Otheriwse, they would be hidden by the non-const versions defined in this file
-    
-    inline CoeffReturnType operator()(Index row, Index col) const
-    {
-      eigen_assert(row >= 0 && row < rows() && col >= 0 && col < cols());
-      return coeff(row, col);
-    }
-
-    inline CoeffReturnType operator()(Index index) const
-    {
-      eigen_assert(index >= 0 && index < m_matrix.size());
-      return coeff(index);
-    }
+    // next line is necessary because otherwise const version of operator()
+    // is hidden by non-const version defined in this file
+    using Base::operator(); 
 
   protected:
     enum {

Eigen/src/Core/Select.h

@@ -60,10 +60,10 @@ class Select : internal::no_assignment_operator,
     typedef typename internal::dense_xpr_base<Select>::type Base;
     EIGEN_DENSE_PUBLIC_INTERFACE(Select)
 
-    Select(const ConditionMatrixType& a_conditionMatrix,
-           const ThenMatrixType& a_thenMatrix,
-           const ElseMatrixType& a_elseMatrix)
-      : m_condition(a_conditionMatrix), m_then(a_thenMatrix), m_else(a_elseMatrix)
+    Select(const ConditionMatrixType& conditionMatrix,
+           const ThenMatrixType& thenMatrix,
+           const ElseMatrixType& elseMatrix)
+      : m_condition(conditionMatrix), m_then(thenMatrix), m_else(elseMatrix)
     {
       eigen_assert(m_condition.rows() == m_then.rows() && m_condition.rows() == m_else.rows());
       eigen_assert(m_condition.cols() == m_then.cols() && m_condition.cols() == m_else.cols());
@@ -136,7 +136,7 @@ template<typename Derived>
 template<typename ThenDerived>
 inline const Select<Derived,ThenDerived, typename ThenDerived::ConstantReturnType>
 DenseBase<Derived>::select(const DenseBase<ThenDerived>& thenMatrix,
-                           const typename ThenDerived::Scalar& elseScalar) const
+                            typename ThenDerived::Scalar elseScalar) const
 {
   return Select<Derived,ThenDerived,typename ThenDerived::ConstantReturnType>(
     derived(), thenMatrix.derived(), ThenDerived::Constant(rows(),cols(),elseScalar));
@@ -150,8 +150,8 @@ DenseBase<Derived>::select(const DenseBase<ThenDerived>& thenMatrix,
 template<typename Derived>
 template<typename ElseDerived>
 inline const Select<Derived, typename ElseDerived::ConstantReturnType, ElseDerived >
-DenseBase<Derived>::select(const typename ElseDerived::Scalar& thenScalar,
-                           const DenseBase<ElseDerived>& elseMatrix) const
+DenseBase<Derived>::select(typename ElseDerived::Scalar thenScalar,
+                            const DenseBase<ElseDerived>& elseMatrix) const
 {
   return Select<Derived,typename ElseDerived::ConstantReturnType,ElseDerived>(
     derived(), ElseDerived::Constant(rows(),cols(),thenScalar), elseMatrix.derived());

Eigen/src/Core/SelfAdjointView.h

@@ -132,7 +132,7 @@ template<typename MatrixType, unsigned int UpLo> class SelfAdjointView
       * \sa rankUpdate(const MatrixBase<DerivedU>&, Scalar)
       */
     template<typename DerivedU, typename DerivedV>
-    SelfAdjointView& rankUpdate(const MatrixBase<DerivedU>& u, const MatrixBase<DerivedV>& v, const Scalar& alpha = Scalar(1));
+    SelfAdjointView& rankUpdate(const MatrixBase<DerivedU>& u, const MatrixBase<DerivedV>& v, Scalar alpha = Scalar(1));
 
     /** Perform a symmetric rank K update of the selfadjoint matrix \c *this:
       * \f$ this = this + \alpha ( u u^* ) \f$ where \a u is a vector or matrix.
@@ -145,7 +145,7 @@ template<typename MatrixType, unsigned int UpLo> class SelfAdjointView
       * \sa rankUpdate(const MatrixBase<DerivedU>&, const MatrixBase<DerivedV>&, Scalar)
       */
     template<typename DerivedU>
-    SelfAdjointView& rankUpdate(const MatrixBase<DerivedU>& u, const Scalar& alpha = Scalar(1));
+    SelfAdjointView& rankUpdate(const MatrixBase<DerivedU>& u, Scalar alpha = Scalar(1));
 
 /////////// Cholesky module ///////////
 
@@ -214,9 +214,9 @@ struct triangular_assignment_selector<Derived1, Derived2, (SelfAdjoint|Upper), U
     triangular_assignment_selector<Derived1, Derived2, (SelfAdjoint|Upper), UnrollCount-1, ClearOpposite>::run(dst, src);
 
     if(row == col)
-      dst.coeffRef(row, col) = numext::real(src.coeff(row, col));
+      dst.coeffRef(row, col) = real(src.coeff(row, col));
     else if(row < col)
-      dst.coeffRef(col, row) = numext::conj(dst.coeffRef(row, col) = src.coeff(row, col));
+      dst.coeffRef(col, row) = conj(dst.coeffRef(row, col) = src.coeff(row, col));
   }
 };
 
@@ -239,9 +239,9 @@ struct triangular_assignment_selector<Derived1, Derived2, (SelfAdjoint|Lower), U
     triangular_assignment_selector<Derived1, Derived2, (SelfAdjoint|Lower), UnrollCount-1, ClearOpposite>::run(dst, src);
 
     if(row == col)
-      dst.coeffRef(row, col) = numext::real(src.coeff(row, col));
+      dst.coeffRef(row, col) = real(src.coeff(row, col));
     else if(row > col)
-      dst.coeffRef(col, row) = numext::conj(dst.coeffRef(row, col) = src.coeff(row, col));
+      dst.coeffRef(col, row) = conj(dst.coeffRef(row, col) = src.coeff(row, col));
   }
 };
 
@@ -262,7 +262,7 @@ struct triangular_assignment_selector<Derived1, Derived2, SelfAdjoint|Upper, Dyn
       for(Index i = 0; i < j; ++i)
       {
         dst.copyCoeff(i, j, src);
-        dst.coeffRef(j,i) = numext::conj(dst.coeff(i,j));
+        dst.coeffRef(j,i) = conj(dst.coeff(i,j));
       }
       dst.copyCoeff(j, j, src);
     }
@@ -280,7 +280,7 @@ struct triangular_assignment_selector<Derived1, Derived2, SelfAdjoint|Lower, Dyn
       for(Index j = 0; j < i; ++j)
       {
         dst.copyCoeff(i, j, src);
-        dst.coeffRef(j,i) = numext::conj(dst.coeff(i,j));
+        dst.coeffRef(j,i) = conj(dst.coeff(i,j));
       }
       dst.copyCoeff(i, i, src);
     }

Eigen/src/Core/SelfCwiseBinaryOp.h

@@ -180,9 +180,12 @@ inline Derived& DenseBase<Derived>::operator*=(const Scalar& other)
 template<typename Derived>
 inline Derived& DenseBase<Derived>::operator/=(const Scalar& other)
 {
+  typedef typename internal::conditional<NumTraits<Scalar>::IsInteger,
+                                        internal::scalar_quotient_op<Scalar>,
+                                        internal::scalar_product_op<Scalar> >::type BinOp;
   typedef typename Derived::PlainObject PlainObject;
-  SelfCwiseBinaryOp<internal::scalar_quotient_op<Scalar>, Derived, typename PlainObject::ConstantReturnType> tmp(derived());
-  tmp = PlainObject::Constant(rows(),cols(), other);
+  SelfCwiseBinaryOp<BinOp, Derived, typename PlainObject::ConstantReturnType> tmp(derived());
+  tmp = PlainObject::Constant(rows(),cols(), NumTraits<Scalar>::IsInteger ? other : Scalar(1)/other);
   return derived();
 }
 

Eigen/src/Core/SolveTriangular.h

@@ -116,17 +116,17 @@ template<typename Lhs, typename Rhs, int Mode, int Index, int Size>
 struct triangular_solver_unroller<Lhs,Rhs,Mode,Index,Size,false> {
   enum {
     IsLower = ((Mode&Lower)==Lower),
-    RowIndex = IsLower ? Index : Size - Index - 1,
-    S = IsLower ? 0     : RowIndex+1
+    I = IsLower ? Index : Size - Index - 1,
+    S = IsLower ? 0     : I+1
   };
   static void run(const Lhs& lhs, Rhs& rhs)
   {
     if (Index>0)
-      rhs.coeffRef(RowIndex) -= lhs.row(RowIndex).template segment<Index>(S).transpose()
+      rhs.coeffRef(I) -= lhs.row(I).template segment<Index>(S).transpose()
                          .cwiseProduct(rhs.template segment<Index>(S)).sum();
 
     if(!(Mode & UnitDiag))
-      rhs.coeffRef(RowIndex) /= lhs.coeff(RowIndex,RowIndex);
+      rhs.coeffRef(I) /= lhs.coeff(I,I);
 
     triangular_solver_unroller<Lhs,Rhs,Mode,Index+1,Size>::run(lhs,rhs);
   }
@@ -243,8 +243,7 @@ template<int Side, typename TriangularType, typename Rhs> struct triangular_solv
 
   template<typename Dest> inline void evalTo(Dest& dst) const
   {
-    const typename Dest::Scalar *dst_data = internal::extract_data(dst);
-    if(!(is_same<RhsNestedCleaned,Dest>::value && dst_data!=0 && extract_data(dst) == extract_data(m_rhs)))
+    if(!(is_same<RhsNestedCleaned,Dest>::value && extract_data(dst) == extract_data(m_rhs)))
       dst = m_rhs;
     m_triangularMatrix.template solveInPlace<Side>(dst);
   }

Eigen/src/Core/StableNorm.h

@@ -17,127 +17,19 @@ namespace internal {
 template<typename ExpressionType, typename Scalar>
 inline void stable_norm_kernel(const ExpressionType& bl, Scalar& ssq, Scalar& scale, Scalar& invScale)
 {
-  using std::max;
-  Scalar maxCoeff = bl.cwiseAbs().maxCoeff();
-  
-  if (maxCoeff>scale)
+  Scalar max = bl.cwiseAbs().maxCoeff();
+  if (max>scale)
   {
-    ssq = ssq * numext::abs2(scale/maxCoeff);
-    Scalar tmp = Scalar(1)/maxCoeff;
-    if(tmp > NumTraits<Scalar>::highest())
-    {
-      invScale = NumTraits<Scalar>::highest();
-      scale = Scalar(1)/invScale;
-    }
-    else
-    {
-      scale = maxCoeff;
-      invScale = tmp;
-    }
+    ssq = ssq * abs2(scale/max);
+    scale = max;
+    invScale = Scalar(1)/scale;
   }
-  
-  // TODO if the maxCoeff is much much smaller than the current scale,
+  // TODO if the max is much much smaller than the current scale,
   // then we can neglect this sub vector
-  if(scale>Scalar(0)) // if scale==0, then bl is 0 
-    ssq += (bl*invScale).squaredNorm();
+  ssq += (bl*invScale).squaredNorm();
 }
-
-template<typename Derived>
-inline typename NumTraits<typename traits<Derived>::Scalar>::Real
-blueNorm_impl(const EigenBase<Derived>& _vec)
-{
-  typedef typename Derived::RealScalar RealScalar;  
-  typedef typename Derived::Index Index;
-  using std::pow;
-  using std::min;
-  using std::max;
-  using std::sqrt;
-  using std::abs;
-  const Derived& vec(_vec.derived());
-  static bool initialized = false;
-  static RealScalar b1, b2, s1m, s2m, overfl, rbig, relerr;
-  if(!initialized)
-  {
-    int ibeta, it, iemin, iemax, iexp;
-    RealScalar eps;
-    // This program calculates the machine-dependent constants
-    // bl, b2, slm, s2m, relerr overfl
-    // from the "basic" machine-dependent numbers
-    // nbig, ibeta, it, iemin, iemax, rbig.
-    // The following define the basic machine-dependent constants.
-    // For portability, the PORT subprograms "ilmaeh" and "rlmach"
-    // are used. For any specific computer, each of the assignment
-    // statements can be replaced
-    ibeta = std::numeric_limits<RealScalar>::radix;                 // base for floating-point numbers
-    it    = std::numeric_limits<RealScalar>::digits;                // number of base-beta digits in mantissa
-    iemin = std::numeric_limits<RealScalar>::min_exponent;          // minimum exponent
-    iemax = std::numeric_limits<RealScalar>::max_exponent;          // maximum exponent
-    rbig  = (std::numeric_limits<RealScalar>::max)();               // largest floating-point number
-
-    iexp  = -((1-iemin)/2);
-    b1    = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));    // lower boundary of midrange
-    iexp  = (iemax + 1 - it)/2;
-    b2    = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));    // upper boundary of midrange
-
-    iexp  = (2-iemin)/2;
-    s1m   = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));    // scaling factor for lower range
-    iexp  = - ((iemax+it)/2);
-    s2m   = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));    // scaling factor for upper range
-
-    overfl  = rbig*s2m;                                             // overflow boundary for abig
-    eps     = RealScalar(pow(double(ibeta), 1-it));
-    relerr  = sqrt(eps);                                            // tolerance for neglecting asml
-    initialized = true;
-  }
-  Index n = vec.size();
-  RealScalar ab2 = b2 / RealScalar(n);
-  RealScalar asml = RealScalar(0);
-  RealScalar amed = RealScalar(0);
-  RealScalar abig = RealScalar(0);
-  for(typename Derived::InnerIterator it(vec, 0); it; ++it)
-  {
-    RealScalar ax = abs(it.value());
-    if(ax > ab2)     abig += numext::abs2(ax*s2m);
-    else if(ax < b1) asml += numext::abs2(ax*s1m);
-    else             amed += numext::abs2(ax);
-  }
-  if(abig > RealScalar(0))
-  {
-    abig = sqrt(abig);
-    if(abig > overfl)
-    {
-      return rbig;
-    }
-    if(amed > RealScalar(0))
-    {
-      abig = abig/s2m;
-      amed = sqrt(amed);
-    }
-    else
-      return abig/s2m;
-  }
-  else if(asml > RealScalar(0))
-  {
-    if (amed > RealScalar(0))
-    {
-      abig = sqrt(amed);
-      amed = sqrt(asml) / s1m;
-    }
-    else
-      return sqrt(asml)/s1m;
-  }
-  else
-    return sqrt(amed);
-  asml = (min)(abig, amed);
-  abig = (max)(abig, amed);
-  if(asml <= abig*relerr)
-    return abig;
-  else
-    return abig * sqrt(RealScalar(1) + numext::abs2(asml/abig));
 }
 
-} // end namespace internal
-
 /** \returns the \em l2 norm of \c *this avoiding underflow and overflow.
   * This version use a blockwise two passes algorithm:
   *  1 - find the absolute largest coefficient \c s
@@ -153,7 +45,6 @@ inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real
 MatrixBase<Derived>::stableNorm() const
 {
   using std::min;
-  using std::sqrt;
   const Index blockSize = 4096;
   RealScalar scale(0);
   RealScalar invScale(1);
@@ -167,7 +58,7 @@ MatrixBase<Derived>::stableNorm() const
     internal::stable_norm_kernel(this->head(bi), ssq, scale, invScale);
   for (; bi<n; bi+=blockSize)
     internal::stable_norm_kernel(this->segment(bi,(min)(blockSize, n - bi)).template forceAlignedAccessIf<Alignment>(), ssq, scale, invScale);
-  return scale * sqrt(ssq);
+  return scale * internal::sqrt(ssq);
 }
 
 /** \returns the \em l2 norm of \c *this using the Blue's algorithm.
@@ -183,7 +74,90 @@ template<typename Derived>
 inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real
 MatrixBase<Derived>::blueNorm() const
 {
-  return internal::blueNorm_impl(*this);
+  using std::pow;
+  using std::min;
+  using std::max;
+  static bool initialized = false;
+  static RealScalar b1, b2, s1m, s2m, overfl, rbig, relerr;
+  if(!initialized)
+  {
+    int ibeta, it, iemin, iemax, iexp;
+    RealScalar abig, eps;
+    // This program calculates the machine-dependent constants
+    // bl, b2, slm, s2m, relerr overfl
+    // from the "basic" machine-dependent numbers
+    // ibeta, it, iemin, iemax, rbig.
+    // The following define the basic machine-dependent constants.
+    // For portability, the PORT subprograms "ilmaeh" and "rlmach"
+    // are used. For any specific computer, each of the assignment
+    // statements can be replaced
+    ibeta = std::numeric_limits<RealScalar>::radix;         // base for floating-point numbers
+    it    = std::numeric_limits<RealScalar>::digits;        // number of base-beta digits in mantissa
+    iemin = std::numeric_limits<RealScalar>::min_exponent;  // minimum exponent
+    iemax = std::numeric_limits<RealScalar>::max_exponent;  // maximum exponent
+    rbig  = (std::numeric_limits<RealScalar>::max)();         // largest floating-point number
+
+    iexp  = -((1-iemin)/2);
+    b1    = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));  // lower boundary of midrange
+    iexp  = (iemax + 1 - it)/2;
+    b2    = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));   // upper boundary of midrange
+
+    iexp  = (2-iemin)/2;
+    s1m   = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));   // scaling factor for lower range
+    iexp  = - ((iemax+it)/2);
+    s2m   = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp)));   // scaling factor for upper range
+
+    overfl  = rbig*s2m;             // overflow boundary for abig
+    eps     = RealScalar(pow(double(ibeta), 1-it));
+    relerr  = internal::sqrt(eps);         // tolerance for neglecting asml
+    abig    = RealScalar(1.0/eps - 1.0);
+    initialized = true;
+  }
+  Index n = size();
+  RealScalar ab2 = b2 / RealScalar(n);
+  RealScalar asml = RealScalar(0);
+  RealScalar amed = RealScalar(0);
+  RealScalar abig = RealScalar(0);
+  for(Index j=0; j<n; ++j)
+  {
+    RealScalar ax = internal::abs(coeff(j));
+    if(ax > ab2)     abig += internal::abs2(ax*s2m);
+    else if(ax < b1) asml += internal::abs2(ax*s1m);
+    else             amed += internal::abs2(ax);
+  }
+  if(abig > RealScalar(0))
+  {
+    abig = internal::sqrt(abig);
+    if(abig > overfl)
+    {
+      return rbig;
+    }
+    if(amed > RealScalar(0))
+    {
+      abig = abig/s2m;
+      amed = internal::sqrt(amed);
+    }
+    else
+      return abig/s2m;
+  }
+  else if(asml > RealScalar(0))
+  {
+    if (amed > RealScalar(0))
+    {
+      abig = internal::sqrt(amed);
+      amed = internal::sqrt(asml) / s1m;
+    }
+    else
+      return internal::sqrt(asml)/s1m;
+  }
+  else
+    return internal::sqrt(amed);
+  asml = (min)(abig, amed);
+  abig = (max)(abig, amed);
+  if(asml <= abig*relerr)
+    return abig;
+  else
+    return abig * internal::sqrt(RealScalar(1) + internal::abs2(asml/abig));
 }
 
 /** \returns the \em l2 norm of \c *this avoiding undeflow and overflow.

