phynexis-deps/arpack-ng
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
master
arpack-ng/CHANGES
T
Markus Mützel 6ae2072b0c PARPACK: Initialize machine dependent constants only once in PDNAITR 
The current implementation checks the value of the variable `aitr_first` but
sets `first`.

Change that to set `aitr_first` to false after the initialization has run.

This also brings that part of the implementation of `PDNAITR` in line with
the corresponding logic that is used in the implementation of the other
`P?NAITR`.
2026-08-01 15:59:19 +02:00

466 lines
18 KiB
Plaintext
Raw Permalink Blame History

arpack-ng - next
[ Evan Biederstedt ]
* Fix warnings for "Warning: Obsolescent feature: Old-style character length at (1)", compiled with `gfortran-16 -fpic -g -O2 -Wall -pedantic -mtune=native`
* Changed GitHub CI workflows for Mac OS to use 'brew install open-mpi', as `mpich` fails.
[ Theodore Chang ]
* CMake: Improve dependency linking, bump up minimum C++ standard version.
[ Franck Houssen ]
* [BUG FIX] Fix arpack solver to handle real and imag shifts.
* [BUG FIX] Make sure the restart file of the arpack solver i written.
* [BUG FIX] Change arpack solver API.
[ Henri Menke ]
* [BUG FIX] Add missing stdexcept header
[ Kyle Guinn ]
* Build PARPACK p[sd]lamch10.f with FFLAGS from ./configure instead of forcing -O0. Build all of PARPACK with AM_FFLAGS.
* [BUG FIX] Don't install pkg-config files for libraries that were not built.
[ Markus Mützel ]
* CMake: Fix running CTests on Windows.
* Use GNU Fortran ABI in CI tests on macOS.
* Initialize machine dependent constants only once in `PDNAITR`.
[ Szabolcs Horvát ]
* fix obsolescent character length warnings
-- TODO <TODO@users.noreply.github.com> DATE
arpack-ng - 3.9.1
[ Fabien Péan ]
* pyarpack: Ensure that the matrix properties (symmetric/hermitian) fit the solver (CG/LDL) with which they are used in the tests
* [BUG FIX] Tests for PARPACK with C/C++ bindings icb_parpack_c and icb_parpack_cpp are now really parallel and split the problem across MPI processes.
* Update arpackmm test suite: enable solving non-symmetric tests with BiCG solver
* README: Add details on Windows installation.
[ Szabolcs Horvát ]
* [BUG FIX] Ensure that LAPACK RNG state is propagated (regression in 3.9.0).
* [BUG FIX] Ensure that separate random seeds are used on different parallel thread in D and S versions of functions (issue from original ARPACK).
[ Ruoyu Feng ]
* ICB(arpackdef.h): distinct intel llvm compiler (icx with clang-cl) from msvc on windows
* ICB(arpackdef.h): Undef macro I if complex.h from msvc version is loaded, which is an usual name and causes issues on arpackSolver.
[ Franck Houssen ]
* [BUG FIX] Fix install: headers in /path/to/local/include/arpack for ICB samples
* [BUG FIX] Fix install: headers in /path/to/local/include/arpack
* arpackmm: allow for using LA/SA magnitudes.
* Rename icbexmm option into eigen option.
* README: document how to use ICB.
* [BUG FIX] arpackmm: fix restart.
* README: document where to find arpack user's guide.
-- Franck Houssen <fghoussen@users.noreply.github.com> Sat, 14 Oct 2023 13:37:37 +0200
[ Franck Houssen ]
* Add documentation on ILP64.
arpack-ng - 3.9.0
[ Vikas Sharma ]
* Improve README.
[ Fabien Péan ]
* CI: Enable job `windows_latest_cmake` to run all tests
* CMake: Fix BLAS and LAPACK static library order needed to consume the library on Windows with static linkage
* Fix using ARPACK on Windows with MSVC compiler from C++17 onwards
[ Zhentao Wang ]
* [BUG FIX] parpack.h and parpack.hpp: type of rwork should be real instead of complex.
