changelog

LAPACK++ (Linear Algebra PACKage in C++) is a software library for numerical linear algebra that sol

	* configure.ac: Release lapackpp-2.4.4

	* include/genmd.h, src/genmc.cc: Move the inlined function
	definitions into normal, non-inlined functions.

	* include/*.h: Add DLLIMPORT for win32 compatibility.

	* include/gmd.h, gmc.h, gmi.h, gmf.h, gmli.h: Add missing
	declaration for operator<<(). Add alias typedef la::mat et al. for
	easy itpp compatibility.

	* configure.ac, lapackpp.iss.in: Add option for including the
	compiled MSVC file into the self-installing exe. The resulting exe
	may usually not be distributed!

2005-12-31  Christian Stimming  <[e-mail]>

	* include/latmpl.h: Add more template functions for matrix
	classes.

2005-12-29  Christian Stimming  <[e-mail]>

	* include/latmpl.h: Add public template functions for common
	matrix constructions.

	* matrix/src/gmconv.cc: Compile complex converters only when
	LA_COMPLEX_SUPPORT is enabled.

	* include/gmd.h, gmc.h, gmi.h, gmf.h, gmli.h: Add has_unitstride()
	and is_submatrixview() predicates to matrix classes.

2005-12-28  Christian Stimming  <[e-mail]>

	* include/lavi.h, lavli.h, matrix/src/mtmpl.h: Add constructor for
	index lists that calculate the list of indices as given by one
	LaIndex() object.

	* include/gmd.h, gmc.h, gmi.h, gmf.h, gmli.h,
	matrix/src/gmconv.cc: Add conversion functions into all directions
	between the five GenMat types.

2005-12-27  Christian Stimming  <[e-mail]>

	* include/gmc.h: Add constructor and copy() method for creation of
	complex matrix from separate real and imag part as real matrices.

	* include/laslv.h, include/lapackc.h, src/eigslv.cc: Add
	calculation of complex matrix eigenvalues by zgeev.

	* src/eigslv.cc: Fix missing check of eigvec matrix size.

	* include/laslv.h: Add documentation.

2005-12-23   Christian Stimming <[e-mail]>

	* blaspp/src/*.cc, matrix/src/*.cc, src/*.cc: Make sure files can
	be compiled even if LA_COMPLEX_SUPPORT is switched off. Makes
	initial porting to MSVC much easier.

	* lapackpp.vcproj: Add project file for Microsoft Visual Studio
	Compiler (MSVC).

2005-12-21   Christian Stimming <[e-mail]>

	* include/gmd.h, gm*.h: Add rows() and cols() member for
	convenience compatibility with IT++.

2005-12-15  Christian Stimming  <[e-mail]>

	* configure.ac: Release lapackpp-2.4.3

2005-12-02   Christian Stimming <[e-mail]>

	* blaspp/src/blas3pp.cc: Fix matrix norms which might have been
	wrong with submatrices.

2005-11-25   Christian Stimming <[e-mail]>

	* matrix/src/gmd.cc: Slightly optimize operator(LaIndex,LaIndex).

	* matrix/src/mtmpl.h (assign): Heavily optimize scalar assignment
	for submatrix view matrices.

2005-11-24   Christian Stimming <[e-mail]>

	* src/Makefile.am: Link main library already against dependencies.

2005-11-02   Christian Stimming <[e-mail]>

	* configure.ac: Release lapackpp-2.4.2

2005-10-28   Christian Stimming <[e-mail]>

	* include/gmc.h, matrix/src/mtmpl.h, gmc.cc etc: All LaGenMatXY
	classes have been greatly improved: Use common function templates
	for the actual member functions. This makes the optimized methods
	available in all classes at once.

	* include/vc.h, matrix/src/vtmpl.h, vc.cc etc: All VectorXY
	classes have been greatly improved: Use common function templates
	for the actual member functions; enforce constructor use of
	management structure. Unifies all the vector classes.

	* matrix/src/vc.cc, vd.cc, vf.cc, vi.cc, vli.cc: Implement
	additional reference-counting for the vector's management
	structure. Fixes a memory leak as described in vc.cc.