Eigen/src/Core/Swap.h

@@ -49,9 +49,9 @@ template<typename ExpressionType> class SwapWrapper
     inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); }
     inline const Scalar* data() const { return m_expression.data(); }
 
-    inline Scalar& coeffRef(Index rowId, Index colId)
+    inline Scalar& coeffRef(Index row, Index col)
     {
-      return m_expression.const_cast_derived().coeffRef(rowId, colId);
+      return m_expression.const_cast_derived().coeffRef(row, col);
     }
 
     inline Scalar& coeffRef(Index index)
@@ -59,9 +59,9 @@ template<typename ExpressionType> class SwapWrapper
       return m_expression.const_cast_derived().coeffRef(index);
     }
 
-    inline Scalar& coeffRef(Index rowId, Index colId) const
+    inline Scalar& coeffRef(Index row, Index col) const
     {
-      return m_expression.coeffRef(rowId, colId);
+      return m_expression.coeffRef(row, col);
     }
 
     inline Scalar& coeffRef(Index index) const
@@ -70,14 +70,14 @@ template<typename ExpressionType> class SwapWrapper
     }
 
     template<typename OtherDerived>
-    void copyCoeff(Index rowId, Index colId, const DenseBase<OtherDerived>& other)
+    void copyCoeff(Index row, Index col, const DenseBase<OtherDerived>& other)
     {
       OtherDerived& _other = other.const_cast_derived();
-      eigen_internal_assert(rowId >= 0 && rowId < rows()
-                         && colId >= 0 && colId < cols());
-      Scalar tmp = m_expression.coeff(rowId, colId);
-      m_expression.coeffRef(rowId, colId) = _other.coeff(rowId, colId);
-      _other.coeffRef(rowId, colId) = tmp;
+      eigen_internal_assert(row >= 0 && row < rows()
+                         && col >= 0 && col < cols());
+      Scalar tmp = m_expression.coeff(row, col);
+      m_expression.coeffRef(row, col) = _other.coeff(row, col);
+      _other.coeffRef(row, col) = tmp;
     }
 
     template<typename OtherDerived>
@@ -91,16 +91,16 @@ template<typename ExpressionType> class SwapWrapper
     }
 
     template<typename OtherDerived, int StoreMode, int LoadMode>
-    void copyPacket(Index rowId, Index colId, const DenseBase<OtherDerived>& other)
+    void copyPacket(Index row, Index col, const DenseBase<OtherDerived>& other)
     {
       OtherDerived& _other = other.const_cast_derived();
-      eigen_internal_assert(rowId >= 0 && rowId < rows()
-                        && colId >= 0 && colId < cols());
-      Packet tmp = m_expression.template packet<StoreMode>(rowId, colId);
-      m_expression.template writePacket<StoreMode>(rowId, colId,
-        _other.template packet<LoadMode>(rowId, colId)
+      eigen_internal_assert(row >= 0 && row < rows()
+                        && col >= 0 && col < cols());
+      Packet tmp = m_expression.template packet<StoreMode>(row, col);
+      m_expression.template writePacket<StoreMode>(row, col,
+        _other.template packet<LoadMode>(row, col)
       );
-      _other.template writePacket<LoadMode>(rowId, colId, tmp);
+      _other.template writePacket<LoadMode>(row, col, tmp);
     }
 
     template<typename OtherDerived, int StoreMode, int LoadMode>

Eigen/src/Core/Transpose.h

@@ -62,7 +62,7 @@ template<typename MatrixType> class Transpose
     typedef typename TransposeImpl<MatrixType,typename internal::traits<MatrixType>::StorageKind>::Base Base;
     EIGEN_GENERIC_PUBLIC_INTERFACE(Transpose)
 
-    inline Transpose(MatrixType& a_matrix) : m_matrix(a_matrix) {}
+    inline Transpose(MatrixType& matrix) : m_matrix(matrix) {}
 
     EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Transpose)
 
@@ -118,10 +118,10 @@ template<typename MatrixType> class TransposeImpl<MatrixType,Dense>
     inline ScalarWithConstIfNotLvalue* data() { return derived().nestedExpression().data(); }
     inline const Scalar* data() const { return derived().nestedExpression().data(); }
 
-    inline ScalarWithConstIfNotLvalue& coeffRef(Index rowId, Index colId)
+    inline ScalarWithConstIfNotLvalue& coeffRef(Index row, Index col)
     {
       EIGEN_STATIC_ASSERT_LVALUE(MatrixType)
-      return derived().nestedExpression().const_cast_derived().coeffRef(colId, rowId);
+      return derived().nestedExpression().const_cast_derived().coeffRef(col, row);
     }
 
     inline ScalarWithConstIfNotLvalue& coeffRef(Index index)
@@ -130,9 +130,9 @@ template<typename MatrixType> class TransposeImpl<MatrixType,Dense>
       return derived().nestedExpression().const_cast_derived().coeffRef(index);
     }
 
-    inline const Scalar& coeffRef(Index rowId, Index colId) const
+    inline const Scalar& coeffRef(Index row, Index col) const
     {
-      return derived().nestedExpression().coeffRef(colId, rowId);
+      return derived().nestedExpression().coeffRef(col, row);
     }
 
     inline const Scalar& coeffRef(Index index) const
@@ -140,9 +140,9 @@ template<typename MatrixType> class TransposeImpl<MatrixType,Dense>
       return derived().nestedExpression().coeffRef(index);
     }
 
-    inline CoeffReturnType coeff(Index rowId, Index colId) const
+    inline CoeffReturnType coeff(Index row, Index col) const
     {
-      return derived().nestedExpression().coeff(colId, rowId);
+      return derived().nestedExpression().coeff(col, row);
     }
 
     inline CoeffReturnType coeff(Index index) const
@@ -151,15 +151,15 @@ template<typename MatrixType> class TransposeImpl<MatrixType,Dense>
     }
 
     template<int LoadMode>
-    inline const PacketScalar packet(Index rowId, Index colId) const
+    inline const PacketScalar packet(Index row, Index col) const
     {
-      return derived().nestedExpression().template packet<LoadMode>(colId, rowId);
+      return derived().nestedExpression().template packet<LoadMode>(col, row);
     }
 
     template<int LoadMode>
-    inline void writePacket(Index rowId, Index colId, const PacketScalar& x)
+    inline void writePacket(Index row, Index col, const PacketScalar& x)
     {
-      derived().nestedExpression().const_cast_derived().template writePacket<LoadMode>(colId, rowId, x);
+      derived().nestedExpression().const_cast_derived().template writePacket<LoadMode>(col, row, x);
     }
 
     template<int LoadMode>
@@ -207,7 +207,7 @@ DenseBase<Derived>::transpose()
   *
   * \sa transposeInPlace(), adjoint() */
 template<typename Derived>
-inline typename DenseBase<Derived>::ConstTransposeReturnType
+inline const typename DenseBase<Derived>::ConstTransposeReturnType
 DenseBase<Derived>::transpose() const
 {
   return ConstTransposeReturnType(derived());
@@ -279,20 +279,17 @@ struct inplace_transpose_selector<MatrixType,false> { // non square matrix
   * m = m.transpose().eval();
   * \endcode
   * and is faster and also safer because in the latter line of code, forgetting the eval() results
-  * in a bug caused by \ref TopicAliasing "aliasing".
+  * in a bug caused by aliasing.
   *
   * Notice however that this method is only useful if you want to replace a matrix by its own transpose.
   * If you just need the transpose of a matrix, use transpose().
   *
-  * \note if the matrix is not square, then \c *this must be a resizable matrix. 
-  * This excludes (non-square) fixed-size matrices, block-expressions and maps.
+  * \note if the matrix is not square, then \c *this must be a resizable matrix.
   *
   * \sa transpose(), adjoint(), adjointInPlace() */
 template<typename Derived>
 inline void DenseBase<Derived>::transposeInPlace()
 {
-  eigen_assert((rows() == cols() || (RowsAtCompileTime == Dynamic && ColsAtCompileTime == Dynamic))
-               && "transposeInPlace() called on a non-square non-resizable matrix");
   internal::inplace_transpose_selector<Derived>::run(derived());
 }
 
@@ -316,7 +313,6 @@ inline void DenseBase<Derived>::transposeInPlace()
   * If you just need the adjoint of a matrix, use adjoint().
   *
   * \note if the matrix is not square, then \c *this must be a resizable matrix.
-  * This excludes (non-square) fixed-size matrices, block-expressions and maps.
   *
   * \sa transpose(), adjoint(), transposeInPlace() */
 template<typename Derived>
@@ -331,11 +327,11 @@ inline void MatrixBase<Derived>::adjointInPlace()
 
 namespace internal {
 
-template<typename BinOp,typename Xpr,typename Rhs>
-struct blas_traits<SelfCwiseBinaryOp<BinOp,Xpr,Rhs> >
- : blas_traits<typename internal::remove_all<typename Xpr::Nested>::type>
+template<typename BinOp,typename NestedXpr,typename Rhs>
+struct blas_traits<SelfCwiseBinaryOp<BinOp,NestedXpr,Rhs> >
+ : blas_traits<NestedXpr>
 {
-  typedef SelfCwiseBinaryOp<BinOp,Xpr,Rhs> XprType;
+  typedef SelfCwiseBinaryOp<BinOp,NestedXpr,Rhs> XprType;
   static inline const XprType extract(const XprType& x) { return x; }
 };
 
@@ -390,8 +386,9 @@ struct checkTransposeAliasing_impl
         eigen_assert((!check_transpose_aliasing_run_time_selector
                       <typename Derived::Scalar,blas_traits<Derived>::IsTransposed,OtherDerived>
                       ::run(extract_data(dst), other))
-          && "aliasing detected during transposition, use transposeInPlace() "
+          && "aliasing detected during tranposition, use transposeInPlace() "
              "or evaluate the rhs into a temporary using .eval()");
+
     }
 };
 

Eigen/src/Core/Transpositions.h

@@ -99,9 +99,9 @@ class TranspositionsBase
     IndicesType& indices() { return derived().indices(); }
 
     /** Resizes to given size. */
-    inline void resize(int newSize)
+    inline void resize(int size)
     {
-      indices().resize(newSize);
+      indices().resize(size);
     }
 
     /** Sets \c *this to represents an identity transformation */
@@ -177,7 +177,7 @@ class Transpositions : public TranspositionsBase<Transpositions<SizeAtCompileTim
 
     /** Generic constructor from expression of the transposition indices. */
     template<typename Other>
-    explicit inline Transpositions(const MatrixBase<Other>& a_indices) : m_indices(a_indices)
+    explicit inline Transpositions(const MatrixBase<Other>& indices) : m_indices(indices)
     {}
 
     /** Copies the \a other transpositions into \c *this */
@@ -234,12 +234,12 @@ class Map<Transpositions<SizeAtCompileTime,MaxSizeAtCompileTime,IndexType>,Packe
     typedef typename Traits::IndicesType IndicesType;
     typedef typename IndicesType::Scalar Index;
 
-    inline Map(const Index* indicesPtr)
-      : m_indices(indicesPtr)
+    inline Map(const Index* indices)
+      : m_indices(indices)
     {}
 
-    inline Map(const Index* indicesPtr, Index size)
-      : m_indices(indicesPtr,size)
+    inline Map(const Index* indices, Index size)
+      : m_indices(indices,size)
     {}
 
     /** Copies the \a other transpositions into \c *this */
@@ -291,8 +291,8 @@ class TranspositionsWrapper
     typedef typename Traits::IndicesType IndicesType;
     typedef typename IndicesType::Scalar Index;
 
-    inline TranspositionsWrapper(IndicesType& a_indices)
-      : m_indices(a_indices)
+    inline TranspositionsWrapper(IndicesType& indices)
+      : m_indices(indices)
     {}
 
     /** Copies the \a other transpositions into \c *this */
@@ -376,8 +376,7 @@ struct transposition_matrix_product_retval
       const int size = m_transpositions.size();
       Index j = 0;
 
-      const typename Dest::Scalar *dst_data = internal::extract_data(dst);
-      if(!(is_same<MatrixTypeNestedCleaned,Dest>::value && dst_data!=0 && dst_data == extract_data(m_matrix)))
+      if(!(is_same<MatrixTypeNestedCleaned,Dest>::value && extract_data(dst) == extract_data(m_matrix)))
         dst = m_matrix;
 
       for(int k=(Transposed?size-1:0) ; Transposed?k>=0:k<size ; Transposed?--k:++k)

Eigen/src/Core/TriangularMatrix.h

@@ -278,21 +278,21 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
 
     /** Efficient triangular matrix times vector/matrix product */
     template<typename OtherDerived>
-    TriangularProduct<Mode, true, MatrixType, false, OtherDerived, OtherDerived::ColsAtCompileTime==1>
+    TriangularProduct<Mode,true,MatrixType,false,OtherDerived, OtherDerived::IsVectorAtCompileTime>
     operator*(const MatrixBase<OtherDerived>& rhs) const
     {
       return TriangularProduct
-              <Mode, true, MatrixType, false, OtherDerived, OtherDerived::ColsAtCompileTime==1>
+              <Mode,true,MatrixType,false,OtherDerived,OtherDerived::IsVectorAtCompileTime>
               (m_matrix, rhs.derived());
     }
 
     /** Efficient vector/matrix times triangular matrix product */
     template<typename OtherDerived> friend
-    TriangularProduct<Mode, false, OtherDerived, OtherDerived::RowsAtCompileTime==1, MatrixType, false>
+    TriangularProduct<Mode,false,OtherDerived,OtherDerived::IsVectorAtCompileTime,MatrixType,false>
     operator*(const MatrixBase<OtherDerived>& lhs, const TriangularView& rhs)
     {
       return TriangularProduct
-              <Mode, false, OtherDerived, OtherDerived::RowsAtCompileTime==1, MatrixType, false>
+              <Mode,false,OtherDerived,OtherDerived::IsVectorAtCompileTime,MatrixType,false>
               (lhs.derived(),rhs.m_matrix);
     }
 
@@ -380,19 +380,19 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
     EIGEN_STRONG_INLINE TriangularView& operator=(const ProductBase<ProductDerived, Lhs,Rhs>& other)
     {
       setZero();
-      return assignProduct(other.derived(),1);
+      return assignProduct(other,1);
     }
     
     template<typename ProductDerived, typename Lhs, typename Rhs>
     EIGEN_STRONG_INLINE TriangularView& operator+=(const ProductBase<ProductDerived, Lhs,Rhs>& other)
     {
-      return assignProduct(other.derived(),1);
+      return assignProduct(other,1);
     }
     
     template<typename ProductDerived, typename Lhs, typename Rhs>
     EIGEN_STRONG_INLINE TriangularView& operator-=(const ProductBase<ProductDerived, Lhs,Rhs>& other)
     {
-      return assignProduct(other.derived(),-1);
+      return assignProduct(other,-1);
     }
     
     
@@ -400,34 +400,25 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
     EIGEN_STRONG_INLINE TriangularView& operator=(const ScaledProduct<ProductDerived>& other)
     {
       setZero();
-      return assignProduct(other.derived(),other.alpha());
+      return assignProduct(other,other.alpha());
     }
     
     template<typename ProductDerived>
     EIGEN_STRONG_INLINE TriangularView& operator+=(const ScaledProduct<ProductDerived>& other)
     {
-      return assignProduct(other.derived(),other.alpha());
+      return assignProduct(other,other.alpha());
     }
     
     template<typename ProductDerived>
     EIGEN_STRONG_INLINE TriangularView& operator-=(const ScaledProduct<ProductDerived>& other)
     {
-      return assignProduct(other.derived(),-other.alpha());
+      return assignProduct(other,-other.alpha());
     }
     
   protected:
     
     template<typename ProductDerived, typename Lhs, typename Rhs>
     EIGEN_STRONG_INLINE TriangularView& assignProduct(const ProductBase<ProductDerived, Lhs,Rhs>& prod, const Scalar& alpha);
-    
-    template<int Mode, bool LhsIsTriangular,
-         typename Lhs, bool LhsIsVector,
-         typename Rhs, bool RhsIsVector>
-    EIGEN_STRONG_INLINE TriangularView& assignProduct(const TriangularProduct<Mode, LhsIsTriangular, Lhs, LhsIsVector, Rhs, RhsIsVector>& prod, const Scalar& alpha)
-    {
-      lazyAssign(alpha*prod.eval());
-      return *this;
-    }
 
     MatrixTypeNested m_matrix;
 };
@@ -788,23 +779,22 @@ MatrixBase<Derived>::triangularView() const
   * \sa isLowerTriangular()
   */
 template<typename Derived>
-bool MatrixBase<Derived>::isUpperTriangular(const RealScalar& prec) const
+bool MatrixBase<Derived>::isUpperTriangular(RealScalar prec) const
 {
-  using std::abs;
   RealScalar maxAbsOnUpperPart = static_cast<RealScalar>(-1);
   for(Index j = 0; j < cols(); ++j)
   {
     Index maxi = (std::min)(j, rows()-1);
     for(Index i = 0; i <= maxi; ++i)
     {
-      RealScalar absValue = abs(coeff(i,j));
+      RealScalar absValue = internal::abs(coeff(i,j));
       if(absValue > maxAbsOnUpperPart) maxAbsOnUpperPart = absValue;
     }
   }
   RealScalar threshold = maxAbsOnUpperPart * prec;
   for(Index j = 0; j < cols(); ++j)
     for(Index i = j+1; i < rows(); ++i)
-      if(abs(coeff(i, j)) > threshold) return false;
+      if(internal::abs(coeff(i, j)) > threshold) return false;
   return true;
 }
 
@@ -814,14 +804,13 @@ bool MatrixBase<Derived>::isUpperTriangular(const RealScalar& prec) const
   * \sa isUpperTriangular()
   */
 template<typename Derived>
-bool MatrixBase<Derived>::isLowerTriangular(const RealScalar& prec) const
+bool MatrixBase<Derived>::isLowerTriangular(RealScalar prec) const
 {
-  using std::abs;
   RealScalar maxAbsOnLowerPart = static_cast<RealScalar>(-1);
   for(Index j = 0; j < cols(); ++j)
     for(Index i = j; i < rows(); ++i)
     {
-      RealScalar absValue = abs(coeff(i,j));
+      RealScalar absValue = internal::abs(coeff(i,j));
       if(absValue > maxAbsOnLowerPart) maxAbsOnLowerPart = absValue;
     }
   RealScalar threshold = maxAbsOnLowerPart * prec;
@@ -829,7 +818,7 @@ bool MatrixBase<Derived>::isLowerTriangular(const RealScalar& prec) const
   {
     Index maxi = (std::min)(j, rows()-1);
     for(Index i = 0; i < maxi; ++i)
-      if(abs(coeff(i, j)) > threshold) return false;
+      if(internal::abs(coeff(i, j)) > threshold) return false;
   }
   return true;
 }

Eigen/src/Core/VectorBlock.h

@@ -90,6 +90,195 @@ template<typename VectorType, int Size> class VectorBlock
 };
 
 
+/** \returns a dynamic-size expression of a segment (i.e. a vector block) in *this.
+  *
+  * \only_for_vectors
+  *
+  * \param start the first coefficient in the segment
+  * \param size the number of coefficients in the segment
+  *
+  * Example: \include MatrixBase_segment_int_int.cpp
+  * Output: \verbinclude MatrixBase_segment_int_int.out
+  *
+  * \note Even though the returned expression has dynamic size, in the case
+  * when it is applied to a fixed-size vector, it inherits a fixed maximal size,
+  * which means that evaluating it does not cause a dynamic memory allocation.
+  *
+  * \sa class Block, segment(Index)
+  */
+template<typename Derived>
+inline typename DenseBase<Derived>::SegmentReturnType
+DenseBase<Derived>::segment(Index start, Index size)
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return SegmentReturnType(derived(), start, size);
+}
+
+/** This is the const version of segment(Index,Index).*/
+template<typename Derived>
+inline typename DenseBase<Derived>::ConstSegmentReturnType
+DenseBase<Derived>::segment(Index start, Index size) const
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return ConstSegmentReturnType(derived(), start, size);
+}
+
+/** \returns a dynamic-size expression of the first coefficients of *this.
+  *
+  * \only_for_vectors
+  *
+  * \param size the number of coefficients in the block
+  *
+  * Example: \include MatrixBase_start_int.cpp
+  * Output: \verbinclude MatrixBase_start_int.out
+  *
+  * \note Even though the returned expression has dynamic size, in the case
+  * when it is applied to a fixed-size vector, it inherits a fixed maximal size,
+  * which means that evaluating it does not cause a dynamic memory allocation.
+  *
+  * \sa class Block, block(Index,Index)
+  */
+template<typename Derived>
+inline typename DenseBase<Derived>::SegmentReturnType
+DenseBase<Derived>::head(Index size)
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return SegmentReturnType(derived(), 0, size);
+}
+
+/** This is the const version of head(Index).*/
+template<typename Derived>
+inline typename DenseBase<Derived>::ConstSegmentReturnType
+DenseBase<Derived>::head(Index size) const
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return ConstSegmentReturnType(derived(), 0, size);
+}
+
+/** \returns a dynamic-size expression of the last coefficients of *this.
+  *
+  * \only_for_vectors
+  *
+  * \param size the number of coefficients in the block
+  *
+  * Example: \include MatrixBase_end_int.cpp
+  * Output: \verbinclude MatrixBase_end_int.out
+  *
+  * \note Even though the returned expression has dynamic size, in the case
+  * when it is applied to a fixed-size vector, it inherits a fixed maximal size,
+  * which means that evaluating it does not cause a dynamic memory allocation.
+  *
+  * \sa class Block, block(Index,Index)
+  */
+template<typename Derived>
+inline typename DenseBase<Derived>::SegmentReturnType
+DenseBase<Derived>::tail(Index size)
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return SegmentReturnType(derived(), this->size() - size, size);
+}
+
+/** This is the const version of tail(Index).*/
+template<typename Derived>
+inline typename DenseBase<Derived>::ConstSegmentReturnType
+DenseBase<Derived>::tail(Index size) const
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return ConstSegmentReturnType(derived(), this->size() - size, size);
+}
+
+/** \returns a fixed-size expression of a segment (i.e. a vector block) in \c *this
+  *
+  * \only_for_vectors
+  *
+  * The template parameter \a Size is the number of coefficients in the block
+  *
+  * \param start the index of the first element of the sub-vector
+  *
+  * Example: \include MatrixBase_template_int_segment.cpp
+  * Output: \verbinclude MatrixBase_template_int_segment.out
+  *
+  * \sa class Block
+  */
+template<typename Derived>
+template<int Size>
+inline typename DenseBase<Derived>::template FixedSegmentReturnType<Size>::Type
+DenseBase<Derived>::segment(Index start)
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return typename FixedSegmentReturnType<Size>::Type(derived(), start);
+}
+
+/** This is the const version of segment<int>(Index).*/
+template<typename Derived>
+template<int Size>
+inline typename DenseBase<Derived>::template ConstFixedSegmentReturnType<Size>::Type
+DenseBase<Derived>::segment(Index start) const
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return typename ConstFixedSegmentReturnType<Size>::Type(derived(), start);
+}
+
+/** \returns a fixed-size expression of the first coefficients of *this.
+  *
+  * \only_for_vectors
+  *
+  * The template parameter \a Size is the number of coefficients in the block
+  *
+  * Example: \include MatrixBase_template_int_start.cpp
+  * Output: \verbinclude MatrixBase_template_int_start.out
+  *
+  * \sa class Block
+  */
+template<typename Derived>
+template<int Size>
+inline typename DenseBase<Derived>::template FixedSegmentReturnType<Size>::Type
+DenseBase<Derived>::head()
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return typename FixedSegmentReturnType<Size>::Type(derived(), 0);
+}
+
+/** This is the const version of head<int>().*/
+template<typename Derived>
+template<int Size>
+inline typename DenseBase<Derived>::template ConstFixedSegmentReturnType<Size>::Type
+DenseBase<Derived>::head() const
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return typename ConstFixedSegmentReturnType<Size>::Type(derived(), 0);
+}
+
+/** \returns a fixed-size expression of the last coefficients of *this.
+  *
+  * \only_for_vectors
+  *
+  * The template parameter \a Size is the number of coefficients in the block
+  *
+  * Example: \include MatrixBase_template_int_end.cpp
+  * Output: \verbinclude MatrixBase_template_int_end.out
+  *
+  * \sa class Block
+  */
+template<typename Derived>
+template<int Size>
+inline typename DenseBase<Derived>::template FixedSegmentReturnType<Size>::Type
+DenseBase<Derived>::tail()
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return typename FixedSegmentReturnType<Size>::Type(derived(), size() - Size);
+}
+
+/** This is the const version of tail<int>.*/
+template<typename Derived>
+template<int Size>
+inline typename DenseBase<Derived>::template ConstFixedSegmentReturnType<Size>::Type
+DenseBase<Derived>::tail() const
+{
+  EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
+  return typename ConstFixedSegmentReturnType<Size>::Type(derived(), size() - Size);
+}
+
 } // end namespace Eigen
 