* Allow ritz_option {"LR", "SR", "LI", "SI"} for complex eigenvalue problems in ICB.
[ Jose E. Roman ]
* Avoid using isnan() in tests, since is GNU-specific
[ Tom Payerle ]
* Change the continuation line format for stat.h, debug.h
[ John Doe ]
* Avoid calling [c|z]dotc for better portability on macOS
[ Dima Pasechnik ]
* [BUG FIX] autotools: replace obsolete AC_TRY_COMPILE macros.
* Support for NAG's nagfor Fortran compiler
[ Franck Houssen ]
* Create one .cmake file per arpack-ng flavor (32-bits, 64-bits, ILP64).
* Test autotools pkg-config (*.pc files) with/without LIBSUFFIX/ITF64SUFFIX.
* Test CMake find_package (*.cmake files) with/without LIBSUFFIX/ITF64SUFFIX.
* [BUG FIX] autotools: ICB must be checked first (MPI changes compilers).
* [BUG FIX] BLAS/LAPACK: allow suffixes in case BLAS/LAPACK can not provide ICB.
* [BUG FIX] Compile C programs with ICB.
* arpackmm: command line bug fix.
* arpackmm: restart bug fix.
* pyarpack: fix compilation warning, test on macos and latest boost-python (1.79).
* arpackSolver: fix error messages.
* [BUG FIX] Make sure iseed is always initialized to values allowed by lapack ?larnv.
* [BUG FIX] According to lapack doc of ?larnv, iseed(4) must be odd.
* [BUG FIX] Use MPI ICB types (mpi_f08) instead of integer(kind=i_int).
* parpack: no ILP64 support.
[ Haoyang Liu ]
* CMake: minimum required version changed to 3.0
* CMake: add C99 standard checking
* CI: Support for centos7 added.
* CI: Add `scripts/travis_centos.sh` for centos builds
[ Robert Schütz ]
* use CMAKE_INSTALL_FULL_<dir> in arpack.pc
[ Markus Mützel ]
* CMake: Handle libraries without "lib" prefix.
* CMake: Don't override BLAS/LAPACK/MPI flags. Directly use results from the Find* modules instead.
[ Juan José García-Ripoll ]
* Adapt the C/C++ interface to accept also MSVC's non-standard complex types.
* Propagate dependencies to CMake targets that use arpack-ng:
- Create CMake-generated targets and configuration files that keep track of
arpack's dependencies (libraries, directories) and expose them to users.
- Install those files under ${prefix}/lib/cmake/arpackng* so that arpack can be
found using 'find_package(arpackng)' from CMake files.
- Add code to the arpackng-config.cmake to find required dependencies when this
module is loaded by find_package(arpackng).
-- Sylvestre Ledru <sylvestre@debian.org> Mon, 07 Dec 2020 11:37:40 +0100
arpack-ng - 3.8.0
[ Myron Oikonomakis ]
* [BUG FIX]: bmat return "G" instead of "B" for generalized matrix in arpack.hpp
* [BUG FIX]: pass arrays of chars as scalar in fortran calls in order not to crash
* when calling subroutines through icb interface
[ Izaak "Zaak" Beekman ]
* [BUG FIX]: fix 'Unknown CMake command "check_symbol_exists".' when ICB=ON.
[ Franck Houssen ]
* CI: Support for Mac OS X added in automation (GNU + "-ff2c -fno-second-underscore" options).
* CI: Support for centos added in automation.
* CI: Support for opensuse added in automation.
* arpackSolver/arpackmm: switch eigen version to 3.3.
* [BUG FIX] fix arpackdef.h (resp. arpackicb.h) must be included only by C/C++ (resp. F77/F90).
* [BUG FIX] iparam/ipntr sizes may change depending on cases.
