changelog

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

2007-07-14  Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.5.2.

	* include/gmd.h: Add GenMat::copy() method. Add documentation of
	"named return value optimization" issues when expecting a
	deep-copy out of the copy constructor. SF bug item #1729023.

2007-07-01  Christian Stimming  <[e-mail]>

	* matrix/src/laindex.cc: Fix broken LaIndex::set method. SF bug
	item #1745670.

2007-01-30  Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.5.1.

2007-01-22  Christian Stimming  <[e-mail]>

	* configure.ac, include/blas1.h: The functions zdotc() and zdotu()
	return their result on the stack even on Linux (opensuse10.2's
	version of BLAS), not only on Windows. Added a configure-time
	check for this.

2007-01-21  Christian Stimming  <[e-mail]>

	* include/blas3pp.h: Fix ambiguity in overloaded Mat_Mat_Mult
	function: When using the version with extra bool arguments,
	require at least the first bool to be specified.

2007-01-11  Christian Stimming <[e-mail]>

	* include/lavd.h: Fix operator()(LaIndex) that returns a new copy
	on MSVC compilers but not on gcc.

	* matrix/testing/tvd.cc: Add extensive test suite for
	LaVectorDouble after reported inject() error on non-gcc compilers.

2006-12-29  Christian Stimming  <[e-mail]>

	* src/lasvd.cc (LaSVD_IP): Fix dimension argument for U and Vt if
	it isn't referenced. Fixes #1624333, by Matti Varjokallio
	[e-mail].

2006-12-05  Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.5.0. Dump SO_VERSION to 15.0.0.

	* include/blas*.h, lapack*.h: Mask the C++ directives by #ifdef
	__cplusplus so that these headers can be used by C as well.

2006-11-24  Christian Stimming <[e-mail]>

	* include/laexcp.h: Add extra constructor of LaException that can
	be used to replace the standard assert() macro. If you want to
	replace the standard assert macro, you have to uncomment some
	lines in laexcp.h - see the instructions there.

2006-11-22  Christian Stimming <[e-mail]>

	* include/lavd.h, symd.h, ltgmd.h: Add resize() methods where they
	were still missing. Patch by Dominik Wagenfuehr.

2006-11-21  Christian Stimming <[e-mail]>

	* include/blas3pp.h: Add documentation for multiplications of
	LaSymmMatDouble; add argument for choosing the transposition.

	* include/sybfd.h, matrix/src/sybfd.cc: Completely remove separate
	factorization class because it isn't needed, as submitted by
	Dominik Wagenfuehr. Move function implementations into cc file.

2006-11-17  Christian Stimming <[e-mail]>

	* include/sybfd.h: Add initial new class implementation by Dominik
	Wagenfuehr.

	* include/sybmd.h, matrix/src/sybmd.cc: Add documentation and some
	missing functions. Patch by Dominik Wagenfuehr.

2006-11-14  Christian Stimming <[e-mail]>

	* matrix/src/sybmd.cc: Fix the storage dimensions of symmetric
	banded matrix construction.

2006-08-21  Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.4.13

	* lapackpp.vcproj: Enable LA_COMPLEX_SUPPORT for MSVC
	compilation. MSVC 7.1 seems to accept this directly. Added
	matrix/testing/tgc.vcproj as a test case.

	* include/vc.h: Fix missing include.

2006-08-20  Christian Stimming  <[e-mail]>

	* include/lacomplex.h: is no longer generated but directly
	included in CVS.

2006-08-17  Christian Stimming <[e-mail]>

	* include/laslv, src/linslv.c: Add LaLUInverseIP for complex
	matrix.

	* include/lapackc.h: Add declaration of zgetri.

2006-08-10  Christian Stimming <[e-mail]>

	* include/trfd.h: Fix LaTridiagFactDouble class that was
	completely broken in most of its member functions. Aaarg.

	* include/trfd.h, matrix/src/trmd.cc: Move LaTridiagMatFactorize()
	from inline function into trmd.cc to avoid linker trouble on MSVC.

2006-08-09  Christian Stimming <[e-mail]>

	* configure.ac: Update win32/mingw DLL build rules.

2006-08-06  Christian Stimming  <[e-mail]>

	* configure.ac: Release version 2.4.12

2006-08-01   Christian Stimming <[e-mail]>

	* include/trmd.h, matrix/src/trmd.cc: Completely overhaul the
	LaTridiagMatDouble class: Move stupid inline functions into the cc
	file and vice versa. Fix operator(int,int) that was completely
	broken. Document everything. Explain how the diagonal assignments
	should be done by using inject().

2006-07-24   Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.4.11

2006-07-20   Christian Stimming <[e-mail]>

	* blaspp/testing/check_blas1pp.cc: Add new tests for
	Blas_H_Dot_Prod and LaVectorDouble::inject.

2006-07-14   Christian Stimming <[e-mail]>

	* src/linslv.cc, testing/tGenSolve.cc, testing/tComplexSolve.cc:
	Finally fix the solving functions for non-square matrices using QR
	decomposition. These were broken all along, especially since the
	"fix" on 2004-08-20. Now they work with all matrices again and the
	testing programs have checking routines for "make check".

2006-05-23   Christian Stimming <[e-mail]>

	* configure.ac: Cleanup of OS_foo macro to ensure it doesn't
	contain decimal points.

2006-05-17   Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.4.10

2006-05-16   Christian Stimming <[e-mail]>

	* lapackpp.vcproj: Update the project file so that MSVC builds a
	real DLL. Requires MSVC>=7.1 now, though.