 #endif // EIGEN_VECTORBLOCK_H

Eigen/src/Core/VectorwiseOp.h

@@ -50,7 +50,7 @@ struct traits<PartialReduxExpr<MatrixType, MemberOp, Direction> >
     MaxColsAtCompileTime = Direction==Horizontal ? 1 : MatrixType::MaxColsAtCompileTime,
     Flags0 = (unsigned int)_MatrixTypeNested::Flags & HereditaryBits,
     Flags = (Flags0 & ~RowMajorBit) | (RowsAtCompileTime == 1 ? RowMajorBit : 0),
-    TraversalSize = Direction==Vertical ? MatrixType::RowsAtCompileTime :  MatrixType::ColsAtCompileTime
+    TraversalSize = Direction==Vertical ? RowsAtCompileTime : ColsAtCompileTime
   };
   #if EIGEN_GNUC_AT_LEAST(3,4)
   typedef typename MemberOp::template Cost<InputScalar,int(TraversalSize)> CostOpType;
@@ -58,8 +58,7 @@ struct traits<PartialReduxExpr<MatrixType, MemberOp, Direction> >
   typedef typename MemberOp::template Cost<InputScalar,TraversalSize> CostOpType;
   #endif
   enum {
-    CoeffReadCost = TraversalSize==Dynamic ? Dynamic
-                  : TraversalSize * traits<_MatrixTypeNested>::CoeffReadCost + int(CostOpType::value)
+    CoeffReadCost = TraversalSize * traits<_MatrixTypeNested>::CoeffReadCost + int(CostOpType::value)
   };
 };
 }
@@ -104,8 +103,8 @@ class PartialReduxExpr : internal::no_assignment_operator,
 
 #define EIGEN_MEMBER_FUNCTOR(MEMBER,COST)                               \
   template <typename ResultType>                                        \
-  struct member_##MEMBER {                                              \
-    EIGEN_EMPTY_STRUCT_CTOR(member_##MEMBER)                            \
+  struct member_##MEMBER {                                           \
+    EIGEN_EMPTY_STRUCT_CTOR(member_##MEMBER)                         \
     typedef ResultType result_type;                                     \
     template<typename Scalar, int Size> struct Cost                     \
     { enum { value = COST }; };                                         \
@@ -234,28 +233,6 @@ template<typename ExpressionType, int Direction> class VectorwiseOp
                        Direction==Vertical   ? 1 : m_matrix.rows(),
                        Direction==Horizontal ? 1 : m_matrix.cols());
     }
-    
-    template<typename OtherDerived> struct OppositeExtendedType {
-      typedef Replicate<OtherDerived,
-                        Direction==Horizontal ? 1 : ExpressionType::RowsAtCompileTime,
-                        Direction==Vertical   ? 1 : ExpressionType::ColsAtCompileTime> Type;
-    };
-
-    /** \internal
-      * Replicates a vector in the opposite direction to match the size of \c *this */
-    template<typename OtherDerived>
-    typename OppositeExtendedType<OtherDerived>::Type
-    extendedToOpposite(const DenseBase<OtherDerived>& other) const
-    {
-      EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(Direction==Horizontal, OtherDerived::MaxColsAtCompileTime==1),
-                          YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED)
-      EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(Direction==Vertical, OtherDerived::MaxRowsAtCompileTime==1),
-                          YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED)
-      return typename OppositeExtendedType<OtherDerived>::Type
-                      (other.derived(),
-                       Direction==Horizontal  ? 1 : m_matrix.rows(),
-                       Direction==Vertical    ? 1 : m_matrix.cols());
-    }
 
   public:
 
@@ -278,8 +255,6 @@ template<typename ExpressionType, int Direction> class VectorwiseOp
 
     /** \returns a row (or column) vector expression of the smallest coefficient
       * of each column (or row) of the referenced expression.
-      * 
-      * \warning the result is undefined if \c *this contains NaN.
       *
       * Example: \include PartialRedux_minCoeff.cpp
       * Output: \verbinclude PartialRedux_minCoeff.out
@@ -290,8 +265,6 @@ template<typename ExpressionType, int Direction> class VectorwiseOp
 
     /** \returns a row (or column) vector expression of the largest coefficient
       * of each column (or row) of the referenced expression.
-      * 
-      * \warning the result is undefined if \c *this contains NaN.
       *
       * Example: \include PartialRedux_maxCoeff.cpp
       * Output: \verbinclude PartialRedux_maxCoeff.out
@@ -531,23 +504,6 @@ template<typename ExpressionType, int Direction> class VectorwiseOp
       EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived)
       return m_matrix / extendedTo(other.derived());
     }
-    
-    /** \returns an expression where each column of row of the referenced matrix are normalized.
-      * The referenced matrix is \b not modified.
-      * \sa MatrixBase::normalized(), normalize()
-      */
-    CwiseBinaryOp<internal::scalar_quotient_op<Scalar>,
-                  const ExpressionTypeNestedCleaned,
-                  const typename OppositeExtendedType<typename ReturnType<internal::member_norm,RealScalar>::Type>::Type>
-    normalized() const { return m_matrix.cwiseQuotient(extendedToOpposite(this->norm())); }
-    
-    
-    /** Normalize in-place each row or columns of the referenced matrix.
-      * \sa MatrixBase::normalize(), normalized()
-      */
-    void normalize() {
-      m_matrix = this->normalized();
-    }
 
 /////////// Geometry module ///////////
 

Eigen/src/Core/Visitor.h

@@ -76,17 +76,14 @@ template<typename Derived>
 template<typename Visitor>
 void DenseBase<Derived>::visit(Visitor& visitor) const
 {
-  typedef typename internal::remove_all<typename Derived::Nested>::type ThisNested;
-  typename Derived::Nested thisNested(derived());
-
   enum { unroll = SizeAtCompileTime != Dynamic
                    && CoeffReadCost != Dynamic
                    && (SizeAtCompileTime == 1 || internal::functor_traits<Visitor>::Cost != Dynamic)
                    && SizeAtCompileTime * CoeffReadCost + (SizeAtCompileTime-1) * internal::functor_traits<Visitor>::Cost
                       <= EIGEN_UNROLLING_LIMIT };
-  return internal::visitor_impl<Visitor, ThisNested,
+  return internal::visitor_impl<Visitor, Derived,
       unroll ? int(SizeAtCompileTime) : Dynamic
-    >::run(thisNested, visitor);
+    >::run(derived(), visitor);
 }
 
 namespace internal {
@@ -167,25 +164,25 @@ struct functor_traits<max_coeff_visitor<Scalar> > {
 
 } // end namespace internal
 
-/** \returns the minimum of all coefficients of *this and puts in *row and *col its location.
-  * \warning the result is undefined if \c *this contains NaN.
+/** \returns the minimum of all coefficients of *this
+  * and puts in *row and *col its location.
   *
   * \sa DenseBase::minCoeff(Index*), DenseBase::maxCoeff(Index*,Index*), DenseBase::visitor(), DenseBase::minCoeff()
   */
 template<typename Derived>
 template<typename IndexType>
 typename internal::traits<Derived>::Scalar
-DenseBase<Derived>::minCoeff(IndexType* rowId, IndexType* colId) const
+DenseBase<Derived>::minCoeff(IndexType* row, IndexType* col) const
 {
   internal::min_coeff_visitor<Derived> minVisitor;
   this->visit(minVisitor);
-  *rowId = minVisitor.row;
-  if (colId) *colId = minVisitor.col;
+  *row = minVisitor.row;
+  if (col) *col = minVisitor.col;
   return minVisitor.res;
 }
 
-/** \returns the minimum of all coefficients of *this and puts in *index its location.
-  * \warning the result is undefined if \c *this contains NaN. 
+/** \returns the minimum of all coefficients of *this
+  * and puts in *index its location.
   *
   * \sa DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::maxCoeff(IndexType*,IndexType*), DenseBase::visitor(), DenseBase::minCoeff()
   */
@@ -201,25 +198,25 @@ DenseBase<Derived>::minCoeff(IndexType* index) const
   return minVisitor.res;
 }
 
-/** \returns the maximum of all coefficients of *this and puts in *row and *col its location.
-  * \warning the result is undefined if \c *this contains NaN. 
+/** \returns the maximum of all coefficients of *this
+  * and puts in *row and *col its location.
   *
   * \sa DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::visitor(), DenseBase::maxCoeff()
   */
 template<typename Derived>
 template<typename IndexType>
 typename internal::traits<Derived>::Scalar
-DenseBase<Derived>::maxCoeff(IndexType* rowPtr, IndexType* colPtr) const
+DenseBase<Derived>::maxCoeff(IndexType* row, IndexType* col) const
 {
   internal::max_coeff_visitor<Derived> maxVisitor;
   this->visit(maxVisitor);
-  *rowPtr = maxVisitor.row;
-  if (colPtr) *colPtr = maxVisitor.col;
+  *row = maxVisitor.row;
+  if (col) *col = maxVisitor.col;
   return maxVisitor.res;
 }
 
-/** \returns the maximum of all coefficients of *this and puts in *index its location.
-  * \warning the result is undefined if \c *this contains NaN.
+/** \returns the maximum of all coefficients of *this
+  * and puts in *index its location.
   *
   * \sa DenseBase::maxCoeff(IndexType*,IndexType*), DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::visitor(), DenseBase::maxCoeff()
   */

Eigen/src/Core/arch/AltiVec/PacketMath.h

@@ -173,9 +173,6 @@ template<> EIGEN_STRONG_INLINE Packet4i psub<Packet4i>(const Packet4i& a, const
 template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) { return psub<Packet4f>(p4f_ZERO, a); }
 template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) { return psub<Packet4i>(p4i_ZERO, a); }
 
-template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; }
-template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; }
-
 template<> EIGEN_STRONG_INLINE Packet4f pmul<Packet4f>(const Packet4f& a, const Packet4f& b) { return vec_madd(a,b,p4f_ZERO); }
 /* Commented out: it's actually slower than processing it scalar
  *

Eigen/src/Core/arch/NEON/Complex.h

@@ -68,6 +68,7 @@ template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a)
 template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
 {
   Packet4f v1, v2;
+  float32x2_t a_lo, a_hi;
 
   // Get the real values of a | a1_re | a1_re | a2_re | a2_re |
   v1 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 0), vdup_lane_f32(vget_high_f32(a.v), 0));
@@ -80,7 +81,9 @@ template<> EIGEN_STRONG_INLINE Packet2cf pmul<Packet2cf>(const Packet2cf& a, con
   // Conjugate v2 
   v2 = vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(v2), p4ui_CONJ_XOR));
   // Swap real/imag elements in v2.
-  v2 = vrev64q_f32(v2);
+  a_lo = vrev64_f32(vget_low_f32(v2));
+  a_hi = vrev64_f32(vget_high_f32(v2));
+  v2 = vcombine_f32(a_lo, a_hi);
   // Add and return the result
   return Packet2cf(vaddq_f32(v1, v2));
 }
@@ -110,7 +113,7 @@ template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<
 template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); }
 template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); }
 
-template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> *   addr) { EIGEN_ARM_PREFETCH((float *)addr); }
+template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> *   addr) { __pld((float *)addr); }
 
 template<> EIGEN_STRONG_INLINE std::complex<float>  pfirst<Packet2cf>(const Packet2cf& a)
 {
@@ -238,10 +241,13 @@ template<> EIGEN_STRONG_INLINE Packet2cf pdiv<Packet2cf>(const Packet2cf& a, con
   // TODO optimize it for AltiVec
   Packet2cf res = conj_helper<Packet2cf,Packet2cf,false,true>().pmul(a,b);
   Packet4f s, rev_s;
+  float32x2_t a_lo, a_hi;
 
   // this computes the norm
   s = vmulq_f32(b.v, b.v);
-  rev_s = vrev64q_f32(s);
+  a_lo = vrev64_f32(vget_low_f32(s));
+  a_hi = vrev64_f32(vget_high_f32(s));
+  rev_s = vcombine_f32(a_lo, a_hi);
 
   return Packet2cf(pdiv(res.v, vaddq_f32(s,rev_s)));
 }

Eigen/src/Core/arch/NEON/PacketMath.h

@@ -48,18 +48,9 @@ typedef uint32x4_t  Packet4ui;
   #define EIGEN_INIT_NEON_PACKET2(X, Y)       {X, Y}
   #define EIGEN_INIT_NEON_PACKET4(X, Y, Z, W) {X, Y, Z, W}
 #endif
-
-// arm64 does have the pld instruction. If available, let's trust the __builtin_prefetch built-in function
-// which available on LLVM and GCC (at least)
-#if EIGEN_HAS_BUILTIN(__builtin_prefetch) || defined(__GNUC__)
-  #define EIGEN_ARM_PREFETCH(ADDR) __builtin_prefetch(ADDR);
-#elif defined __pld
-  #define EIGEN_ARM_PREFETCH(ADDR) __pld(ADDR)
-#elif !defined(__aarch64__)
-  #define EIGEN_ARM_PREFETCH(ADDR) __asm__ __volatile__ ( "   pld [%[addr]]\n" :: [addr] "r" (ADDR) : "cc" );
-#else
-  // by default no explicit prefetching
-  #define EIGEN_ARM_PREFETCH(ADDR)
+    
+#ifndef __pld
+#define __pld(x) asm volatile ( "   pld [%[addr]]\n" :: [addr] "r" (x) : "cc" );
 #endif
 
 template<> struct packet_traits<float>  : default_packet_traits
@@ -124,9 +115,6 @@ template<> EIGEN_STRONG_INLINE Packet4i psub<Packet4i>(const Packet4i& a, const
 template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) { return vnegq_f32(a); }
 template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) { return vnegq_s32(a); }
 
-template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; }
-template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; }
-
 template<> EIGEN_STRONG_INLINE Packet4f pmul<Packet4f>(const Packet4f& a, const Packet4f& b) { return vmulq_f32(a,b); }
 template<> EIGEN_STRONG_INLINE Packet4i pmul<Packet4i>(const Packet4i& a, const Packet4i& b) { return vmulq_s32(a,b); }
 
@@ -200,15 +188,15 @@ template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int* from)   { EI
 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);
+  lo = vdup_n_f32(*from);
+  hi = vdup_n_f32(*(from+1));
   return vcombine_f32(lo, hi);
 }
 template<> EIGEN_STRONG_INLINE Packet4i ploaddup<Packet4i>(const int*     from)
 {
   int32x2_t lo, hi;
-  lo = vld1_dup_s32(from);
-  hi = vld1_dup_s32(from+1);
+  lo = vdup_n_s32(*from);
+  hi = vdup_n_s32(*(from+1));
   return vcombine_s32(lo, hi);
 }
 
@@ -218,8 +206,8 @@ template<> EIGEN_STRONG_INLINE void pstore<int>(int*       to, const Packet4i& f
 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 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) { __pld(addr); }
+template<> EIGEN_STRONG_INLINE void prefetch<int>(const int*     addr) { __pld(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]; }
@@ -249,12 +237,15 @@ template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) { return vabsq_s
 template<> EIGEN_STRONG_INLINE float predux<Packet4f>(const Packet4f& a)
 {
   float32x2_t a_lo, a_hi, sum;
+  float s[2];
 
   a_lo = vget_low_f32(a);
   a_hi = vget_high_f32(a);
   sum = vpadd_f32(a_lo, a_hi);
   sum = vpadd_f32(sum, sum);
-  return vget_lane_f32(sum, 0);
+  vst1_f32(s, sum);
+
+  return s[0];
 }
 
 template<> EIGEN_STRONG_INLINE Packet4f preduxp<Packet4f>(const Packet4f* vecs)
@@ -280,12 +271,15 @@ template<> EIGEN_STRONG_INLINE Packet4f preduxp<Packet4f>(const Packet4f* vecs)
 template<> EIGEN_STRONG_INLINE int predux<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, sum;
+  int32_t s[2];
 
   a_lo = vget_low_s32(a);
   a_hi = vget_high_s32(a);
   sum = vpadd_s32(a_lo, a_hi);
   sum = vpadd_s32(sum, sum);
-  return vget_lane_s32(sum, 0);
+  vst1_s32(s, sum);
+
+  return s[0];
 }
 
 template<> EIGEN_STRONG_INLINE Packet4i preduxp<Packet4i>(const Packet4i* vecs)
@@ -313,6 +307,7 @@ template<> EIGEN_STRONG_INLINE Packet4i preduxp<Packet4i>(const Packet4i* vecs)
 template<> EIGEN_STRONG_INLINE float predux_mul<Packet4f>(const Packet4f& a)
 {
   float32x2_t a_lo, a_hi, prod;
+  float s[2];
 
   // Get a_lo = |a1|a2| and a_hi = |a3|a4|
   a_lo = vget_low_f32(a);
@@ -321,12 +316,14 @@ template<> EIGEN_STRONG_INLINE float predux_mul<Packet4f>(const Packet4f& a)
   prod = vmul_f32(a_lo, a_hi);
   // Multiply prod with its swapped value |a2*a4|a1*a3|
   prod = vmul_f32(prod, vrev64_f32(prod));
+  vst1_f32(s, prod);
 
-  return vget_lane_f32(prod, 0);
+  return s[0];
 }
 template<> EIGEN_STRONG_INLINE int predux_mul<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, prod;
+  int32_t s[2];
 
   // Get a_lo = |a1|a2| and a_hi = |a3|a4|
   a_lo = vget_low_s32(a);
@@ -335,58 +332,65 @@ template<> EIGEN_STRONG_INLINE int predux_mul<Packet4i>(const Packet4i& a)
   prod = vmul_s32(a_lo, a_hi);
   // Multiply prod with its swapped value |a2*a4|a1*a3|
   prod = vmul_s32(prod, vrev64_s32(prod));
+  vst1_s32(s, prod);
 
-  return vget_lane_s32(prod, 0);
+  return s[0];
 }
 
 // min
 template<> EIGEN_STRONG_INLINE float predux_min<Packet4f>(const Packet4f& a)
 {
   float32x2_t a_lo, a_hi, min;
+  float s[2];
 
   a_lo = vget_low_f32(a);
   a_hi = vget_high_f32(a);
   min = vpmin_f32(a_lo, a_hi);
   min = vpmin_f32(min, min);
+  vst1_f32(s, min);
 
-  return vget_lane_f32(min, 0);
+  return s[0];
 }
-
 template<> EIGEN_STRONG_INLINE int predux_min<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, min;
+  int32_t s[2];
 
   a_lo = vget_low_s32(a);
   a_hi = vget_high_s32(a);
   min = vpmin_s32(a_lo, a_hi);
   min = vpmin_s32(min, min);
-  
-  return vget_lane_s32(min, 0);
+  vst1_s32(s, min);
+
+  return s[0];
 }
 
 // max
 template<> EIGEN_STRONG_INLINE float predux_max<Packet4f>(const Packet4f& a)
 {
   float32x2_t a_lo, a_hi, max;
+  float s[2];
 
   a_lo = vget_low_f32(a);
   a_hi = vget_high_f32(a);
   max = vpmax_f32(a_lo, a_hi);
   max = vpmax_f32(max, max);
+  vst1_f32(s, max);
 
-  return vget_lane_f32(max, 0);
+  return s[0];
 }
-
 template<> EIGEN_STRONG_INLINE int predux_max<Packet4i>(const Packet4i& a)
 {
   int32x2_t a_lo, a_hi, max;
+  int32_t s[2];
 
   a_lo = vget_low_s32(a);
   a_hi = vget_high_s32(a);
   max = vpmax_s32(a_lo, a_hi);
   max = vpmax_s32(max, max);
+  vst1_s32(s, max);
 
-  return vget_lane_s32(max, 0);
+  return s[0];
 }
 
 // this PALIGN_NEON business is to work around a bug in LLVM Clang 3.0 causing incorrect compilation errors,

Eigen/src/Core/arch/SSE/Complex.h

@@ -81,31 +81,25 @@ template<> EIGEN_STRONG_INLINE Packet2cf por    <Packet2cf>(const Packet2cf& a,
 template<> EIGEN_STRONG_INLINE Packet2cf pxor   <Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_xor_ps(a.v,b.v)); }
 template<> EIGEN_STRONG_INLINE Packet2cf pandnot<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_andnot_ps(a.v,b.v)); }
 
-template<> EIGEN_STRONG_INLINE Packet2cf pload <Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>(&numext::real_ref(*from))); }
-template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>(&numext::real_ref(*from))); }
+template<> EIGEN_STRONG_INLINE Packet2cf pload <Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload<Packet4f>(&real_ref(*from))); }
+template<> EIGEN_STRONG_INLINE Packet2cf ploadu<Packet2cf>(const std::complex<float>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu<Packet4f>(&real_ref(*from))); }
 
 template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>&  from)
 {
   Packet2cf res;
-#if EIGEN_GNUC_AT_MOST(4,2)
-  // Workaround annoying "may be used uninitialized in this function" warning with gcc 4.2
+  #if EIGEN_GNUC_AT_MOST(4,2)
+  // workaround annoying "may be used uninitialized in this function" warning with gcc 4.2
   res.v = _mm_loadl_pi(_mm_set1_ps(0.0f), reinterpret_cast<const __m64*>(&from));
-#elif EIGEN_GNUC_AT_LEAST(4,6)
-  // Suppress annoying "may be used uninitialized in this function" warning with gcc >= 4.6
-  #pragma GCC diagnostic push
-  #pragma GCC diagnostic ignored "-Wuninitialized"
+  #else
   res.v = _mm_loadl_pi(res.v, (const __m64*)&from);
-  #pragma GCC diagnostic pop
-#else
-  res.v = _mm_loadl_pi(res.v, (const __m64*)&from);
-#endif
+  #endif
   return Packet2cf(_mm_movelh_ps(res.v,res.v));
 }
 
 template<> EIGEN_STRONG_INLINE Packet2cf ploaddup<Packet2cf>(const std::complex<float>* from) { return pset1<Packet2cf>(*from); }
 