2005-10-26   Christian Stimming <[e-mail]>

	* include/lacomplex: Fix gcc4 and <complex> header file.

	* matrix/src/gmc.cc, gmd.cc, gmf.cc, gmi.cc, gmli.cc: Make
	optional error message a fatal one. Will probably break some
	applications, but that code never worked correctly anyway.

	* configure.ac: Release lapackpp-2.4.1

	* matrix/src/gmc.cc, gmd.cc, gmf.cc, gmi.cc, gmli.cc: Added extra
	check for submatrix in resize(). This will point out errorneous
	usages of operator=() which has copy() semantics, where in reality
	inject() should have been used. This will probably show up in much
	existing code, but that code is broken and never worked correctly
	to begin with. 

	Nevertheless: If you compile lapackpp with --enable-debug=yes,
	then this error will immediately throw an exception. If you
	compile with --enable-debug=no, then this error will be ignored,
	although wrong results will still be calculated. If you call
	LaException::enablePrint(true) in your application, then even with
	--enable-debug=no you will get an error message on stderr when
	this check fails and your calculations are probably wrong. In a
	future release an exception will be thrown always.

	* matrix/src/gmc.cc, gmd.cc: Fixed optimized inject() and
	operator=(scalar) for submatrices. Used to be broken, but is fixed
	now.

	* include/v*.h: Introduce vref_ref_count as an extra reference
	counting for the management data structure in order to fix the
	memory leak as described in vd.cc. Unfinished so far.
	
	* matrix/src/vd.cc: The size checking of VectorDouble::inject did
	print an error message but didn't abort the operation. Changed, so
	that it will throw an LaException now.

	* matrix/src/*.cc: Make fatal error messages really fatal.
	
	* blaspp/src/blas3pp.cc (Blas_NormF): Check for submatrix view and
	non-unit strides. If anything besides the simple case occurs,
	calculate the result by hand.

2005-10-25   Christian Stimming <[e-mail]>

	* matrix/src/gmd.cc, gmc.cc (inject): Use optimized inject()
	copying really only when both LaIndex's are identical.

	* include/laindex.h: Add operator== for equality testing.

2005-10-19   Christian Stimming <[e-mail]>

	* include/gmd.h: Don't cast an address to an unsigned int; it
	breaks on i86_64 platforms.

2005-10-06   Christian Stimming <[e-mail]>

	* include/lacomplex.h.in: Make USE_GCC3 conditional on __GNUC__
	for compatibility with other compilers, as pointed out by Vincent
	Acary <[e-mail]>.

	* src/lasvd.cc: Remove definition of min() because std::min is
	used everywhere.

2005-09-16   Christian Stimming <[e-mail]>

	* matrix/src/vc.cc, vd.cc, vf.cc, vi.cc, vli.cc: Add sanity checks
	for all arguments. Also check for successful memory allocation.

	* matrix/src/laindex.cc: Replace assert()s by LaExceptions.

2005-09-12   Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.4.0

2005-09-12   Christian Stimming <[e-mail]>

	* include/laindex.h, matrix/source/laindex.cc: Make the
	return-by-reference functions start(), end() and incr() protected
	because they are too error-prone when being manipulated
	arbitrarily. Instead, add a set() method that follows the
	conventions of the constructors. Additionally, add assertions so
	that the increment is nonzero.

	* lapackpp.pc.in: Add configuration file for pkg-config, let this
	also be installed.

	* matrix/src/gnc.cc, gmd.cc (operator=): Fix errorneous assignment
	when matrix has non-unit stride. Reported by Brian White
	<[e-mail]>.

	* include/laindex.h: Fix error in LaIndex::operator=; increment
	was not copied. Reported by Brian White <[e-mail]>.

2005-08-05   Christian Stimming <[e-mail]>

	* configure.in: Release version 2.3.0

	* include/blas1.h, blaspp/src/blas1pp.cc: Fix declaration of zdotu
	and zdotc functions on Windows, because the return value of LAPACK
	is there actually returned on the stack -- contrary to LAPACK on
	linux.