	* src/Makefile.am: On windows only install the liblapackpp-1.dll,
	not the old lapackpp32.dll any longer.
	
	* contrib/README, lapack.diff: Add explanation for creating real
	MSVC dlls.

	* include/arch.h: Add appropriate checks for Windows/MSVC
	dllexport declaration.

	* include/blaspp.h, blas/src/blaspp.cc: Move operator function
	bodies into extra file to avoid function importing issues.

2006-05-15   Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.4.9

	* README: Clarify pre-compiled DLL. Improve documentation.

2006-05-08   Christian Stimming <[e-mail]>

	* include/laslv.h: Rename LaEigSolveVecIP into
	LaEigSolveSymmetricVecIP because since 2004-09-08 the former
	looked as if it accepted any non-symmetric matrix.

2006-04-19   Christian Stimming <[e-mail]>

	* include/Makefile.am: Don't install the <blas++.h> headers with
	the "++" in the filename anymore.

	* include/blaspp.h: Fix some stupid arguments in the operators.

2006-04-11   Christian Stimming <[e-mail]>

	* blaspp/src/blas1pp.cc: Disable abs() definition which seems
	unneeded anyway?

	* lapackpp.vcproj: Fix the project file for MSVCs VC71
	compiler. Remove obsolete files.

2006-04-10   Christian Stimming <[e-mail]>

	* configure.ac: Release version 2.4.8

	* lapackpp.vcproj: Switch the "Runtime library" to from
	multi-threaded debug to "single-threaded debug" since that's what
	obviously most other projects expect.

2006-03-24   Christian Stimming <[e-mail]>

	* include/gmd.h: Remove erroneous extra trailing semicolon.

	* testing/tGenSolve.cc: On amd64/gcc3.4, LaGenerateMat is broken!
	Use la::rand() instead.

	* src/dtimmg.c: Add F77NAME macro and comment out function
	prototype to fix compile error on gcc3.4/amd64.

2006-03-23   Christian Stimming <[e-mail]>

	* blaspp/src/blas3pp.cc: Fix wrong matrix size checking as pointed
	out by Christian Fellenberg.

2006-03-13   Christian Stimming <[e-mail]>

	* include/laversion.h.in: Add header with lapackpp version numbers
	so that code conditional on particular lapackpp versions can be
	written.

	* blaspp/src/blas3pp.cc: Abort by assertion only if matrices are
	actually being used.

2006-02-24   Christian Stimming <[e-mail]>

	* configure.ac: Release lapackpp-2.4.7

	* include/gmd.h, gmc.h: Add scale() method and operator*= as an
	alias to Blas_Scale.

	* include/gmd.h, gmc.h, gmi.h, gmf.h, gmli.h: Add add() method in
	addition to operator+=.

	* include/blas3pp.h: Add Blas_Scale for matrices.

	* include/latmpl.h: Add compile dependency on MSVC version number.

2006-02-10   Christian Stimming <[e-mail]>

	* blaspp/src/blas3pp.cc (max_fro_sum): Fix erroneous NormF for
	non-square matrices. Reported by Christian Fellenberg. This was
	broken in lapackpp-2.4.3 through lapackpp-2.4.6 and is fixed in
	lapackpp-2.4.7.

2006-02-01   Christian Stimming <[e-mail]>

	* configure.ac: release lapackpp-2.4.6

2006-01-31   Christian Stimming <[e-mail]>

	* matrix/src/gmtmpl.cc, mtmpl.h: Consolidate much more methods
	from the five matrix classes into this templated macro
	file. Unifies the matrices even more.

 	* include/gmd.h, gmc.h, gmi.h, gmf.h, gmli.h: Add row() and
	col() methods to access single rows and columns as submatrix
	views more easily. Implementation is in matrix/src/gmtmpl.cc.

2006-01-18   Christian Stimming <[e-mail]>

	* configure.ac: Release lapackpp-2.4.5

2006-01-04   Christian Stimming <[e-mail]>

	* macros/acx_lapackpp.m4: Remove extra linker flags FLIBS from
	LAPACKPP_LIBS since they are now already linked into
	liblapackpp.so.

	* include/gmd.h, gmc.h, gmi.h, gmf.h, gmli.h,
	matrix/src/gmtmpl.cc: Add a whole new lot of static constructors
	for elementary matrices. These are in fact just wrappers for the
	template functions from latmpl.h, and to save some typing the
	implementations are defined by macros in one common file
	gmtmpl.cc.

	* include/latmpl.h: Add la::int_rand template. Fix rand() usage.

	* include/lacomplex.h.in, laindex.h: Add operator== and operator!=
	where useful.

	* include/gmc.h: Add LaGenMatComplex::real() and imag() methods.

2006-01-03   Christian Stimming <[e-mail]>

	* include/latmpl.h: Add many more constructors for elementary
	matrices.

	* src/genmd.cc: Fix errorneous call of rand().

	* configure.ac: Add workaround for gfortran-4.0.2 which says
	libgfortranbegin.a were to be linked to the library, although it
	shouldn't.

	* src/dopla.c, src/dtimmg.c: Add used subroutines directly into
	source file.

	* include/latmpl.h, matrix/src/gmconv.cc: Renamed la::diag to
	la::from_diag. Added la::convert_mat.

2006-01-02  Christian Stimming <[e-mail]>