-template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&numext::real_ref(*to), from.v); }
-template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&numext::real_ref(*to), from.v); }
+template<> EIGEN_STRONG_INLINE void pstore <std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&real_ref(*to), from.v); }
+template<> EIGEN_STRONG_INLINE void pstoreu<std::complex<float> >(std::complex<float> *   to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&real_ref(*to), from.v); }
 
 template<> EIGEN_STRONG_INLINE void prefetch<std::complex<float> >(const std::complex<float> *   addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); }
 

Eigen/src/Core/arch/SSE/MathFunctions.h

@@ -52,7 +52,7 @@ Packet4f plog<Packet4f>(const Packet4f& _x)
 
   Packet4i emm0;
 
-  Packet4f invalid_mask = _mm_cmpnge_ps(x, _mm_setzero_ps()); // not greater equal is true if x is NaN
+  Packet4f invalid_mask = _mm_cmplt_ps(x, _mm_setzero_ps());
   Packet4f iszero_mask = _mm_cmpeq_ps(x, _mm_setzero_ps());
 
   x = pmax(x, p4f_min_norm_pos);  /* cut off denormalized stuff */
@@ -126,7 +126,7 @@ Packet4f pexp<Packet4f>(const Packet4f& _x)
   _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p4, 1.6666665459E-1f);
   _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p5, 5.0000001201E-1f);
 
-  Packet4f tmp, fx;
+  Packet4f tmp = _mm_setzero_ps(), fx;
   Packet4i emm0;
 
   // clamp x
@@ -135,16 +135,13 @@ Packet4f pexp<Packet4f>(const Packet4f& _x)
   /* express exp(x) as exp(g + n*log(2)) */
   fx = pmadd(x, p4f_cephes_LOG2EF, p4f_half);
 
-#ifdef EIGEN_VECTORIZE_SSE4_1
-  fx = _mm_floor_ps(fx);
-#else
+  /* how to perform a floorf with SSE: just below */
   emm0 = _mm_cvttps_epi32(fx);
   tmp  = _mm_cvtepi32_ps(emm0);
   /* if greater, substract 1 */
   Packet4f mask = _mm_cmpgt_ps(tmp, fx);
   mask = _mm_and_ps(mask, p4f_1);
   fx = psub(tmp, mask);
-#endif
 
   tmp = pmul(fx, p4f_cephes_exp_C1);
   Packet4f z = pmul(fx, p4f_cephes_exp_C2);
@@ -166,80 +163,7 @@ Packet4f pexp<Packet4f>(const Packet4f& _x)
   emm0 = _mm_cvttps_epi32(fx);
   emm0 = _mm_add_epi32(emm0, p4i_0x7f);
   emm0 = _mm_slli_epi32(emm0, 23);
-  return pmax(pmul(y, Packet4f(_mm_castsi128_ps(emm0))), _x);
-}
-template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED
-Packet2d pexp<Packet2d>(const Packet2d& _x)
-{
-  Packet2d x = _x;
-
-  _EIGEN_DECLARE_CONST_Packet2d(1 , 1.0);
-  _EIGEN_DECLARE_CONST_Packet2d(2 , 2.0);
-  _EIGEN_DECLARE_CONST_Packet2d(half, 0.5);
-
-  _EIGEN_DECLARE_CONST_Packet2d(exp_hi,  709.437);
-  _EIGEN_DECLARE_CONST_Packet2d(exp_lo, -709.436139303);
-
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_LOG2EF, 1.4426950408889634073599);
-
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p0, 1.26177193074810590878e-4);
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p1, 3.02994407707441961300e-2);
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p2, 9.99999999999999999910e-1);
-
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q0, 3.00198505138664455042e-6);
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q1, 2.52448340349684104192e-3);
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q2, 2.27265548208155028766e-1);
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q3, 2.00000000000000000009e0);
-
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C1, 0.693145751953125);
-  _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C2, 1.42860682030941723212e-6);
-  static const __m128i p4i_1023_0 = _mm_setr_epi32(1023, 1023, 0, 0);
-
-  Packet2d tmp, fx;
-  Packet4i emm0;
-
-  // clamp x
-  x = pmax(pmin(x, p2d_exp_hi), p2d_exp_lo);
-  /* express exp(x) as exp(g + n*log(2)) */
-  fx = pmadd(p2d_cephes_LOG2EF, x, p2d_half);
-
-#ifdef EIGEN_VECTORIZE_SSE4_1
-  fx = _mm_floor_pd(fx);
-#else
-  emm0 = _mm_cvttpd_epi32(fx);
-  tmp  = _mm_cvtepi32_pd(emm0);
-  /* if greater, substract 1 */
-  Packet2d mask = _mm_cmpgt_pd(tmp, fx);
-  mask = _mm_and_pd(mask, p2d_1);
-  fx = psub(tmp, mask);
-#endif
-
-  tmp = pmul(fx, p2d_cephes_exp_C1);
-  Packet2d z = pmul(fx, p2d_cephes_exp_C2);
-  x = psub(x, tmp);
-  x = psub(x, z);
-
-  Packet2d x2 = pmul(x,x);
-
-  Packet2d px = p2d_cephes_exp_p0;
-  px = pmadd(px, x2, p2d_cephes_exp_p1);
-  px = pmadd(px, x2, p2d_cephes_exp_p2);
-  px = pmul (px, x);
-
-  Packet2d qx = p2d_cephes_exp_q0;
-  qx = pmadd(qx, x2, p2d_cephes_exp_q1);
-  qx = pmadd(qx, x2, p2d_cephes_exp_q2);
-  qx = pmadd(qx, x2, p2d_cephes_exp_q3);
-
-  x = pdiv(px,psub(qx,px));
-  x = pmadd(p2d_2,x,p2d_1);
-
-  // build 2^n
-  emm0 = _mm_cvttpd_epi32(fx);
-  emm0 = _mm_add_epi32(emm0, p4i_1023_0);
-  emm0 = _mm_slli_epi32(emm0, 20);
-  emm0 = _mm_shuffle_epi32(emm0, _MM_SHUFFLE(1,2,0,3));
-  return pmax(pmul(x, Packet2d(_mm_castsi128_pd(emm0))), _x);
+  return pmul(y, _mm_castsi128_ps(emm0));
 }
 
 /* evaluation of 4 sines at onces, using SSE2 intrinsics.
@@ -279,7 +203,7 @@ Packet4f psin<Packet4f>(const Packet4f& _x)
   _EIGEN_DECLARE_CONST_Packet4f(coscof_p2,  4.166664568298827E-002f);
   _EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4 / M_PI
 
-  Packet4f xmm1, xmm2, xmm3, sign_bit, y;
+  Packet4f xmm1, xmm2 = _mm_setzero_ps(), xmm3, sign_bit, y;
 
   Packet4i emm0, emm2;
   sign_bit = x;
@@ -378,7 +302,7 @@ Packet4f pcos<Packet4f>(const Packet4f& _x)
   _EIGEN_DECLARE_CONST_Packet4f(coscof_p2,  4.166664568298827E-002f);
   _EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4 / M_PI
 
-  Packet4f xmm1, xmm2, xmm3, y;
+  Packet4f xmm1, xmm2 = _mm_setzero_ps(), xmm3, y;
   Packet4i emm0, emm2;
 
   x = pabs(x);
@@ -442,32 +366,21 @@ Packet4f pcos<Packet4f>(const Packet4f& _x)
   return _mm_xor_ps(y, sign_bit);
 }
 
-#if EIGEN_FAST_MATH
-
 // This is based on Quake3's fast inverse square root.
 // For detail see here: http://www.beyond3d.com/content/articles/8/
-// It lacks 1 (or 2 bits in some rare cases) of precision, and does not handle negative, +inf, or denormalized numbers correctly.
 template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED
 Packet4f psqrt<Packet4f>(const Packet4f& _x)
 {
   Packet4f half = pmul(_x, pset1<Packet4f>(.5f));
 
   /* select only the inverse sqrt of non-zero inputs */
-  Packet4f non_zero_mask = _mm_cmpge_ps(_x, pset1<Packet4f>((std::numeric_limits<float>::min)()));
+  Packet4f non_zero_mask = _mm_cmpgt_ps(_x, pset1<Packet4f>((std::numeric_limits<float>::min)()));
   Packet4f x = _mm_and_ps(non_zero_mask, _mm_rsqrt_ps(_x));
 
   x = pmul(x, psub(pset1<Packet4f>(1.5f), pmul(half, pmul(x,x))));
   return pmul(_x,x);
 }
 
-#else
-
-template<> EIGEN_STRONG_INLINE Packet4f psqrt<Packet4f>(const Packet4f& x) { return _mm_sqrt_ps(x); }
-
-#endif
-
-template<> EIGEN_STRONG_INLINE Packet2d psqrt<Packet2d>(const Packet2d& x) { return _mm_sqrt_pd(x); }
-
 } // end namespace internal
 
 } // end namespace Eigen

Eigen/src/Core/arch/SSE/PacketMath.h

@@ -48,9 +48,6 @@ template<> struct is_arithmetic<__m128d> { enum { value = true }; };
 #define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \
   const Packet4f p4f_##NAME = pset1<Packet4f>(X)
 
-#define _EIGEN_DECLARE_CONST_Packet2d(NAME,X) \
-  const Packet2d p2d_##NAME = pset1<Packet2d>(X)
-
 #define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \
   const Packet4f p4f_##NAME = _mm_castsi128_ps(pset1<Packet4i>(X))
 
@@ -66,7 +63,7 @@ template<> struct packet_traits<float>  : default_packet_traits
     AlignedOnScalar = 1,
     size=4,
 
-    HasDiv  = 1,
+    HasDiv    = 1,
     HasSin  = EIGEN_FAST_MATH,
     HasCos  = EIGEN_FAST_MATH,
     HasLog  = 1,
@@ -82,9 +79,7 @@ template<> struct packet_traits<double> : default_packet_traits
     AlignedOnScalar = 1,
     size=2,
 
-    HasDiv  = 1,
-    HasExp  = 1,
-    HasSqrt = 1
+    HasDiv    = 1
   };
 };
 template<> struct packet_traits<int>    : default_packet_traits
@@ -142,10 +137,6 @@ template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a)
   return psub(_mm_setr_epi32(0,0,0,0), a);
 }
 
-template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; }
-template<> EIGEN_STRONG_INLINE Packet2d pconj(const Packet2d& a) { return a; }
-template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; }
-
 template<> EIGEN_STRONG_INLINE Packet4f pmul<Packet4f>(const Packet4f& a, const Packet4f& b) { return _mm_mul_ps(a,b); }
 template<> EIGEN_STRONG_INLINE Packet2d pmul<Packet2d>(const Packet2d& a, const Packet2d& b) { return _mm_mul_pd(a,b); }
 template<> EIGEN_STRONG_INLINE Packet4i pmul<Packet4i>(const Packet4i& a, const Packet4i& b)
@@ -178,26 +169,18 @@ template<> EIGEN_STRONG_INLINE Packet4f pmin<Packet4f>(const Packet4f& a, const
 template<> EIGEN_STRONG_INLINE Packet2d pmin<Packet2d>(const Packet2d& a, const Packet2d& b) { return _mm_min_pd(a,b); }
 template<> EIGEN_STRONG_INLINE Packet4i pmin<Packet4i>(const Packet4i& a, const Packet4i& b)
 {
-#ifdef EIGEN_VECTORIZE_SSE4_1
-  return _mm_min_epi32(a,b);
-#else
   // after some bench, this version *is* faster than a scalar implementation
   Packet4i mask = _mm_cmplt_epi32(a,b);
   return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b));
-#endif
 }
 
 template<> EIGEN_STRONG_INLINE Packet4f pmax<Packet4f>(const Packet4f& a, const Packet4f& b) { return _mm_max_ps(a,b); }
 template<> EIGEN_STRONG_INLINE Packet2d pmax<Packet2d>(const Packet2d& a, const Packet2d& b) { return _mm_max_pd(a,b); }
 template<> EIGEN_STRONG_INLINE Packet4i pmax<Packet4i>(const Packet4i& a, const Packet4i& b)
 {
-#ifdef EIGEN_VECTORIZE_SSE4_1
-  return _mm_max_epi32(a,b);
-#else
   // after some bench, this version *is* faster than a scalar implementation
   Packet4i mask = _mm_cmpgt_epi32(a,b);
   return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b));
-#endif
 }
 
 template<> EIGEN_STRONG_INLINE Packet4f pand<Packet4f>(const Packet4f& a, const Packet4f& b) { return _mm_and_ps(a,b); }
@@ -235,27 +218,63 @@ template<> EIGEN_STRONG_INLINE Packet4i pload<Packet4i>(const int*     from) { E
     return _mm_loadu_ps(from);
     #endif
   }
+  template<> EIGEN_STRONG_INLINE Packet2d ploadu<Packet2d>(const double* from) { EIGEN_DEBUG_UNALIGNED_LOAD return _mm_loadu_pd(from); }
+  template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int*    from) { EIGEN_DEBUG_UNALIGNED_LOAD return _mm_loadu_si128(reinterpret_cast<const Packet4i*>(from)); }
 #else
+// Fast unaligned loads. Note that here we cannot directly use intrinsics: this would
+// require pointer casting to incompatible pointer types and leads to invalid code
+// because of the strict aliasing rule. The "dummy" stuff are required to enforce
+// a correct instruction dependency.
+// TODO: do the same for MSVC (ICC is compatible)
 // NOTE: with the code below, MSVC's compiler crashes!
 
+#if defined(__GNUC__) && defined(__i386__)
+  // bug 195: gcc/i386 emits weird x87 fldl/fstpl instructions for _mm_load_sd
+  #define EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS 1
+#elif defined(__clang__)
+  // bug 201: Segfaults in __mm_loadh_pd with clang 2.8
+  #define EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS 1
+#else
+  #define EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS 0
+#endif
+
 template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float* from)
 {
   EIGEN_DEBUG_UNALIGNED_LOAD
+#if EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS
   return _mm_loadu_ps(from);
-}
+#else
+  __m128d res;
+  res =  _mm_load_sd((const double*)(from)) ;
+  res =  _mm_loadh_pd(res, (const double*)(from+2)) ;
+  return _mm_castpd_ps(res);
 #endif
-
+}
 template<> EIGEN_STRONG_INLINE Packet2d ploadu<Packet2d>(const double* from)
 {
   EIGEN_DEBUG_UNALIGNED_LOAD
+#if EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS
   return _mm_loadu_pd(from);
+#else
+  __m128d res;
+  res = _mm_load_sd(from) ;
+  res = _mm_loadh_pd(res,from+1);
+  return res;
+#endif
 }
 template<> EIGEN_STRONG_INLINE Packet4i ploadu<Packet4i>(const int* from)
 {
   EIGEN_DEBUG_UNALIGNED_LOAD
-  return _mm_loadu_si128(reinterpret_cast<const __m128i*>(from));
+#if EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS
+  return _mm_loadu_si128(reinterpret_cast<const Packet4i*>(from));
+#else
+  __m128d res;
+  res =  _mm_load_sd((const double*)(from)) ;
+  res =  _mm_loadh_pd(res, (const double*)(from+2)) ;
+  return _mm_castpd_si128(res);
+#endif
 }
-
+#endif
 
 template<> EIGEN_STRONG_INLINE Packet4f ploaddup<Packet4f>(const float*   from)
 {
@@ -472,8 +491,8 @@ template<> EIGEN_STRONG_INLINE int predux_min<Packet4i>(const Packet4i& a)
   // for GCC (eg., it does not like using std::min after the pstore !!)
   EIGEN_ALIGN16 int aux[4];
   pstore(aux, a);
-  int aux0 = aux[0]<aux[1] ? aux[0] : aux[1];
-  int aux2 = aux[2]<aux[3] ? aux[2] : aux[3];
+  register int aux0 = aux[0]<aux[1] ? aux[0] : aux[1];
+  register int aux2 = aux[2]<aux[3] ? aux[2] : aux[3];
   return aux0<aux2 ? aux0 : aux2;
 }
 
@@ -493,8 +512,8 @@ template<> EIGEN_STRONG_INLINE int predux_max<Packet4i>(const Packet4i& a)
   // for GCC (eg., it does not like using std::min after the pstore !!)
   EIGEN_ALIGN16 int aux[4];
   pstore(aux, a);
-  int aux0 = aux[0]>aux[1] ? aux[0] : aux[1];
-  int aux2 = aux[2]>aux[3] ? aux[2] : aux[3];
+  register int aux0 = aux[0]>aux[1] ? aux[0] : aux[1];
+  register int aux2 = aux[2]>aux[3] ? aux[2] : aux[3];
   return aux0>aux2 ? aux0 : aux2;
 }
 

Eigen/src/Core/products/CoeffBasedProduct.h

@@ -90,7 +90,6 @@ struct traits<CoeffBasedProduct<LhsNested,RhsNested,NestingFlags> >
             | (SameType && (CanVectorizeLhs || CanVectorizeRhs) ? PacketAccessBit : 0),
 
       CoeffReadCost = InnerSize == Dynamic ? Dynamic
-                    : InnerSize == 0 ? 0
                     : InnerSize * (NumTraits<Scalar>::MulCost + LhsCoeffReadCost + RhsCoeffReadCost)
                       + (InnerSize - 1) * NumTraits<Scalar>::AddCost,
 
@@ -134,7 +133,7 @@ class CoeffBasedProduct
     };
 
     typedef internal::product_coeff_impl<CanVectorizeInner ? InnerVectorizedTraversal : DefaultTraversal,
-                                   Unroll ? InnerSize : Dynamic,
+                                   Unroll ? InnerSize-1 : Dynamic,
                                    _LhsNested, _RhsNested, Scalar> ScalarCoeffImpl;
 
     typedef CoeffBasedProduct<LhsNested,RhsNested,NestByRefBit> LazyCoeffBasedProductType;
@@ -151,7 +150,7 @@ class CoeffBasedProduct
     {
       // we don't allow taking products of matrices of different real types, as that wouldn't be vectorizable.
       // We still allow to mix T and complex<T>.
-      EIGEN_STATIC_ASSERT((internal::scalar_product_traits<typename Lhs::RealScalar, typename Rhs::RealScalar>::Defined),
+      EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::RealScalar, typename Rhs::RealScalar>::value),
         YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
       eigen_assert(lhs.cols() == rhs.rows()
         && "invalid matrix product"
@@ -185,7 +184,7 @@ class CoeffBasedProduct
     {
       PacketScalar res;
       internal::product_packet_impl<Flags&RowMajorBit ? RowMajor : ColMajor,
-                              Unroll ? InnerSize : Dynamic,
+                              Unroll ? InnerSize-1 : Dynamic,
                               _LhsNested, _RhsNested, PacketScalar, LoadMode>
         ::run(row, col, m_lhs, m_rhs, res);
       return res;
@@ -243,12 +242,12 @@ struct product_coeff_impl<DefaultTraversal, UnrollingIndex, Lhs, Rhs, RetScalar>
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, RetScalar &res)
   {
     product_coeff_impl<DefaultTraversal, UnrollingIndex-1, Lhs, Rhs, RetScalar>::run(row, col, lhs, rhs, res);
-    res += lhs.coeff(row, UnrollingIndex-1) * rhs.coeff(UnrollingIndex-1, col);
+    res += lhs.coeff(row, UnrollingIndex) * rhs.coeff(UnrollingIndex, col);
   }
 };
 
 template<typename Lhs, typename Rhs, typename RetScalar>
-struct product_coeff_impl<DefaultTraversal, 1, Lhs, Rhs, RetScalar>
+struct product_coeff_impl<DefaultTraversal, 0, Lhs, Rhs, RetScalar>
 {
   typedef typename Lhs::Index Index;
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, RetScalar &res)
@@ -257,23 +256,16 @@ struct product_coeff_impl<DefaultTraversal, 1, Lhs, Rhs, RetScalar>
   }
 };
 
-template<typename Lhs, typename Rhs, typename RetScalar>
-struct product_coeff_impl<DefaultTraversal, 0, Lhs, Rhs, RetScalar>
-{
-  typedef typename Lhs::Index Index;
-  static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, RetScalar &res)
-  {
-    res = RetScalar(0);
-  }
-};
-
 template<typename Lhs, typename Rhs, typename RetScalar>
 struct product_coeff_impl<DefaultTraversal, Dynamic, Lhs, Rhs, RetScalar>
 {
   typedef typename Lhs::Index Index;
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, RetScalar& res)
   {
-    res = (lhs.row(row).transpose().cwiseProduct( rhs.col(col) )).sum();
+    eigen_assert(lhs.cols()>0 && "you are using a non initialized matrix");
+    res = lhs.coeff(row, 0) * rhs.coeff(0, col);
+      for(Index i = 1; i < lhs.cols(); ++i)
+        res += lhs.coeff(row, i) * rhs.coeff(i, col);
   }
 };
 
@@ -303,16 +295,6 @@ struct product_coeff_vectorized_unroller<0, Lhs, Rhs, Packet>
   }
 };
 