	* include/bmd.h, matrix/src/bmd.cc, include/blas1pp.h: Fix
	LaBandMatDouble methods as reported by Edward Baudrez
	<[e-mail]>

2005-07-20   Christian Stimming <[e-mail]>

	* macros/acx_lapackpp.m4: Copy required macro ACX_SEARCH_FOR_PATH
	into this file.

2005-04-26   Christian Stimming <[e-mail]>

	* macros/acx_lapackpp.m4: Fix library lookup on win32 because it
	was still looked for liblapackpp32.dll instead of lapackpp32.dll.

2005-04-19  Jacob (Jack) Gryn <jgryn at cs dot yorku dot ca>

	* include/laslv.h, src/linslv.cc: Add work size checker to
	LaLUInverseIP

2005-04-18  Jacob (Jack) Gryn <jgryn at cs dot yorku dot ca>
        * include/gm*.h, matrix/src/gm*.cc
          Add operator+=(scalar) functions

2005-04-04  Jacob (Jack) Gryn <jgryn at cs dot yorku dot ca>
        * src/linslv.cc src/eigslv.cc include/laslv.h:
          Add LaLUInverseIP (to get inverse from LU factorization)
          Add LaEigSlv for General Matrices
        * include/lapackd.h : add lapack functions dgeev and dgetri

2005-03-22    Christian Stimming <[e-mail]>

	* configure.in: Release version 2.2.0
	
2005-03-18    Christian Stimming <[e-mail]>

	* configure.in: Changed default of --enable-old-librarynames to
	"no". Increased SO_VERSION and SO_AGE of library.

	* include/laprefs.h: Improve documentation. 

	* include/gmd.h: Copy constructor from LaGenMatFloat made explicit
	in order to avoid unintentional conversions.

2005-03-17  Jacob (Jack) Gryn <jgryn at cs dot yorku dot ca>

	* include/laprefs.h: Put enum in LaPreferences class (no longer
	global)
        * src/{gmd.cc, gmf.cc}: call the same display function for
	operator<<.  Create new copy constructor from LaGenMatFloat to
	LaGenMatDouble
        * include/gmd.h: Defs for new copy constructor
        
2005-03-16  Jacob (Jack) Gryn <jgryn at cs dot yorku dot ca>

        Added support for {NORMAL, MATLAB, MAPLE} display output
        Added two files
        * include/laprefs.h: New LaPreferences class defs
        * src/laprefs.cc: New LaPreferences declarations
        Modified 
        * include/Makefile.am: for laprefs.h
        * src/Makefile.am: for laprefs.cc
        * matrix/src/{gmc.cc, gmd.cc, gmf.cc} : now reads LaPreferences
        class to determine output display type for operator<<.  

2005-03-11  Christian Stimming  <[e-mail]>

	* include/gfqrc.h: Rename _A member variable into _matA because
	for whatever reason Mac OS X barfed on this variable name.

2005-03-04   Christian Stimming <[e-mail]>

	* include/lavc.h (inc), lavd.h, lavi.h, lavli.h: The inc() method
	gave wrong results when the vector was in fact a submatrix view of
	a larger matrix. Fixed this.

2005-03-03   Christian Stimming <[e-mail]>

	* matrix/src/gmc.cc, gmd.cc: Simplify inject to use the Vector's
	function. Speeds up this operation quite a lot.

2005-02-04   Christian Stimming <[e-mail]>

	* src/Makefile.am: Change name of windows DLL to lapackpp32.dll
	without the leading lib.

	* configure.in: Release version 2.1.2

	* src/lasvd.cc, eigslv.cc, linslv.cc: Initialize info variable
	with zero. Fixes bug#1092300 of the sourceforge bug tracker.

	* include/genmd.h: Make all functions inline that are defined in
	the header.

2005-01-21  Christian Stimming <[e-mail]>

	* configure.in: Release version 2.1.1

	* include/blas1pp.h: Fix superfluous semicolons as complained
	about by gcc3.4