-template<typename Lhs, typename Rhs, typename RetScalar>
-struct product_coeff_impl<InnerVectorizedTraversal, 0, Lhs, Rhs, RetScalar>
-{
-  typedef typename Lhs::Index Index;
-  static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, RetScalar &res)
-  {
-    res = 0;
-  }
-};
-
 template<int UnrollingIndex, typename Lhs, typename Rhs, typename RetScalar>
 struct product_coeff_impl<InnerVectorizedTraversal, UnrollingIndex, Lhs, Rhs, RetScalar>
 {
@@ -322,7 +304,8 @@ struct product_coeff_impl<InnerVectorizedTraversal, UnrollingIndex, Lhs, Rhs, Re
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, RetScalar &res)
   {
     Packet pres;
-    product_coeff_vectorized_unroller<UnrollingIndex-PacketSize, Lhs, Rhs, Packet>::run(row, col, lhs, rhs, pres);
+    product_coeff_vectorized_unroller<UnrollingIndex+1-PacketSize, Lhs, Rhs, Packet>::run(row, col, lhs, rhs, pres);
+    product_coeff_impl<DefaultTraversal,UnrollingIndex,Lhs,Rhs,RetScalar>::run(row, col, lhs, rhs, res);
     res = predux(pres);
   }
 };
@@ -390,7 +373,7 @@ struct product_packet_impl<RowMajor, UnrollingIndex, Lhs, Rhs, Packet, LoadMode>
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Packet &res)
   {
     product_packet_impl<RowMajor, UnrollingIndex-1, Lhs, Rhs, Packet, LoadMode>::run(row, col, lhs, rhs, res);
-    res =  pmadd(pset1<Packet>(lhs.coeff(row, UnrollingIndex-1)), rhs.template packet<LoadMode>(UnrollingIndex-1, col), res);
+    res =  pmadd(pset1<Packet>(lhs.coeff(row, UnrollingIndex)), rhs.template packet<LoadMode>(UnrollingIndex, col), res);
   }
 };
 
@@ -401,12 +384,12 @@ struct product_packet_impl<ColMajor, UnrollingIndex, Lhs, Rhs, Packet, LoadMode>
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Packet &res)
   {
     product_packet_impl<ColMajor, UnrollingIndex-1, Lhs, Rhs, Packet, LoadMode>::run(row, col, lhs, rhs, res);
-    res =  pmadd(lhs.template packet<LoadMode>(row, UnrollingIndex-1), pset1<Packet>(rhs.coeff(UnrollingIndex-1, col)), res);
+    res =  pmadd(lhs.template packet<LoadMode>(row, UnrollingIndex), pset1<Packet>(rhs.coeff(UnrollingIndex, col)), res);
   }
 };
 
 template<typename Lhs, typename Rhs, typename Packet, int LoadMode>
-struct product_packet_impl<RowMajor, 1, Lhs, Rhs, Packet, LoadMode>
+struct product_packet_impl<RowMajor, 0, Lhs, Rhs, Packet, LoadMode>
 {
   typedef typename Lhs::Index Index;
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Packet &res)
@@ -416,7 +399,7 @@ struct product_packet_impl<RowMajor, 1, Lhs, Rhs, Packet, LoadMode>
 };
 
 template<typename Lhs, typename Rhs, typename Packet, int LoadMode>
-struct product_packet_impl<ColMajor, 1, Lhs, Rhs, Packet, LoadMode>
+struct product_packet_impl<ColMajor, 0, Lhs, Rhs, Packet, LoadMode>
 {
   typedef typename Lhs::Index Index;
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Packet &res)
@@ -425,35 +408,16 @@ struct product_packet_impl<ColMajor, 1, Lhs, Rhs, Packet, LoadMode>
   }
 };
 
-template<typename Lhs, typename Rhs, typename Packet, int LoadMode>
-struct product_packet_impl<RowMajor, 0, Lhs, Rhs, Packet, LoadMode>
-{
-  typedef typename Lhs::Index Index;
-  static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, Packet &res)
-  {
-    res = pset1<Packet>(0);
-  }
-};
-
-template<typename Lhs, typename Rhs, typename Packet, int LoadMode>
-struct product_packet_impl<ColMajor, 0, Lhs, Rhs, Packet, LoadMode>
-{
-  typedef typename Lhs::Index Index;
-  static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, Packet &res)
-  {
-    res = pset1<Packet>(0);
-  }
-};
-
 template<typename Lhs, typename Rhs, typename Packet, int LoadMode>
 struct product_packet_impl<RowMajor, Dynamic, Lhs, Rhs, Packet, LoadMode>
 {
   typedef typename Lhs::Index Index;
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Packet& res)
   {
-    res = pset1<Packet>(0);
-    for(Index i = 0; i < lhs.cols(); ++i)
-      res =  pmadd(pset1<Packet>(lhs.coeff(row, i)), rhs.template packet<LoadMode>(i, col), res);
+    eigen_assert(lhs.cols()>0 && "you are using a non initialized matrix");
+    res = pmul(pset1<Packet>(lhs.coeff(row, 0)),rhs.template packet<LoadMode>(0, col));
+      for(Index i = 1; i < lhs.cols(); ++i)
+        res =  pmadd(pset1<Packet>(lhs.coeff(row, i)), rhs.template packet<LoadMode>(i, col), res);
   }
 };
 
@@ -463,9 +427,10 @@ struct product_packet_impl<ColMajor, Dynamic, Lhs, Rhs, Packet, LoadMode>
   typedef typename Lhs::Index Index;
   static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Packet& res)
   {
-    res = pset1<Packet>(0);
-    for(Index i = 0; i < lhs.cols(); ++i)
-      res =  pmadd(lhs.template packet<LoadMode>(row, i), pset1<Packet>(rhs.coeff(i, col)), res);
+    eigen_assert(lhs.cols()>0 && "you are using a non initialized matrix");
+    res = pmul(lhs.template packet<LoadMode>(row, 0), pset1<Packet>(rhs.coeff(0, col)));
+      for(Index i = 1; i < lhs.cols(); ++i)
+        res =  pmadd(lhs.template packet<LoadMode>(row, i), pset1<Packet>(rhs.coeff(i, col)), res);
   }
 };
 

Eigen/src/Core/products/GeneralBlockPanelKernel.h

@@ -527,16 +527,9 @@ struct gebp_kernel
     ResPacketSize = Traits::ResPacketSize
   };
 
-  EIGEN_DONT_INLINE
+  EIGEN_DONT_INLINE EIGEN_FLATTEN_ATTRIB
   void operator()(ResScalar* res, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index rows, Index depth, Index cols, ResScalar alpha,
-                  Index strideA=-1, Index strideB=-1, Index offsetA=0, Index offsetB=0, RhsScalar* unpackedB=0);
-};
-
-template<typename LhsScalar, typename RhsScalar, typename Index, int mr, int nr, bool ConjugateLhs, bool ConjugateRhs>
-EIGEN_DONT_INLINE
-void gebp_kernel<LhsScalar,RhsScalar,Index,mr,nr,ConjugateLhs,ConjugateRhs>
-  ::operator()(ResScalar* res, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index rows, Index depth, Index cols, ResScalar alpha,
-               Index strideA, Index strideB, Index offsetA, Index offsetB, RhsScalar* unpackedB)
+                  Index strideA=-1, Index strideB=-1, Index offsetA=0, Index offsetB=0, RhsScalar* unpackedB = 0)
   {
     Traits traits;
     
@@ -1096,7 +1089,7 @@ EIGEN_ASM_COMMENT("mybegin4");
       }
     }
   }
-
+};
 
 #undef CJMADD
 
@@ -1117,85 +1110,80 @@ EIGEN_ASM_COMMENT("mybegin4");
 template<typename Scalar, typename Index, int Pack1, int Pack2, int StorageOrder, bool Conjugate, bool PanelMode>
 struct gemm_pack_lhs
 {
-  EIGEN_DONT_INLINE void operator()(Scalar* blockA, const Scalar* EIGEN_RESTRICT _lhs, Index lhsStride, Index depth, Index rows, Index stride=0, Index offset=0);
-};
-
-template<typename Scalar, typename Index, int Pack1, int Pack2, int StorageOrder, bool Conjugate, bool PanelMode>
-EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, Pack1, Pack2, StorageOrder, Conjugate, PanelMode>
-  ::operator()(Scalar* blockA, const Scalar* EIGEN_RESTRICT _lhs, Index lhsStride, Index depth, Index rows, Index stride, Index offset)
-{
-  typedef typename packet_traits<Scalar>::type Packet;
-  enum { PacketSize = packet_traits<Scalar>::size };
-
-  EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK LHS");
-  EIGEN_UNUSED_VARIABLE(stride)
-  EIGEN_UNUSED_VARIABLE(offset)
-  eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride));
-  eigen_assert( (StorageOrder==RowMajor) || ((Pack1%PacketSize)==0 && Pack1<=4*PacketSize) );
-  conj_if<NumTraits<Scalar>::IsComplex && Conjugate> cj;
-  const_blas_data_mapper<Scalar, Index, StorageOrder> lhs(_lhs,lhsStride);
-  Index count = 0;
-  Index peeled_mc = (rows/Pack1)*Pack1;
-  for(Index i=0; i<peeled_mc; i+=Pack1)
+  EIGEN_DONT_INLINE void operator()(Scalar* blockA, const Scalar* EIGEN_RESTRICT _lhs, Index lhsStride, Index depth, Index rows,
+                  Index stride=0, Index offset=0)
   {
-    if(PanelMode) count += Pack1 * offset;
+    typedef typename packet_traits<Scalar>::type Packet;
+    enum { PacketSize = packet_traits<Scalar>::size };
 
-    if(StorageOrder==ColMajor)
+    EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK LHS");
+    eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride));
+    eigen_assert( (StorageOrder==RowMajor) || ((Pack1%PacketSize)==0 && Pack1<=4*PacketSize) );
+    conj_if<NumTraits<Scalar>::IsComplex && Conjugate> cj;
+    const_blas_data_mapper<Scalar, Index, StorageOrder> lhs(_lhs,lhsStride);
+    Index count = 0;
+    Index peeled_mc = (rows/Pack1)*Pack1;
+    for(Index i=0; i<peeled_mc; i+=Pack1)
     {
-      for(Index k=0; k<depth; k++)
+      if(PanelMode) count += Pack1 * offset;
+
+      if(StorageOrder==ColMajor)
       {
-        Packet A, B, C, D;
-        if(Pack1>=1*PacketSize) A = ploadu<Packet>(&lhs(i+0*PacketSize, k));
-        if(Pack1>=2*PacketSize) B = ploadu<Packet>(&lhs(i+1*PacketSize, k));
-        if(Pack1>=3*PacketSize) C = ploadu<Packet>(&lhs(i+2*PacketSize, k));
-        if(Pack1>=4*PacketSize) D = ploadu<Packet>(&lhs(i+3*PacketSize, k));
-        if(Pack1>=1*PacketSize) { pstore(blockA+count, cj.pconj(A)); count+=PacketSize; }
-        if(Pack1>=2*PacketSize) { pstore(blockA+count, cj.pconj(B)); count+=PacketSize; }
-        if(Pack1>=3*PacketSize) { pstore(blockA+count, cj.pconj(C)); count+=PacketSize; }
-        if(Pack1>=4*PacketSize) { pstore(blockA+count, cj.pconj(D)); count+=PacketSize; }
-      }
-    }
-    else
-    {
-      for(Index k=0; k<depth; k++)
-      {
-        // TODO add a vectorized transpose here
-        Index w=0;
-        for(; w<Pack1-3; w+=4)
+        for(Index k=0; k<depth; k++)
         {
-          Scalar a(cj(lhs(i+w+0, k))),
-                  b(cj(lhs(i+w+1, k))),
-                  c(cj(lhs(i+w+2, k))),
-                  d(cj(lhs(i+w+3, k)));
-          blockA[count++] = a;
-          blockA[count++] = b;
-          blockA[count++] = c;
-          blockA[count++] = d;
+          Packet A, B, C, D;
+          if(Pack1>=1*PacketSize) A = ploadu<Packet>(&lhs(i+0*PacketSize, k));
+          if(Pack1>=2*PacketSize) B = ploadu<Packet>(&lhs(i+1*PacketSize, k));
+          if(Pack1>=3*PacketSize) C = ploadu<Packet>(&lhs(i+2*PacketSize, k));
+          if(Pack1>=4*PacketSize) D = ploadu<Packet>(&lhs(i+3*PacketSize, k));
+          if(Pack1>=1*PacketSize) { pstore(blockA+count, cj.pconj(A)); count+=PacketSize; }
+          if(Pack1>=2*PacketSize) { pstore(blockA+count, cj.pconj(B)); count+=PacketSize; }
+          if(Pack1>=3*PacketSize) { pstore(blockA+count, cj.pconj(C)); count+=PacketSize; }
+          if(Pack1>=4*PacketSize) { pstore(blockA+count, cj.pconj(D)); count+=PacketSize; }
         }
-        if(Pack1%4)
-          for(;w<Pack1;++w)
-            blockA[count++] = cj(lhs(i+w, k));
       }
+      else
+      {
+        for(Index k=0; k<depth; k++)
+        {
+          // TODO add a vectorized transpose here
+          Index w=0;
+          for(; w<Pack1-3; w+=4)
+          {
+            Scalar a(cj(lhs(i+w+0, k))),
+                   b(cj(lhs(i+w+1, k))),
+                   c(cj(lhs(i+w+2, k))),
+                   d(cj(lhs(i+w+3, k)));
+            blockA[count++] = a;
+            blockA[count++] = b;
+            blockA[count++] = c;
+            blockA[count++] = d;
+          }
+          if(Pack1%4)
+            for(;w<Pack1;++w)
+              blockA[count++] = cj(lhs(i+w, k));
+        }
+      }
+      if(PanelMode) count += Pack1 * (stride-offset-depth);
+    }
+    if(rows-peeled_mc>=Pack2)
+    {
+      if(PanelMode) count += Pack2*offset;
+      for(Index k=0; k<depth; k++)
+        for(Index w=0; w<Pack2; w++)
+          blockA[count++] = cj(lhs(peeled_mc+w, k));
+      if(PanelMode) count += Pack2 * (stride-offset-depth);
+      peeled_mc += Pack2;
+    }
+    for(Index i=peeled_mc; i<rows; i++)
+    {
+      if(PanelMode) count += offset;
+      for(Index k=0; k<depth; k++)
+        blockA[count++] = cj(lhs(i, k));
+      if(PanelMode) count += (stride-offset-depth);
     }
-    if(PanelMode) count += Pack1 * (stride-offset-depth);
   }
-  if(rows-peeled_mc>=Pack2)
-  {
-    if(PanelMode) count += Pack2*offset;
-    for(Index k=0; k<depth; k++)
-      for(Index w=0; w<Pack2; w++)
-        blockA[count++] = cj(lhs(peeled_mc+w, k));
-    if(PanelMode) count += Pack2 * (stride-offset-depth);
-    peeled_mc += Pack2;
-  }
-  for(Index i=peeled_mc; i<rows; i++)
-  {
-    if(PanelMode) count += offset;
-    for(Index k=0; k<depth; k++)
-      blockA[count++] = cj(lhs(i, k));
-    if(PanelMode) count += (stride-offset-depth);
-  }
-}
+};
 
 // copy a complete panel of the rhs
 // this version is optimized for column major matrices
@@ -1209,103 +1197,93 @@ struct gemm_pack_rhs<Scalar, Index, nr, ColMajor, Conjugate, PanelMode>
 {
   typedef typename packet_traits<Scalar>::type Packet;
   enum { PacketSize = packet_traits<Scalar>::size };
-  EIGEN_DONT_INLINE void operator()(Scalar* blockB, const Scalar* rhs, Index rhsStride, Index depth, Index cols, Index stride=0, Index offset=0);
+  EIGEN_DONT_INLINE void operator()(Scalar* blockB, const Scalar* rhs, Index rhsStride, Index depth, Index cols,
+                  Index stride=0, Index offset=0)
+  {
+    EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS COLMAJOR");
+    eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride));
+    conj_if<NumTraits<Scalar>::IsComplex && Conjugate> cj;
+    Index packet_cols = (cols/nr) * nr;
+    Index count = 0;
+    for(Index j2=0; j2<packet_cols; j2+=nr)
+    {
+      // skip what we have before
+      if(PanelMode) count += nr * offset;
+      const Scalar* b0 = &rhs[(j2+0)*rhsStride];
+      const Scalar* b1 = &rhs[(j2+1)*rhsStride];
+      const Scalar* b2 = &rhs[(j2+2)*rhsStride];
+      const Scalar* b3 = &rhs[(j2+3)*rhsStride];
+      for(Index k=0; k<depth; k++)
+      {
+                  blockB[count+0] = cj(b0[k]);
+                  blockB[count+1] = cj(b1[k]);
+        if(nr==4) blockB[count+2] = cj(b2[k]);
+        if(nr==4) blockB[count+3] = cj(b3[k]);
+        count += nr;
+      }
+      // skip what we have after
+      if(PanelMode) count += nr * (stride-offset-depth);
+    }
+
+    // copy the remaining columns one at a time (nr==1)
+    for(Index j2=packet_cols; j2<cols; ++j2)
+    {
+      if(PanelMode) count += offset;
+      const Scalar* b0 = &rhs[(j2+0)*rhsStride];
+      for(Index k=0; k<depth; k++)
+      {
+        blockB[count] = cj(b0[k]);
+        count += 1;
+      }
+      if(PanelMode) count += (stride-offset-depth);
+    }
+  }
 };
 
-template<typename Scalar, typename Index, int nr, bool Conjugate, bool PanelMode>
-EIGEN_DONT_INLINE void gemm_pack_rhs<Scalar, Index, nr, ColMajor, Conjugate, PanelMode>
-  ::operator()(Scalar* blockB, const Scalar* rhs, Index rhsStride, Index depth, Index cols, Index stride, Index offset)
-{
-  EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS COLMAJOR");
-  EIGEN_UNUSED_VARIABLE(stride)
-  EIGEN_UNUSED_VARIABLE(offset)
-  eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride));
-  conj_if<NumTraits<Scalar>::IsComplex && Conjugate> cj;
-  Index packet_cols = (cols/nr) * nr;
-  Index count = 0;
-  for(Index j2=0; j2<packet_cols; j2+=nr)
-  {
-    // skip what we have before
-    if(PanelMode) count += nr * offset;
-    const Scalar* b0 = &rhs[(j2+0)*rhsStride];
-    const Scalar* b1 = &rhs[(j2+1)*rhsStride];
-    const Scalar* b2 = &rhs[(j2+2)*rhsStride];
-    const Scalar* b3 = &rhs[(j2+3)*rhsStride];
-    for(Index k=0; k<depth; k++)
-    {
-                blockB[count+0] = cj(b0[k]);
-                blockB[count+1] = cj(b1[k]);
-      if(nr==4) blockB[count+2] = cj(b2[k]);
-      if(nr==4) blockB[count+3] = cj(b3[k]);
-      count += nr;
-    }
-    // skip what we have after
-    if(PanelMode) count += nr * (stride-offset-depth);
-  }
-
-  // copy the remaining columns one at a time (nr==1)
-  for(Index j2=packet_cols; j2<cols; ++j2)
-  {
-    if(PanelMode) count += offset;
-    const Scalar* b0 = &rhs[(j2+0)*rhsStride];
-    for(Index k=0; k<depth; k++)
-    {
-      blockB[count] = cj(b0[k]);
-      count += 1;
-    }
-    if(PanelMode) count += (stride-offset-depth);
-  }
-}
-
 // this version is optimized for row major matrices
 template<typename Scalar, typename Index, int nr, bool Conjugate, bool PanelMode>
 struct gemm_pack_rhs<Scalar, Index, nr, RowMajor, Conjugate, PanelMode>
 {
   enum { PacketSize = packet_traits<Scalar>::size };
-  EIGEN_DONT_INLINE void operator()(Scalar* blockB, const Scalar* rhs, Index rhsStride, Index depth, Index cols, Index stride=0, Index offset=0);
+  EIGEN_DONT_INLINE void operator()(Scalar* blockB, const Scalar* rhs, Index rhsStride, Index depth, Index cols,
+                  Index stride=0, Index offset=0)
+  {
+    EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR");
+    eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride));
+    conj_if<NumTraits<Scalar>::IsComplex && Conjugate> cj;
+    Index packet_cols = (cols/nr) * nr;
+    Index count = 0;
+    for(Index j2=0; j2<packet_cols; j2+=nr)
+    {
+      // skip what we have before
+      if(PanelMode) count += nr * offset;
+      for(Index k=0; k<depth; k++)
+      {
+        const Scalar* b0 = &rhs[k*rhsStride + j2];
+                  blockB[count+0] = cj(b0[0]);
+                  blockB[count+1] = cj(b0[1]);
+        if(nr==4) blockB[count+2] = cj(b0[2]);
+        if(nr==4) blockB[count+3] = cj(b0[3]);
+        count += nr;
+      }
+      // skip what we have after
+      if(PanelMode) count += nr * (stride-offset-depth);
+    }
+    // copy the remaining columns one at a time (nr==1)
+    for(Index j2=packet_cols; j2<cols; ++j2)
+    {
+      if(PanelMode) count += offset;
+      const Scalar* b0 = &rhs[j2];
+      for(Index k=0; k<depth; k++)
+      {
+        blockB[count] = cj(b0[k*rhsStride]);
+        count += 1;
+      }
+      if(PanelMode) count += stride-offset-depth;
+    }
+  }
 };
 
-template<typename Scalar, typename Index, int nr, bool Conjugate, bool PanelMode>
-EIGEN_DONT_INLINE void gemm_pack_rhs<Scalar, Index, nr, RowMajor, Conjugate, PanelMode>
-  ::operator()(Scalar* blockB, const Scalar* rhs, Index rhsStride, Index depth, Index cols, Index stride, Index offset)
-{
-  EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR");
-  EIGEN_UNUSED_VARIABLE(stride)
-  EIGEN_UNUSED_VARIABLE(offset)
-  eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride));
-  conj_if<NumTraits<Scalar>::IsComplex && Conjugate> cj;
-  Index packet_cols = (cols/nr) * nr;
-  Index count = 0;
-  for(Index j2=0; j2<packet_cols; j2+=nr)
-  {
-    // skip what we have before
-    if(PanelMode) count += nr * offset;
-    for(Index k=0; k<depth; k++)
-    {
-      const Scalar* b0 = &rhs[k*rhsStride + j2];
-                blockB[count+0] = cj(b0[0]);
-                blockB[count+1] = cj(b0[1]);
-      if(nr==4) blockB[count+2] = cj(b0[2]);
-      if(nr==4) blockB[count+3] = cj(b0[3]);
-      count += nr;
-    }
-    // skip what we have after
-    if(PanelMode) count += nr * (stride-offset-depth);
-  }
-  // copy the remaining columns one at a time (nr==1)
-  for(Index j2=packet_cols; j2<cols; ++j2)
-  {
-    if(PanelMode) count += offset;
-    const Scalar* b0 = &rhs[j2];
-    for(Index k=0; k<depth; k++)
-    {
-      blockB[count] = cj(b0[k*rhsStride]);
-      count += 1;
-    }
-    if(PanelMode) count += stride-offset-depth;
-  }
-}
-
 } // end namespace internal
 
 /** \returns the currently set level 1 cpu cache size (in bytes) used to estimate the ideal blocking size parameters.

Eigen/src/Core/products/GeneralMatrixMatrix.h

@@ -50,7 +50,6 @@ template<
   typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs>
 struct general_matrix_matrix_product<Index,LhsScalar,LhsStorageOrder,ConjugateLhs,RhsScalar,RhsStorageOrder,ConjugateRhs,ColMajor>
 {
-
 typedef typename scalar_product_traits<LhsScalar, RhsScalar>::ReturnType ResScalar;
 static void run(Index rows, Index cols, Index depth,
   const LhsScalar* _lhs, Index lhsStride,
@@ -140,10 +139,8 @@ static void run(Index rows, Index cols, Index depth,
       // Release all the sub blocks B'_j of B' for the current thread,
       // i.e., we simply decrement the number of users by 1
       for(Index j=0; j<threads; ++j)
-      {
         #pragma omp atomic
-        info[j].users -= 1;
-      }
+        --(info[j].users);
     }
   }
   else
@@ -172,6 +169,7 @@ static void run(Index rows, Index cols, Index depth,
       // vertical panel which is, in practice, a very low number.
       pack_rhs(blockB, &rhs(k2,0), rhsStride, actual_kc, cols);
 
+
       // For each mc x kc block of the lhs's vertical panel...
       // (==GEPP_VAR1)
       for(Index i2=0; i2<rows; i2+=mc)
@@ -185,6 +183,7 @@ static void run(Index rows, Index cols, Index depth,
 
         // Everything is packed, we can now call the block * panel kernel:
         gebp(res+i2, resStride, blockA, blockB, actual_mc, actual_kc, cols, alpha, -1, -1, 0, 0, blockW);
+
       }
     }
   }
@@ -205,7 +204,7 @@ struct traits<GeneralProduct<Lhs,Rhs,GemmProduct> >
 template<typename Scalar, typename Index, typename Gemm, typename Lhs, typename Rhs, typename Dest, typename BlockingType>
 struct gemm_functor
 {
-  gemm_functor(const Lhs& lhs, const Rhs& rhs, Dest& dest, const Scalar& actualAlpha,
+  gemm_functor(const Lhs& lhs, const Rhs& rhs, Dest& dest, Scalar actualAlpha,
                   BlockingType& blocking)
     : m_lhs(lhs), m_rhs(rhs), m_dest(dest), m_actualAlpha(actualAlpha), m_blocking(blocking)
   {}
@@ -392,17 +391,13 @@ class GeneralProduct<Lhs, Rhs, GemmProduct>
 
     GeneralProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs)
     {
-#if !(defined(EIGEN_NO_STATIC_ASSERT) && defined(EIGEN_NO_DEBUG))
       typedef internal::scalar_product_op<LhsScalar,RhsScalar> BinOp;
       EIGEN_CHECK_BINARY_COMPATIBILIY(BinOp,LhsScalar,RhsScalar);
-#endif
     }
 
-    template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
+    template<typename Dest> void scaleAndAddTo(Dest& dst, Scalar alpha) const
     {
       eigen_assert(dst.rows()==m_lhs.rows() && dst.cols()==m_rhs.cols());
-      if(m_lhs.cols()==0 || m_lhs.rows()==0 || m_rhs.cols()==0)
-        return;
 
       typename internal::add_const_on_value_type<ActualLhsType>::type lhs = LhsBlasTraits::extract(m_lhs);
       typename internal::add_const_on_value_type<ActualRhsType>::type rhs = RhsBlasTraits::extract(m_rhs);

Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h

@@ -12,9 +12,6 @@
 
 namespace Eigen { 
 
-template<typename Scalar, typename Index, int StorageOrder, int UpLo, bool ConjLhs, bool ConjRhs>
-struct selfadjoint_rank1_update;
-
 namespace internal {
 
 /**********************************************************************
@@ -42,7 +39,7 @@ struct general_matrix_matrix_triangular_product<Index,LhsScalar,LhsStorageOrder,
 {
   typedef typename scalar_product_traits<LhsScalar, RhsScalar>::ReturnType ResScalar;
   static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* lhs, Index lhsStride,
-                                      const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resStride, const ResScalar& alpha)
+                                      const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resStride, ResScalar alpha)
   {
     general_matrix_matrix_triangular_product<Index,
         RhsScalar, RhsStorageOrder==RowMajor ? ColMajor : RowMajor, ConjugateRhs,
@@ -58,7 +55,7 @@ struct general_matrix_matrix_triangular_product<Index,LhsScalar,LhsStorageOrder,
 {
   typedef typename scalar_product_traits<LhsScalar, RhsScalar>::ReturnType ResScalar;
   static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* _lhs, Index lhsStride,
-                                      const RhsScalar* _rhs, Index rhsStride, ResScalar* res, Index resStride, const ResScalar& alpha)
+                                      const RhsScalar* _rhs, Index rhsStride, ResScalar* res, Index resStride, ResScalar alpha)
   {
     const_blas_data_mapper<LhsScalar, Index, LhsStorageOrder> lhs(_lhs,lhsStride);
     const_blas_data_mapper<RhsScalar, Index, RhsStorageOrder> rhs(_rhs,rhsStride);
@@ -136,7 +133,7 @@ struct tribb_kernel
   enum {
     BlockSize  = EIGEN_PLAIN_ENUM_MAX(mr,nr)
   };
-  void operator()(ResScalar* res, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha, RhsScalar* workspace)
+  void operator()(ResScalar* res, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, ResScalar alpha, RhsScalar* workspace)
   {
     gebp_kernel<LhsScalar, RhsScalar, Index, mr, nr, ConjLhs, ConjRhs> gebp_kernel;
     Matrix<ResScalar,BlockSize,BlockSize,ColMajor> buffer;
@@ -183,92 +180,31 @@ struct tribb_kernel
 
 // high level API
 
-template<typename MatrixType, typename ProductType, int UpLo, bool IsOuterProduct>
-struct general_product_to_triangular_selector;
-
-
-template<typename MatrixType, typename ProductType, int UpLo>
-struct general_product_to_triangular_selector<MatrixType,ProductType,UpLo,true>
-{
-  static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha)
-  {
-    typedef typename MatrixType::Scalar Scalar;
-    typedef typename MatrixType::Index Index;
-    
-    typedef typename internal::remove_all<typename ProductType::LhsNested>::type Lhs;
-    typedef internal::blas_traits<Lhs> LhsBlasTraits;
-    typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs;
-    typedef typename internal::remove_all<ActualLhs>::type _ActualLhs;
-    typename internal::add_const_on_value_type<ActualLhs>::type actualLhs = LhsBlasTraits::extract(prod.lhs());
-    
-    typedef typename internal::remove_all<typename ProductType::RhsNested>::type Rhs;
-    typedef internal::blas_traits<Rhs> RhsBlasTraits;
-    typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs;
-    typedef typename internal::remove_all<ActualRhs>::type _ActualRhs;
-    typename internal::add_const_on_value_type<ActualRhs>::type actualRhs = RhsBlasTraits::extract(prod.rhs());
-
-    Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived());
-
-    enum {
-      StorageOrder = (internal::traits<MatrixType>::Flags&RowMajorBit) ? RowMajor : ColMajor,
-      UseLhsDirectly = _ActualLhs::InnerStrideAtCompileTime==1,
-      UseRhsDirectly = _ActualRhs::InnerStrideAtCompileTime==1
-    };
-    
-    internal::gemv_static_vector_if<Scalar,Lhs::SizeAtCompileTime,Lhs::MaxSizeAtCompileTime,!UseLhsDirectly> static_lhs;
-    ei_declare_aligned_stack_constructed_variable(Scalar, actualLhsPtr, actualLhs.size(),
-      (UseLhsDirectly ? const_cast<Scalar*>(actualLhs.data()) : static_lhs.data()));
-    if(!UseLhsDirectly) Map<typename _ActualLhs::PlainObject>(actualLhsPtr, actualLhs.size()) = actualLhs;
-    
-    internal::gemv_static_vector_if<Scalar,Rhs::SizeAtCompileTime,Rhs::MaxSizeAtCompileTime,!UseRhsDirectly> static_rhs;
-    ei_declare_aligned_stack_constructed_variable(Scalar, actualRhsPtr, actualRhs.size(),
-      (UseRhsDirectly ? const_cast<Scalar*>(actualRhs.data()) : static_rhs.data()));
-    if(!UseRhsDirectly) Map<typename _ActualRhs::PlainObject>(actualRhsPtr, actualRhs.size()) = actualRhs;
-    
-    
-    selfadjoint_rank1_update<Scalar,Index,StorageOrder,UpLo,
-                              LhsBlasTraits::NeedToConjugate && NumTraits<Scalar>::IsComplex,
-                              RhsBlasTraits::NeedToConjugate && NumTraits<Scalar>::IsComplex>
-          ::run(actualLhs.size(), mat.data(), mat.outerStride(), actualLhsPtr, actualRhsPtr, actualAlpha);
-  }
-};
-
-template<typename MatrixType, typename ProductType, int UpLo>
-struct general_product_to_triangular_selector<MatrixType,ProductType,UpLo,false>
-{
-  static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha)
-  {
-    typedef typename MatrixType::Index Index;
-    
-    typedef typename internal::remove_all<typename ProductType::LhsNested>::type Lhs;
-    typedef internal::blas_traits<Lhs> LhsBlasTraits;
-    typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs;
-    typedef typename internal::remove_all<ActualLhs>::type _ActualLhs;
-    typename internal::add_const_on_value_type<ActualLhs>::type actualLhs = LhsBlasTraits::extract(prod.lhs());
-    
-    typedef typename internal::remove_all<typename ProductType::RhsNested>::type Rhs;
-    typedef internal::blas_traits<Rhs> RhsBlasTraits;
-    typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs;
-    typedef typename internal::remove_all<ActualRhs>::type _ActualRhs;
-    typename internal::add_const_on_value_type<ActualRhs>::type actualRhs = RhsBlasTraits::extract(prod.rhs());
-
-    typename ProductType::Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived());
-
-    internal::general_matrix_matrix_triangular_product<Index,
-      typename Lhs::Scalar, _ActualLhs::Flags&RowMajorBit ? RowMajor : ColMajor, LhsBlasTraits::NeedToConjugate,
-      typename Rhs::Scalar, _ActualRhs::Flags&RowMajorBit ? RowMajor : ColMajor, RhsBlasTraits::NeedToConjugate,
-      MatrixType::Flags&RowMajorBit ? RowMajor : ColMajor, UpLo>
-      ::run(mat.cols(), actualLhs.cols(),
-            &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &actualRhs.coeffRef(0,0), actualRhs.outerStride(),
-            mat.data(), mat.outerStride(), actualAlpha);
-  }
-};
-
 template<typename MatrixType, unsigned int UpLo>
 template<typename ProductDerived, typename _Lhs, typename _Rhs>
 TriangularView<MatrixType,UpLo>& TriangularView<MatrixType,UpLo>::assignProduct(const ProductBase<ProductDerived, _Lhs,_Rhs>& prod, const Scalar& alpha)
 {
-  general_product_to_triangular_selector<MatrixType, ProductDerived, UpLo, (_Lhs::ColsAtCompileTime==1) || (_Rhs::RowsAtCompileTime==1)>::run(m_matrix.const_cast_derived(), prod.derived(), alpha);
+  typedef typename internal::remove_all<typename ProductDerived::LhsNested>::type Lhs;
+  typedef internal::blas_traits<Lhs> LhsBlasTraits;
+  typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs;
+  typedef typename internal::remove_all<ActualLhs>::type _ActualLhs;
+  typename internal::add_const_on_value_type<ActualLhs>::type actualLhs = LhsBlasTraits::extract(prod.lhs());
+  
+  typedef typename internal::remove_all<typename ProductDerived::RhsNested>::type Rhs;
+  typedef internal::blas_traits<Rhs> RhsBlasTraits;
+  typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs;
+  typedef typename internal::remove_all<ActualRhs>::type _ActualRhs;
+  typename internal::add_const_on_value_type<ActualRhs>::type actualRhs = RhsBlasTraits::extract(prod.rhs());
+
+  typename ProductDerived::Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived());
+
+  internal::general_matrix_matrix_triangular_product<Index,
+    typename Lhs::Scalar, _ActualLhs::Flags&RowMajorBit ? RowMajor : ColMajor, LhsBlasTraits::NeedToConjugate,
+    typename Rhs::Scalar, _ActualRhs::Flags&RowMajorBit ? RowMajor : ColMajor, RhsBlasTraits::NeedToConjugate,
+    MatrixType::Flags&RowMajorBit ? RowMajor : ColMajor, UpLo>
+    ::run(m_matrix.cols(), actualLhs.cols(),
+          &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &actualRhs.coeffRef(0,0), actualRhs.outerStride(),
+          const_cast<Scalar*>(m_matrix.data()), m_matrix.outerStride(), actualAlpha);
   
   return *this;
 }

Eigen/src/Core/products/GeneralMatrixVector.h

@@ -52,17 +52,12 @@ EIGEN_DONT_INLINE static void run(
   Index rows, Index cols,
   const LhsScalar* lhs, Index lhsStride,
   const RhsScalar* rhs, Index rhsIncr,
-  ResScalar* res, Index resIncr, RhsScalar alpha);
-};
-
-template<typename Index, typename LhsScalar, bool ConjugateLhs, typename RhsScalar, bool ConjugateRhs, int Version>
-EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,ColMajor,ConjugateLhs,RhsScalar,ConjugateRhs,Version>::run(
-  Index rows, Index cols,
-  const LhsScalar* lhs, Index lhsStride,
-  const RhsScalar* rhs, Index rhsIncr,
-  ResScalar* res, Index resIncr, RhsScalar alpha)
+  ResScalar* res, Index
+  #ifdef EIGEN_INTERNAL_DEBUGGING
+    resIncr
+  #endif
+  , RhsScalar alpha)
 {
-  EIGEN_UNUSED_VARIABLE(resIncr)
   eigen_internal_assert(resIncr==1);
   #ifdef _EIGEN_ACCUMULATE_PACKETS
   #error _EIGEN_ACCUMULATE_PACKETS has already been defined
@@ -79,14 +74,13 @@ EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,ColMajor,Co
   conj_helper<LhsScalar,RhsScalar,ConjugateLhs,ConjugateRhs> cj;
   conj_helper<LhsPacket,RhsPacket,ConjugateLhs,ConjugateRhs> pcj;
   if(ConjugateRhs)
-    alpha = numext::conj(alpha);
+    alpha = conj(alpha);
 
   enum { AllAligned = 0, EvenAligned, FirstAligned, NoneAligned };
   const Index columnsAtOnce = 4;
   const Index peels = 2;
   const Index LhsPacketAlignedMask = LhsPacketSize-1;
   const Index ResPacketAlignedMask = ResPacketSize-1;
-//  const Index PeelAlignedMask = ResPacketSize*peels-1;
   const Index size = rows;
   
   // How many coeffs of the result do we have to skip to be aligned.
@@ -258,7 +252,7 @@ EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,ColMajor,Co
         // process aligned result's coeffs
         if ((size_t(lhs0+alignedStart)%sizeof(LhsPacket))==0)
           for (Index i = alignedStart;i<alignedSize;i+=ResPacketSize)
-            pstore(&res[i], pcj.pmadd(pload<LhsPacket>(&lhs0[i]), ptmp0, pload<ResPacket>(&res[i])));
+            pstore(&res[i], pcj.pmadd(ploadu<LhsPacket>(&lhs0[i]), ptmp0, pload<ResPacket>(&res[i])));
         else
           for (Index i = alignedStart;i<alignedSize;i+=ResPacketSize)
             pstore(&res[i], pcj.pmadd(ploadu<LhsPacket>(&lhs0[i]), ptmp0, pload<ResPacket>(&res[i])));
@@ -279,6 +273,7 @@ EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,ColMajor,Co
   } while(Vectorizable);
   #undef _EIGEN_ACCUMULATE_PACKETS
 }
+};
 
 /* Optimized row-major matrix * vector product:
  * This algorithm processes 4 rows at onces that allows to both reduce
@@ -312,15 +307,6 @@ typedef typename conditional<Vectorizable,_RhsPacket,RhsScalar>::type RhsPacket;
 typedef typename conditional<Vectorizable,_ResPacket,ResScalar>::type ResPacket;
   
 EIGEN_DONT_INLINE static void run(
-  Index rows, Index cols,
-  const LhsScalar* lhs, Index lhsStride,
-  const RhsScalar* rhs, Index rhsIncr,
-  ResScalar* res, Index resIncr,
-  ResScalar alpha);
-};
-
-template<typename Index, typename LhsScalar, bool ConjugateLhs, typename RhsScalar, bool ConjugateRhs, int Version>
-EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,RowMajor,ConjugateLhs,RhsScalar,ConjugateRhs,Version>::run(
   Index rows, Index cols,
   const LhsScalar* lhs, Index lhsStride,
   const RhsScalar* rhs, Index rhsIncr,
@@ -348,7 +334,6 @@ EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,RowMajor,Co
   const Index peels = 2;
   const Index RhsPacketAlignedMask = RhsPacketSize-1;
   const Index LhsPacketAlignedMask = LhsPacketSize-1;
-//   const Index PeelAlignedMask = RhsPacketSize*peels-1;
   const Index depth = cols;
 
   // How many coeffs of the result do we have to skip to be aligned.
@@ -558,6 +543,7 @@ EIGEN_DONT_INLINE void general_matrix_vector_product<Index,LhsScalar,RowMajor,Co
 
   #undef _EIGEN_ACCUMULATE_PACKETS
 }
+};
 
 } // end namespace internal
 

Eigen/src/Core/products/GeneralMatrixVector_MKL.h

@@ -53,7 +53,7 @@ struct general_matrix_vector_product_gemv :
 #define EIGEN_MKL_GEMV_SPECIALIZE(Scalar) \
 template<typename Index, bool ConjugateLhs, bool ConjugateRhs> \
 struct general_matrix_vector_product<Index,Scalar,ColMajor,ConjugateLhs,Scalar,ConjugateRhs,Specialized> { \
-static void run( \
+static EIGEN_DONT_INLINE void run( \
   Index rows, Index cols, \
   const Scalar* lhs, Index lhsStride, \
   const Scalar* rhs, Index rhsIncr, \
@@ -70,7 +70,7 @@ static void run( \
 }; \
 template<typename Index, bool ConjugateLhs, bool ConjugateRhs> \
 struct general_matrix_vector_product<Index,Scalar,RowMajor,ConjugateLhs,Scalar,ConjugateRhs,Specialized> { \
-static void run( \
+static EIGEN_DONT_INLINE void run( \
   Index rows, Index cols, \
   const Scalar* lhs, Index lhsStride, \
   const Scalar* rhs, Index rhsIncr, \
@@ -92,7 +92,7 @@ struct general_matrix_vector_product_gemv<Index,EIGTYPE,LhsStorageOrder,Conjugat
 { \
 typedef Matrix<EIGTYPE,Dynamic,1,ColMajor> GEMVVector;\
 \
-static void run( \
+static EIGEN_DONT_INLINE void run( \
   Index rows, Index cols, \
   const EIGTYPE* lhs, Index lhsStride, \
   const EIGTYPE* rhs, Index rhsIncr, \

Eigen/src/Core/products/Parallelizer.h

@@ -125,22 +125,19 @@ void parallelize_gemm(const Functor& func, Index rows, Index cols, bool transpos
   if(transpose)
     std::swap(rows,cols);
 
+  Index blockCols = (cols / threads) & ~Index(0x3);
+  Index blockRows = (rows / threads) & ~Index(0x7);
+  
   GemmParallelInfo<Index>* info = new GemmParallelInfo<Index>[threads];
 
-  #pragma omp parallel num_threads(threads)
+  #pragma omp parallel for schedule(static,1) num_threads(threads)
+  for(Index i=0; i<threads; ++i)
   {
-    Index i = omp_get_thread_num();
-    // Note that the actual number of threads might be lower than the number of request ones.
-    Index actual_threads = omp_get_num_threads();
-    
-    Index blockCols = (cols / actual_threads) & ~Index(0x3);
-    Index blockRows = (rows / actual_threads) & ~Index(0x7);
-    
     Index r0 = i*blockRows;
-    Index actualBlockRows = (i+1==actual_threads) ? rows-r0 : blockRows;
+    Index actualBlockRows = (i+1==threads) ? rows-r0 : blockRows;
 
     Index c0 = i*blockCols;
-    Index actualBlockCols = (i+1==actual_threads) ? cols-c0 : blockCols;
+    Index actualBlockCols = (i+1==threads) ? cols-c0 : blockCols;
 
     info[i].rhs_start = c0;
     info[i].rhs_length = actualBlockCols;

Eigen/src/Core/products/SelfadjointMatrixMatrix.h

@@ -30,9 +30,9 @@ struct symm_pack_lhs
     for(Index k=i; k<i+BlockRows; k++)
     {
       for(Index w=0; w<h; w++)
-        blockA[count++] = numext::conj(lhs(k, i+w)); // transposed
+        blockA[count++] = conj(lhs(k, i+w)); // transposed
 
-      blockA[count++] = numext::real(lhs(k,k));   // real (diagonal)
+      blockA[count++] = real(lhs(k,k));   // real (diagonal)
 
       for(Index w=h+1; w<BlockRows; w++)
         blockA[count++] = lhs(i+w, k);          // normal
@@ -41,7 +41,7 @@ struct symm_pack_lhs
     // transposed copy
     for(Index k=i+BlockRows; k<cols; k++)
       for(Index w=0; w<BlockRows; w++)
-        blockA[count++] = numext::conj(lhs(k, i+w)); // transposed
+        blockA[count++] = conj(lhs(k, i+w)); // transposed
   }
   void operator()(Scalar* blockA, const Scalar* _lhs, Index lhsStride, Index cols, Index rows)
   {
@@ -65,10 +65,10 @@ struct symm_pack_lhs
       for(Index k=0; k<i; k++)
         blockA[count++] = lhs(i, k);              // normal
 
-      blockA[count++] = numext::real(lhs(i, i));       // real (diagonal)
+      blockA[count++] = real(lhs(i, i));       // real (diagonal)
 
       for(Index k=i+1; k<cols; k++)
-        blockA[count++] = numext::conj(lhs(k, i));     // transposed
+        blockA[count++] = conj(lhs(k, i));     // transposed
     }
   }
 };
@@ -107,12 +107,12 @@ struct symm_pack_rhs
       // transpose
       for(Index k=k2; k<j2; k++)
       {
-        blockB[count+0] = numext::conj(rhs(j2+0,k));
-        blockB[count+1] = numext::conj(rhs(j2+1,k));
+        blockB[count+0] = conj(rhs(j2+0,k));
+        blockB[count+1] = conj(rhs(j2+1,k));
         if (nr==4)
         {
-          blockB[count+2] = numext::conj(rhs(j2+2,k));
-          blockB[count+3] = numext::conj(rhs(j2+3,k));
+          blockB[count+2] = conj(rhs(j2+2,k));
+          blockB[count+3] = conj(rhs(j2+3,k));
         }
         count += nr;
       }
@@ -124,11 +124,11 @@ struct symm_pack_rhs
         for (Index w=0 ; w<h; ++w)
           blockB[count+w] = rhs(k,j2+w);
 
-        blockB[count+h] = numext::real(rhs(k,k));
+        blockB[count+h] = real(rhs(k,k));
 
         // transpose
         for (Index w=h+1 ; w<nr; ++w)
-          blockB[count+w] = numext::conj(rhs(j2+w,k));
+          blockB[count+w] = conj(rhs(j2+w,k));
         count += nr;
         ++h;
       }
@@ -151,12 +151,12 @@ struct symm_pack_rhs
     {
       for(Index k=k2; k<end_k; k++)
       {
-        blockB[count+0] = numext::conj(rhs(j2+0,k));
-        blockB[count+1] = numext::conj(rhs(j2+1,k));
+        blockB[count+0] = conj(rhs(j2+0,k));
+        blockB[count+1] = conj(rhs(j2+1,k));
         if (nr==4)
         {
-          blockB[count+2] = numext::conj(rhs(j2+2,k));
-          blockB[count+3] = numext::conj(rhs(j2+3,k));
+          blockB[count+2] = conj(rhs(j2+2,k));
+          blockB[count+3] = conj(rhs(j2+3,k));
         }
         count += nr;
       }
@@ -169,13 +169,13 @@ struct symm_pack_rhs
       Index half = (std::min)(end_k,j2);
       for(Index k=k2; k<half; k++)
       {
-        blockB[count] = numext::conj(rhs(j2,k));
+        blockB[count] = conj(rhs(j2,k));
         count += 1;
       }
 
       if(half==j2 && half<k2+rows)
       {
-        blockB[count] = numext::real(rhs(j2,j2));
+        blockB[count] = real(rhs(j2,j2));
         count += 1;
       }
       else
@@ -211,7 +211,7 @@ struct product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,LhsSelfAdjoint,Co
     const Scalar* lhs, Index lhsStride,
     const Scalar* rhs, Index rhsStride,
     Scalar* res,       Index resStride,
-    const Scalar& alpha)
+    Scalar alpha)
   {
     product_selfadjoint_matrix<Scalar, Index,
       EIGEN_LOGICAL_XOR(RhsSelfAdjoint,RhsStorageOrder==RowMajor) ? ColMajor : RowMajor,
@@ -234,18 +234,7 @@ struct product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,true,ConjugateLhs
     const Scalar* _lhs, Index lhsStride,
     const Scalar* _rhs, Index rhsStride,
     Scalar* res,        Index resStride,
-    const Scalar& alpha);
-};
-
-template <typename Scalar, typename Index,
-          int LhsStorageOrder, bool ConjugateLhs,
-          int RhsStorageOrder, bool ConjugateRhs>
-EIGEN_DONT_INLINE void product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,true,ConjugateLhs, RhsStorageOrder,false,ConjugateRhs,ColMajor>::run(
-    Index rows, Index cols,
-    const Scalar* _lhs, Index lhsStride,
-    const Scalar* _rhs, Index rhsStride,
-    Scalar* res,        Index resStride,
-    const Scalar& alpha)
+    Scalar alpha)
   {
     Index size = rows;
 
@@ -312,6 +301,7 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,t
       }
     }
   }
+};
 
 // matrix * selfadjoint product
 template <typename Scalar, typename Index,
@@ -325,18 +315,7 @@ struct product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,false,ConjugateLh
     const Scalar* _lhs, Index lhsStride,
     const Scalar* _rhs, Index rhsStride,
     Scalar* res,        Index resStride,
-    const Scalar& alpha);
-};
-
-template <typename Scalar, typename Index,
-          int LhsStorageOrder, bool ConjugateLhs,
-          int RhsStorageOrder, bool ConjugateRhs>
-EIGEN_DONT_INLINE void product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,false,ConjugateLhs, RhsStorageOrder,true,ConjugateRhs,ColMajor>::run(
-    Index rows, Index cols,
-    const Scalar* _lhs, Index lhsStride,
-    const Scalar* _rhs, Index rhsStride,
-    Scalar* res,        Index resStride,
-    const Scalar& alpha)
+    Scalar alpha)
   {
     Index size = cols;
 
@@ -374,6 +353,7 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix<Scalar,Index,LhsStorageOrder,f
       }
     }
   }
+};
 
 } // end namespace internal
 
@@ -403,7 +383,7 @@ struct SelfadjointProductMatrix<Lhs,LhsMode,false,Rhs,RhsMode,false>
     RhsIsSelfAdjoint = (RhsMode&SelfAdjoint)==SelfAdjoint
   };
 
-  template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
+  template<typename Dest> void scaleAndAddTo(Dest& dst, Scalar alpha) const
   {
     eigen_assert(dst.rows()==m_lhs.rows() && dst.cols()==m_rhs.cols());
 

Eigen/src/Core/products/SelfadjointMatrixMatrix_MKL.h

@@ -23,7 +23,7 @@
  ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-//
+
  ********************************************************************************
  *   Content : Eigen bindings to Intel(R) MKL
  *   Self adjoint matrix * matrix product functionality based on ?SYMM/?HEMM.
@@ -47,7 +47,7 @@ template <typename Index, \
 struct product_selfadjoint_matrix<EIGTYPE,Index,LhsStorageOrder,true,ConjugateLhs,RhsStorageOrder,false,ConjugateRhs,ColMajor> \
 {\
 \
-  static void run( \
+  static EIGEN_DONT_INLINE void run( \
     Index rows, Index cols, \
     const EIGTYPE* _lhs, Index lhsStride, \
     const EIGTYPE* _rhs, Index rhsStride, \
@@ -98,7 +98,7 @@ template <typename Index, \
           int RhsStorageOrder, bool ConjugateRhs> \
 struct product_selfadjoint_matrix<EIGTYPE,Index,LhsStorageOrder,true,ConjugateLhs,RhsStorageOrder,false,ConjugateRhs,ColMajor> \
 {\
-  static void run( \
+  static EIGEN_DONT_INLINE void run( \
     Index rows, Index cols, \
     const EIGTYPE* _lhs, Index lhsStride, \
     const EIGTYPE* _rhs, Index rhsStride, \
@@ -174,7 +174,7 @@ template <typename Index, \
 struct product_selfadjoint_matrix<EIGTYPE,Index,LhsStorageOrder,false,ConjugateLhs,RhsStorageOrder,true,ConjugateRhs,ColMajor> \
 {\
 \
-  static void run( \
+  static EIGEN_DONT_INLINE void run( \
     Index rows, Index cols, \
     const EIGTYPE* _lhs, Index lhsStride, \
     const EIGTYPE* _rhs, Index rhsStride, \
@@ -224,7 +224,7 @@ template <typename Index, \
           int RhsStorageOrder, bool ConjugateRhs> \
 struct product_selfadjoint_matrix<EIGTYPE,Index,LhsStorageOrder,false,ConjugateLhs,RhsStorageOrder,true,ConjugateRhs,ColMajor> \
 {\
-  static void run( \
+  static EIGEN_DONT_INLINE void run( \
     Index rows, Index cols, \
     const EIGTYPE* _lhs, Index lhsStride, \
     const EIGTYPE* _rhs, Index rhsStride, \

Eigen/src/Core/products/SelfadjointMatrixVector.h

@@ -28,15 +28,6 @@ struct selfadjoint_matrix_vector_product
 
 {
 static EIGEN_DONT_INLINE void run(
-  Index size,
-  const Scalar*  lhs, Index lhsStride,
-  const Scalar* _rhs, Index rhsIncr,
-  Scalar* res,
-  Scalar alpha);
-};
-
-template<typename Scalar, typename Index, int StorageOrder, int UpLo, bool ConjugateLhs, bool ConjugateRhs, int Version>
-EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrder,UpLo,ConjugateLhs,ConjugateRhs,Version>::run(
   Index size,
   const Scalar*  lhs, Index lhsStride,
   const Scalar* _rhs, Index rhsIncr,
@@ -44,6 +35,7 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
   Scalar alpha)
 {
   typedef typename packet_traits<Scalar>::type Packet;
+  typedef typename NumTraits<Scalar>::Real RealScalar;
   const Index PacketSize = sizeof(Packet)/sizeof(Scalar);
 
   enum {
@@ -59,7 +51,7 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
   conj_helper<Packet,Packet,NumTraits<Scalar>::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs,  IsRowMajor), ConjugateRhs> pcj0;
   conj_helper<Packet,Packet,NumTraits<Scalar>::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, !IsRowMajor), ConjugateRhs> pcj1;
 
-  Scalar cjAlpha = ConjugateRhs ? numext::conj(alpha) : alpha;
+  Scalar cjAlpha = ConjugateRhs ? conj(alpha) : alpha;
 
   // FIXME this copy is now handled outside product_selfadjoint_vector, so it could probably be removed.
   // if the rhs is not sequentially stored in memory we copy it to a temporary buffer,
@@ -79,8 +71,8 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
   for (Index j=FirstTriangular ? bound : 0;
        j<(FirstTriangular ? size : bound);j+=2)
   {
-    const Scalar* EIGEN_RESTRICT A0 = lhs + j*lhsStride;
-    const Scalar* EIGEN_RESTRICT A1 = lhs + (j+1)*lhsStride;
+    register const Scalar* EIGEN_RESTRICT A0 = lhs + j*lhsStride;
+    register const Scalar* EIGEN_RESTRICT A1 = lhs + (j+1)*lhsStride;
 
     Scalar t0 = cjAlpha * rhs[j];
     Packet ptmp0 = pset1<Packet>(t0);
@@ -98,8 +90,8 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
     size_t alignedEnd = alignedStart + ((endi-alignedStart)/(PacketSize))*(PacketSize);
 
     // TODO make sure this product is a real * complex and that the rhs is properly conjugated if needed
-    res[j]   += cjd.pmul(numext::real(A0[j]), t0);
-    res[j+1] += cjd.pmul(numext::real(A1[j+1]), t1);
+    res[j]   += cjd.pmul(internal::real(A0[j]), t0);
+    res[j+1] += cjd.pmul(internal::real(A1[j+1]), t1);
     if(FirstTriangular)
     {
       res[j]   += cj0.pmul(A1[j],   t1);
@@ -114,8 +106,8 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
     for (size_t i=starti; i<alignedStart; ++i)
     {
       res[i] += t0 * A0[i] + t1 * A1[i];
-      t2 += numext::conj(A0[i]) * rhs[i];
-      t3 += numext::conj(A1[i]) * rhs[i];
+      t2 += conj(A0[i]) * rhs[i];
+      t3 += conj(A1[i]) * rhs[i];
     }
     // Yes this an optimization for gcc 4.3 and 4.4 (=> huge speed up)
     // gcc 4.2 does this optimization automatically.
@@ -147,12 +139,12 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
   }
   for (Index j=FirstTriangular ? 0 : bound;j<(FirstTriangular ? bound : size);j++)
   {
-    const Scalar* EIGEN_RESTRICT A0 = lhs + j*lhsStride;
+    register const Scalar* EIGEN_RESTRICT A0 = lhs + j*lhsStride;
 
     Scalar t1 = cjAlpha * rhs[j];
     Scalar t2(0);
     // TODO make sure this product is a real * complex and that the rhs is properly conjugated if needed
-    res[j] += cjd.pmul(numext::real(A0[j]), t1);
+    res[j] += cjd.pmul(internal::real(A0[j]), t1);
     for (Index i=FirstTriangular ? 0 : j+1; i<(FirstTriangular ? j : size); i++)
     {
       res[i] += cj0.pmul(A0[i], t1);
@@ -161,6 +153,7 @@ EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product<Scalar,Index,StorageOrd
     res[j] += alpha * t2;
   }
 }
+};
 
 } // end namespace internal 
 
@@ -187,7 +180,7 @@ struct SelfadjointProductMatrix<Lhs,LhsMode,false,Rhs,0,true>
 
   SelfadjointProductMatrix(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs) {}
 
-  template<typename Dest> void scaleAndAddTo(Dest& dest, const Scalar& alpha) const
+  template<typename Dest> void scaleAndAddTo(Dest& dest, Scalar alpha) const
   {
     typedef typename Dest::Scalar ResScalar;
     typedef typename Base::RhsScalar RhsScalar;
@@ -267,7 +260,7 @@ struct SelfadjointProductMatrix<Lhs,0,true,Rhs,RhsMode,false>
 
   SelfadjointProductMatrix(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs) {}
 
-  template<typename Dest> void scaleAndAddTo(Dest& dest, const Scalar& alpha) const
+  template<typename Dest> void scaleAndAddTo(Dest& dest, Scalar alpha) const
   {
     // let's simply transpose the product
     Transpose<Dest> destT(dest);

Eigen/src/Core/products/SelfadjointMatrixVector_MKL.h

@@ -50,7 +50,7 @@ struct selfadjoint_matrix_vector_product_symv :
 #define EIGEN_MKL_SYMV_SPECIALIZE(Scalar) \
 template<typename Index, int StorageOrder, int UpLo, bool ConjugateLhs, bool ConjugateRhs> \
 struct selfadjoint_matrix_vector_product<Scalar,Index,StorageOrder,UpLo,ConjugateLhs,ConjugateRhs,Specialized> { \
-static void run( \
+static EIGEN_DONT_INLINE void run( \
   Index size, const Scalar*  lhs, Index lhsStride, \
   const Scalar* _rhs, Index rhsIncr, Scalar* res, Scalar alpha) { \
     enum {\
@@ -77,7 +77,7 @@ struct selfadjoint_matrix_vector_product_symv<EIGTYPE,Index,StorageOrder,UpLo,Co
 { \
 typedef Matrix<EIGTYPE,Dynamic,1,ColMajor> SYMVVector;\
 \
-static void run( \
+static EIGEN_DONT_INLINE void run( \
 Index size, const EIGTYPE*  lhs, Index lhsStride, \
 const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* res, EIGTYPE alpha) \
 { \

Eigen/src/Core/products/SelfadjointProduct.h

@@ -18,19 +18,21 @@
 
 namespace Eigen { 
 
+template<typename Scalar, typename Index, int StorageOrder, int UpLo, bool ConjLhs, bool ConjRhs>
+struct selfadjoint_rank1_update;
 
 template<typename Scalar, typename Index, int UpLo, bool ConjLhs, bool ConjRhs>
 struct selfadjoint_rank1_update<Scalar,Index,ColMajor,UpLo,ConjLhs,ConjRhs>
 {
-  static void run(Index size, Scalar* mat, Index stride, const Scalar* vecX, const Scalar* vecY, const Scalar& alpha)
+  static void run(Index size, Scalar* mat, Index stride, const Scalar* vec, Scalar alpha)
   {
     internal::conj_if<ConjRhs> cj;
     typedef Map<const Matrix<Scalar,Dynamic,1> > OtherMap;
-    typedef typename internal::conditional<ConjLhs,typename OtherMap::ConjugateReturnType,const OtherMap&>::type ConjLhsType;
+    typedef typename internal::conditional<ConjLhs,typename OtherMap::ConjugateReturnType,const OtherMap&>::type ConjRhsType;
     for (Index i=0; i<size; ++i)
     {
       Map<Matrix<Scalar,Dynamic,1> >(mat+stride*i+(UpLo==Lower ? i : 0), (UpLo==Lower ? size-i : (i+1)))
-          += (alpha * cj(vecY[i])) * ConjLhsType(OtherMap(vecX+(UpLo==Lower ? i : 0),UpLo==Lower ? size-i : (i+1)));
+          += (alpha * cj(vec[i])) * ConjRhsType(OtherMap(vec+(UpLo==Lower ? i : 0),UpLo==Lower ? size-i : (i+1)));
     }
   }
 };
@@ -38,9 +40,9 @@ struct selfadjoint_rank1_update<Scalar,Index,ColMajor,UpLo,ConjLhs,ConjRhs>
 template<typename Scalar, typename Index, int UpLo, bool ConjLhs, bool ConjRhs>
 struct selfadjoint_rank1_update<Scalar,Index,RowMajor,UpLo,ConjLhs,ConjRhs>
 {
-  static void run(Index size, Scalar* mat, Index stride, const Scalar* vecX, const Scalar* vecY, const Scalar& alpha)
+  static void run(Index size, Scalar* mat, Index stride, const Scalar* vec, Scalar alpha)
   {
-    selfadjoint_rank1_update<Scalar,Index,ColMajor,UpLo==Lower?Upper:Lower,ConjRhs,ConjLhs>::run(size,mat,stride,vecY,vecX,alpha);
+    selfadjoint_rank1_update<Scalar,Index,ColMajor,UpLo==Lower?Upper:Lower,ConjRhs,ConjLhs>::run(size,mat,stride,vec,alpha);
   }
 };
 
@@ -50,7 +52,7 @@ struct selfadjoint_product_selector;
 template<typename MatrixType, typename OtherType, int UpLo>
 struct selfadjoint_product_selector<MatrixType,OtherType,UpLo,true>
 {
-  static void run(MatrixType& mat, const OtherType& other, const typename MatrixType::Scalar& alpha)
+  static void run(MatrixType& mat, const OtherType& other, typename MatrixType::Scalar alpha)
   {
     typedef typename MatrixType::Scalar Scalar;
     typedef typename MatrixType::Index Index;
@@ -76,14 +78,14 @@ struct selfadjoint_product_selector<MatrixType,OtherType,UpLo,true>
     selfadjoint_rank1_update<Scalar,Index,StorageOrder,UpLo,
                               OtherBlasTraits::NeedToConjugate  && NumTraits<Scalar>::IsComplex,
                             (!OtherBlasTraits::NeedToConjugate) && NumTraits<Scalar>::IsComplex>
-          ::run(other.size(), mat.data(), mat.outerStride(), actualOtherPtr, actualOtherPtr, actualAlpha);
+          ::run(other.size(), mat.data(), mat.outerStride(), actualOtherPtr, actualAlpha);
   }
 };
 
 template<typename MatrixType, typename OtherType, int UpLo>
 struct selfadjoint_product_selector<MatrixType,OtherType,UpLo,false>
 {
-  static void run(MatrixType& mat, const OtherType& other, const typename MatrixType::Scalar& alpha)
+  static void run(MatrixType& mat, const OtherType& other, typename MatrixType::Scalar alpha)
   {
     typedef typename MatrixType::Scalar Scalar;
     typedef typename MatrixType::Index Index;
@@ -111,7 +113,7 @@ struct selfadjoint_product_selector<MatrixType,OtherType,UpLo,false>
 template<typename MatrixType, unsigned int UpLo>
 template<typename DerivedU>
 SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
-::rankUpdate(const MatrixBase<DerivedU>& u, const Scalar& alpha)
+::rankUpdate(const MatrixBase<DerivedU>& u, Scalar alpha)
 {
   selfadjoint_product_selector<MatrixType,DerivedU,UpLo>::run(_expression().const_cast_derived(), u.derived(), alpha);
 

Eigen/src/Core/products/SelfadjointRank2Update.h

@@ -24,14 +24,14 @@ struct selfadjoint_rank2_update_selector;
 template<typename Scalar, typename Index, typename UType, typename VType>
 struct selfadjoint_rank2_update_selector<Scalar,Index,UType,VType,Lower>
 {
-  static void run(Scalar* mat, Index stride, const UType& u, const VType& v, const Scalar& alpha)
+  static void run(Scalar* mat, Index stride, const UType& u, const VType& v, Scalar alpha)
   {
     const Index size = u.size();
     for (Index i=0; i<size; ++i)
     {
       Map<Matrix<Scalar,Dynamic,1> >(mat+stride*i+i, size-i) +=
-                        (numext::conj(alpha) * numext::conj(u.coeff(i))) * v.tail(size-i)
-                      + (alpha * numext::conj(v.coeff(i))) * u.tail(size-i);
+                        (conj(alpha)  * conj(u.coeff(i))) * v.tail(size-i)
+                      + (alpha * conj(v.coeff(i))) * u.tail(size-i);
     }
   }
 };
@@ -39,13 +39,13 @@ struct selfadjoint_rank2_update_selector<Scalar,Index,UType,VType,Lower>
 template<typename Scalar, typename Index, typename UType, typename VType>
 struct selfadjoint_rank2_update_selector<Scalar,Index,UType,VType,Upper>
 {
-  static void run(Scalar* mat, Index stride, const UType& u, const VType& v, const Scalar& alpha)
+  static void run(Scalar* mat, Index stride, const UType& u, const VType& v, Scalar alpha)
   {
     const Index size = u.size();
     for (Index i=0; i<size; ++i)
       Map<Matrix<Scalar,Dynamic,1> >(mat+stride*i, i+1) +=
-                        (numext::conj(alpha)  * numext::conj(u.coeff(i))) * v.head(i+1)
-                      + (alpha * numext::conj(v.coeff(i))) * u.head(i+1);
+                        (conj(alpha)  * conj(u.coeff(i))) * v.head(i+1)
+                      + (alpha * conj(v.coeff(i))) * u.head(i+1);
   }
 };
 
@@ -58,7 +58,7 @@ template<bool Cond, typename T> struct conj_expr_if
 template<typename MatrixType, unsigned int UpLo>
 template<typename DerivedU, typename DerivedV>
 SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
-::rankUpdate(const MatrixBase<DerivedU>& u, const MatrixBase<DerivedV>& v, const Scalar& alpha)
+::rankUpdate(const MatrixBase<DerivedU>& u, const MatrixBase<DerivedV>& v, Scalar alpha)
 {
   typedef internal::blas_traits<DerivedU> UBlasTraits;
   typedef typename UBlasTraits::DirectLinearAccessType ActualUType;
@@ -75,9 +75,9 @@ SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
 
   enum { IsRowMajor = (internal::traits<MatrixType>::Flags&RowMajorBit) ? 1 : 0 };
   Scalar actualAlpha = alpha * UBlasTraits::extractScalarFactor(u.derived())
-                             * numext::conj(VBlasTraits::extractScalarFactor(v.derived()));
+                             * internal::conj(VBlasTraits::extractScalarFactor(v.derived()));
   if (IsRowMajor)
-    actualAlpha = numext::conj(actualAlpha);
+    actualAlpha = internal::conj(actualAlpha);
 
   internal::selfadjoint_rank2_update_selector<Scalar, Index,
     typename internal::remove_all<typename internal::conj_expr_if<IsRowMajor ^ UBlasTraits::NeedToConjugate,_ActualUType>::type>::type,

Eigen/src/Core/products/TriangularMatrixMatrix.h

@@ -61,7 +61,7 @@ struct product_triangular_matrix_matrix<Scalar,Index,Mode,LhsIsTriangular,
     const Scalar* lhs, Index lhsStride,
     const Scalar* rhs, Index rhsStride,
     Scalar* res,       Index resStride,
-    const Scalar& alpha, level3_blocking<Scalar,Scalar>& blocking)
+    Scalar alpha, level3_blocking<Scalar,Scalar>& blocking)
   {
     product_triangular_matrix_matrix<Scalar, Index,
       (Mode&(UnitDiag|ZeroDiag)) | ((Mode&Upper) ? Lower : Upper),
@@ -96,20 +96,7 @@ struct product_triangular_matrix_matrix<Scalar,Index,Mode,true,
     const Scalar* _lhs, Index lhsStride,
     const Scalar* _rhs, Index rhsStride,
     Scalar* res,        Index resStride,
-    const Scalar& alpha, level3_blocking<Scalar,Scalar>& blocking);
-};
-
-template <typename Scalar, typename Index, int Mode,
-          int LhsStorageOrder, bool ConjugateLhs,
-          int RhsStorageOrder, bool ConjugateRhs, int Version>
-EIGEN_DONT_INLINE void product_triangular_matrix_matrix<Scalar,Index,Mode,true,
-                                                        LhsStorageOrder,ConjugateLhs,
-                                                        RhsStorageOrder,ConjugateRhs,ColMajor,Version>::run(
-    Index _rows, Index _cols, Index _depth,
-    const Scalar* _lhs, Index lhsStride,
-    const Scalar* _rhs, Index rhsStride,
-    Scalar* res,        Index resStride,
-    const Scalar& alpha, level3_blocking<Scalar,Scalar>& blocking)
+    Scalar alpha, level3_blocking<Scalar,Scalar>& blocking)
   {
     // strip zeros
     Index diagSize  = (std::min)(_rows,_depth);
@@ -216,14 +203,15 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix<Scalar,Index,Mode,true,
       }
     }
   }
+};
 
 // implements col-major += alpha * op(general) * op(triangular)
 template <typename Scalar, typename Index, int Mode,
           int LhsStorageOrder, bool ConjugateLhs,
           int RhsStorageOrder, bool ConjugateRhs, int Version>
 struct product_triangular_matrix_matrix<Scalar,Index,Mode,false,
-                                        LhsStorageOrder,ConjugateLhs,
-                                        RhsStorageOrder,ConjugateRhs,ColMajor,Version>
+                                           LhsStorageOrder,ConjugateLhs,
+                                           RhsStorageOrder,ConjugateRhs,ColMajor,Version>
 {
   typedef gebp_traits<Scalar,Scalar> Traits;
   enum {
@@ -237,20 +225,7 @@ struct product_triangular_matrix_matrix<Scalar,Index,Mode,false,
     const Scalar* _lhs, Index lhsStride,
     const Scalar* _rhs, Index rhsStride,
     Scalar* res,        Index resStride,
-    const Scalar& alpha, level3_blocking<Scalar,Scalar>& blocking);
-};
-
-template <typename Scalar, typename Index, int Mode,
-          int LhsStorageOrder, bool ConjugateLhs,
-          int RhsStorageOrder, bool ConjugateRhs, int Version>
-EIGEN_DONT_INLINE void product_triangular_matrix_matrix<Scalar,Index,Mode,false,
-                                                        LhsStorageOrder,ConjugateLhs,
-                                                        RhsStorageOrder,ConjugateRhs,ColMajor,Version>::run(
-    Index _rows, Index _cols, Index _depth,
-    const Scalar* _lhs, Index lhsStride,
-    const Scalar* _rhs, Index rhsStride,
-    Scalar* res,        Index resStride,
-    const Scalar& alpha, level3_blocking<Scalar,Scalar>& blocking)
+    Scalar alpha, level3_blocking<Scalar,Scalar>& blocking)
   {
     // strip zeros
     Index diagSize  = (std::min)(_cols,_depth);
@@ -368,6 +343,7 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix<Scalar,Index,Mode,false,
       }
     }
   }
+};
 
 /***************************************************************************
 * Wrapper to product_triangular_matrix_matrix
@@ -388,7 +364,7 @@ struct TriangularProduct<Mode,LhsIsTriangular,Lhs,false,Rhs,false>
 
   TriangularProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs) {}
 
-  template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
+  template<typename Dest> void scaleAndAddTo(Dest& dst, Scalar alpha) const
   {
     typename internal::add_const_on_value_type<ActualLhsType>::type lhs = LhsBlasTraits::extract(m_lhs);
     typename internal::add_const_on_value_type<ActualRhsType>::type rhs = RhsBlasTraits::extract(m_rhs);

Eigen/src/Core/products/TriangularMatrixMatrix_MKL.h

@@ -91,7 +91,7 @@ struct product_triangular_matrix_matrix_trmm<EIGTYPE,Index,Mode,true, \
     conjA = ((LhsStorageOrder==ColMajor) && ConjugateLhs) ? 1 : 0 \
   }; \
 \
-  static void run( \
+  static EIGEN_DONT_INLINE void run( \
     Index _rows, Index _cols, Index _depth, \
     const EIGTYPE* _lhs, Index lhsStride, \
     const EIGTYPE* _rhs, Index rhsStride, \
@@ -109,7 +109,7 @@ struct product_triangular_matrix_matrix_trmm<EIGTYPE,Index,Mode,true, \
 /* Non-square case - doesn't fit to MKL ?TRMM. Fall to default triangular product or call MKL ?GEMM*/ \
    if (rows != depth) { \
 \
-     int nthr = mkl_domain_get_max_threads(EIGEN_MKL_DOMAIN_BLAS); \
+     int nthr = mkl_domain_get_max_threads(MKL_BLAS); \
 \
      if (((nthr==1) && (((std::max)(rows,depth)-diagSize)/(double)diagSize < 0.5))) { \
      /* Most likely no benefit to call TRMM or GEMM from MKL*/ \
@@ -205,7 +205,7 @@ struct product_triangular_matrix_matrix_trmm<EIGTYPE,Index,Mode,false, \
     conjA = ((RhsStorageOrder==ColMajor) && ConjugateRhs) ? 1 : 0 \
   }; \
 \
-  static void run( \
+  static EIGEN_DONT_INLINE void run( \
     Index _rows, Index _cols, Index _depth, \
     const EIGTYPE* _lhs, Index lhsStride, \
     const EIGTYPE* _rhs, Index rhsStride, \
@@ -223,7 +223,7 @@ struct product_triangular_matrix_matrix_trmm<EIGTYPE,Index,Mode,false, \
 /* Non-square case - doesn't fit to MKL ?TRMM. Fall to default triangular product or call MKL ?GEMM*/ \
    if (cols != depth) { \
 \
-     int nthr = mkl_domain_get_max_threads(EIGEN_MKL_DOMAIN_BLAS); \
+     int nthr = mkl_domain_get_max_threads(MKL_BLAS); \
 \
      if ((nthr==1) && (((std::max)(cols,depth)-diagSize)/(double)diagSize < 0.5)) { \
      /* Most likely no benefit to call TRMM or GEMM from MKL*/ \

Eigen/src/Core/products/TriangularMatrixVector.h

@@ -27,13 +27,7 @@ struct triangular_matrix_vector_product<Index,Mode,LhsScalar,ConjLhs,RhsScalar,C
     HasZeroDiag = (Mode & ZeroDiag)==ZeroDiag
   };
   static EIGEN_DONT_INLINE  void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride,
-                                     const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha);
-};
-
-template<typename Index, int Mode, typename LhsScalar, bool ConjLhs, typename RhsScalar, bool ConjRhs, int Version>
-EIGEN_DONT_INLINE void triangular_matrix_vector_product<Index,Mode,LhsScalar,ConjLhs,RhsScalar,ConjRhs,ColMajor,Version>
-  ::run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride,
-        const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha)
+                                     const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, ResScalar alpha)
   {
     static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH;
     Index size = (std::min)(_rows,_cols);
@@ -84,6 +78,7 @@ EIGEN_DONT_INLINE void triangular_matrix_vector_product<Index,Mode,LhsScalar,Con
           _res, resIncr, alpha);
     }
   }
+};
 
 template<typename Index, int Mode, typename LhsScalar, bool ConjLhs, typename RhsScalar, bool ConjRhs,int Version>
 struct triangular_matrix_vector_product<Index,Mode,LhsScalar,ConjLhs,RhsScalar,ConjRhs,RowMajor,Version>
@@ -94,14 +89,8 @@ struct triangular_matrix_vector_product<Index,Mode,LhsScalar,ConjLhs,RhsScalar,C
     HasUnitDiag = (Mode & UnitDiag)==UnitDiag,
     HasZeroDiag = (Mode & ZeroDiag)==ZeroDiag
   };
-  static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride,
-                                    const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha);
-};
-
-template<typename Index, int Mode, typename LhsScalar, bool ConjLhs, typename RhsScalar, bool ConjRhs,int Version>
-EIGEN_DONT_INLINE void triangular_matrix_vector_product<Index,Mode,LhsScalar,ConjLhs,RhsScalar,ConjRhs,RowMajor,Version>
-  ::run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride,
-        const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha)
+  static void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride,
+                  const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, ResScalar alpha)
   {
     static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH;
     Index diagSize = (std::min)(_rows,_cols);
@@ -152,6 +141,7 @@ EIGEN_DONT_INLINE void triangular_matrix_vector_product<Index,Mode,LhsScalar,Con
             &res.coeffRef(diagSize), resIncr, alpha);
     }
   }
+};
 
 /***************************************************************************
 * Wrapper to product_triangular_vector
@@ -181,7 +171,7 @@ struct TriangularProduct<Mode,true,Lhs,false,Rhs,true>
 
   TriangularProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs) {}
 
-  template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
+  template<typename Dest> void scaleAndAddTo(Dest& dst, Scalar alpha) const
   {
     eigen_assert(dst.rows()==m_lhs.rows() && dst.cols()==m_rhs.cols());
   
@@ -197,7 +187,7 @@ struct TriangularProduct<Mode,false,Lhs,true,Rhs,false>
 
   TriangularProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs) {}
 
-  template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
+  template<typename Dest> void scaleAndAddTo(Dest& dst, Scalar alpha) const
   {
     eigen_assert(dst.rows()==m_lhs.rows() && dst.cols()==m_rhs.cols());
 
@@ -215,7 +205,7 @@ namespace internal {
 template<> struct trmv_selector<ColMajor>
 {
   template<int Mode, typename Lhs, typename Rhs, typename Dest>
-  static void run(const TriangularProduct<Mode,true,Lhs,false,Rhs,true>& prod, Dest& dest, const typename TriangularProduct<Mode,true,Lhs,false,Rhs,true>::Scalar& alpha)
+  static void run(const TriangularProduct<Mode,true,Lhs,false,Rhs,true>& prod, Dest& dest, typename TriangularProduct<Mode,true,Lhs,false,Rhs,true>::Scalar alpha)
   {
     typedef TriangularProduct<Mode,true,Lhs,false,Rhs,true> ProductType;
     typedef typename ProductType::Index Index;
@@ -245,7 +235,7 @@ template<> struct trmv_selector<ColMajor>
 
     gemv_static_vector_if<ResScalar,Dest::SizeAtCompileTime,Dest::MaxSizeAtCompileTime,MightCannotUseDest> static_dest;
 
-    bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0));
+    bool alphaIsCompatible = (!ComplexByReal) || (imag(actualAlpha)==RealScalar(0));
     bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible;
     
     RhsScalar compatibleAlpha = get_factor<ResScalar,RhsScalar>::run(actualAlpha);
@@ -256,7 +246,7 @@ template<> struct trmv_selector<ColMajor>
     if(!evalToDest)
     {
       #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN
-      Index size = dest.size();
+      int size = dest.size();
       EIGEN_DENSE_STORAGE_CTOR_PLUGIN
       #endif
       if(!alphaIsCompatible)
@@ -291,7 +281,7 @@ template<> struct trmv_selector<ColMajor>
 template<> struct trmv_selector<RowMajor>
 {
   template<int Mode, typename Lhs, typename Rhs, typename Dest>
-  static void run(const TriangularProduct<Mode,true,Lhs,false,Rhs,true>& prod, Dest& dest, const typename TriangularProduct<Mode,true,Lhs,false,Rhs,true>::Scalar& alpha)
+  static void run(const TriangularProduct<Mode,true,Lhs,false,Rhs,true>& prod, Dest& dest, typename TriangularProduct<Mode,true,Lhs,false,Rhs,true>::Scalar alpha)
   {
     typedef TriangularProduct<Mode,true,Lhs,false,Rhs,true> ProductType;
     typedef typename ProductType::LhsScalar LhsScalar;

Eigen/src/Core/products/TriangularMatrixVector_MKL.h

@@ -50,7 +50,7 @@ struct triangular_matrix_vector_product_trmv :
 #define EIGEN_MKL_TRMV_SPECIALIZE(Scalar) \
 template<typename Index, int Mode, bool ConjLhs, bool ConjRhs> \
 struct triangular_matrix_vector_product<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,ColMajor,Specialized> { \
- static void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \
+ static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \
                                      const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \
       triangular_matrix_vector_product_trmv<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,ColMajor>::run( \
         _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \
@@ -58,7 +58,7 @@ struct triangular_matrix_vector_product<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs
 }; \
 template<typename Index, int Mode, bool ConjLhs, bool ConjRhs> \
 struct triangular_matrix_vector_product<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,RowMajor,Specialized> { \
- static void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \
+ static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \
                                      const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \
       triangular_matrix_vector_product_trmv<Index,Mode,Scalar,ConjLhs,Scalar,ConjRhs,RowMajor>::run( \
         _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \
@@ -81,8 +81,8 @@ struct triangular_matrix_vector_product_trmv<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,
     IsZeroDiag  = (Mode&ZeroDiag) ? 1 : 0, \
     LowUp = IsLower ? Lower : Upper \
   }; \
- static void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \
-                 const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \
+ static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \
+                             const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \
  { \
    if (ConjLhs || IsZeroDiag) { \
      triangular_matrix_vector_product<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,ConjRhs,ColMajor,BuiltIn>::run( \
@@ -166,8 +166,8 @@ struct triangular_matrix_vector_product_trmv<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,
     IsZeroDiag  = (Mode&ZeroDiag) ? 1 : 0, \
     LowUp = IsLower ? Lower : Upper \
   }; \
- static void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \
-                 const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \
+ static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \
+                             const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \
  { \
    if (IsZeroDiag) { \
      triangular_matrix_vector_product<Index,Mode,EIGTYPE,ConjLhs,EIGTYPE,ConjRhs,RowMajor,BuiltIn>::run( \

Eigen/src/Core/products/TriangularSolverMatrix.h

@@ -18,7 +18,7 @@ namespace internal {
 template <typename Scalar, typename Index, int Side, int Mode, bool Conjugate, int TriStorageOrder>
 struct triangular_solve_matrix<Scalar,Index,Side,Mode,Conjugate,TriStorageOrder,RowMajor>
 {
-  static void run(
+  static EIGEN_DONT_INLINE void run(
     Index size, Index cols,
     const Scalar*  tri, Index triStride,
     Scalar* _other, Index otherStride,
@@ -39,13 +39,6 @@ template <typename Scalar, typename Index, int Mode, bool Conjugate, int TriStor
 struct triangular_solve_matrix<Scalar,Index,OnTheLeft,Mode,Conjugate,TriStorageOrder,ColMajor>
 {
   static EIGEN_DONT_INLINE void run(
-    Index size, Index otherSize,
-    const Scalar* _tri, Index triStride,
-    Scalar* _other, Index otherStride,
-    level3_blocking<Scalar,Scalar>& blocking);
-};
-template <typename Scalar, typename Index, int Mode, bool Conjugate, int TriStorageOrder>
-EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheLeft,Mode,Conjugate,TriStorageOrder,ColMajor>::run(
     Index size, Index otherSize,
     const Scalar* _tri, Index triStride,
     Scalar* _other, Index otherStride,
@@ -81,7 +74,7 @@ EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheLeft,Mode,Conju
     // coherence when accessing the rhs elements
     std::ptrdiff_t l1, l2;
     manage_caching_sizes(GetAction, &l1, &l2);
-    Index subcols = cols>0 ? l2/(4 * sizeof(Scalar) *  std::max<Index>(otherStride,size)) : 0;
+    Index subcols = cols>0 ? l2/(4 * sizeof(Scalar) * otherStride) : 0;
     subcols = std::max<Index>((subcols/Traits::nr)*Traits::nr, Traits::nr);
 
     for(Index k2=IsLower ? 0 : size;
@@ -115,9 +108,8 @@ EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheLeft,Mode,Conju
           {
             // TODO write a small kernel handling this (can be shared with trsv)
             Index i  = IsLower ? k2+k1+k : k2-k1-k-1;
+            Index s  = IsLower ? k2+k1 : i+1;
             Index rs = actualPanelWidth - k - 1; // remaining size
-            Index s  = TriStorageOrder==RowMajor ? (IsLower ? k2+k1 : i+1)
-                                                 :  IsLower ? i+1 : i-rs;
 
             Scalar a = (Mode & UnitDiag) ? Scalar(1) : Scalar(1)/conj(tri(i,i));
             for (Index j=j2; j<j2+actual_cols; ++j)
@@ -134,6 +126,7 @@ EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheLeft,Mode,Conju
               }
               else
               {
+                Index s = IsLower ? i+1 : i-rs;
                 Scalar b = (other(i,j) *= a);
                 Scalar* r = &other(s,j);
                 const Scalar* l = &tri(s,i);
@@ -180,6 +173,7 @@ EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheLeft,Mode,Conju
       }
     }
   }
+};
 
 /* Optimized triangular solver with multiple left hand sides and the trinagular matrix on the right
  */
@@ -187,13 +181,6 @@ template <typename Scalar, typename Index, int Mode, bool Conjugate, int TriStor
 struct triangular_solve_matrix<Scalar,Index,OnTheRight,Mode,Conjugate,TriStorageOrder,ColMajor>
 {
   static EIGEN_DONT_INLINE void run(
-    Index size, Index otherSize,
-    const Scalar* _tri, Index triStride,
-    Scalar* _other, Index otherStride,
-    level3_blocking<Scalar,Scalar>& blocking);
-};
-template <typename Scalar, typename Index, int Mode, bool Conjugate, int TriStorageOrder>
-EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheRight,Mode,Conjugate,TriStorageOrder,ColMajor>::run(
     Index size, Index otherSize,
     const Scalar* _tri, Index triStride,
     Scalar* _other, Index otherStride,
@@ -302,12 +289,9 @@ EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheRight,Mode,Conj
                 for (Index i=0; i<actual_mc; ++i)
                   r[i] -= a[i] * b;
               }
-              if((Mode & UnitDiag)==0)
-              {
-                Scalar b = conj(rhs(j,j));
-                for (Index i=0; i<actual_mc; ++i)
-                  r[i] /= b;
-              }
+              Scalar b = (Mode & UnitDiag) ? Scalar(1) : Scalar(1)/conj(rhs(j,j));
+              for (Index i=0; i<actual_mc; ++i)
+                r[i] *= b;
             }
 
             // pack the just computed part of lhs to A
@@ -324,6 +308,7 @@ EIGEN_DONT_INLINE void triangular_solve_matrix<Scalar,Index,OnTheRight,Mode,Conj
       }
     }
   }
+};
 
 } // end namespace internal