* pyarpack: python binding based on Boost.Python.Numpy exposing C++ API.
* [CLEAN] arpackSolver API: more convenient, suppress template parameters when possible.
* [BUG FIX] ICB using rvec/select: rvec/select turned to integer
bool should be, but, is not always supported (depend on compiler, options).
* extract arpackSolver.hpp from arpakmm.cpp.
* arpackmm: add --slvItrPC option (PC: Jacobi, ILU).
* arpackmm: add --slv LLT LDLT (for SPD matrices).
* arpackmm: add --simplePrec option (to enable use of s*upd).
* arpackmm: add --dense option.
* autotools: provide *.cmake files (in addition to *.pc file).
* [BUG FIX] ILP64 support: using debug_c and stat_c.
* [BUG FIX] fix check precision which may fail with some ATLAS versions.
[ Kyle Guinn ]
* [BUG FIX]: fix 'eval: Syntax error: "(" unexpected' error at build time.
* Only build shared libraries by default. To build static libraries, use
--enable-static (autotools) or -DBUILD_SHARED_LIBS=OFF (cmake).
* Add parpack.pc and arpackSolver.pc.
[ David Schwörer ]
* Support of gfortran 10
-- Sylvestre Ledru <sylvestre@debian.org> Mon, 07 Dec 2020 11:35:57 +0100
arpack-ng - 3.7.0
[ Franck Houssen ]
* [BUG FIX] ICB: missing workev for *[ds]neupd (real+not-sym) => API/ABI change for *[ds]neupd_c.
* [BUG FIX] autotools - make distcheck: fix circular dependencies.
* arpackmm: utility to test arpack with matrix market files.
* ICB: add ILP64 support.
The idea is:
- autoheader/cmake generates arpackdef.h/arpackicb.h from arpackdef.h.in/arpackicb.h.in
- in C/C++ files: arpackdef.h defines a_int according to architecture.
- in F77/F90 files: arpackicb.h defines i_int to architecture.
- MPI does not support ILP64: integer*4 must be imposed in all
calls involving MPI (f90 example/test code).
To enable ILP64 users to compile/link, arpackdef.h/arpackicb.h is added in
the arpack installation (make install).
[ Kyle Guinn ]
* Autoconf/Automake simplifications and fixes.
* Simplify the generation of arpackdef.h.
-- Sylvestre Ledru <sylvestre@debian.org> Sat, 12 Jan 2019 16:24:00 +0100
arpack-ng - 3.6.3
[ Franck Houssen ]
* Add Fortran common initialization (block data).
[ Marco Caliari ]
* Give up forcing the initial residual to be in the range of the operator OP after a restart (Closes: #142).
-- Sylvestre Ledru <sylvestre@debian.org> Wed, 19 Sep 2018 09:59:59 +0200
arpack-ng - 3.6.2
* Remove all trailing whitespaces
[ Franck Houssen ]
* Install: move headers into a dedicated directory (local/include/arpack).
(Closes #126)
* Add configuration summary.
* Improve the flag detection. Hopefully fix the ppc64el and other archs
issues in Debian
-- Sylvestre Ledru <sylvestre@debian.org> Sat, 23 Jun 2018 14:56:54 +0200
arpack-ng - 3.6.1
[ Ruslan Kabatsayev ]
* Fix a regression on i386 and other archs (Closes #123)
-- Sylvestre Ledru <sylvestre@debian.org> Thu Jun 7 21:41:16 2018 +0200
arpack-ng - 3.6.0
[ Franck Houssen ]
* Add support for ISO_C_BINDING (Fortran 2003) for ARPACK, PARPACK (Fortran <-> C/C++).
ARPACK: example of C/Fortran binding can be found in the TESTS/icb_arpack_c.c file.
ARPACK: example of C++/Fortran binding can be found in the TESTS/icb_arpack_cpp.cpp file.
PARPACK: example of C/Fortran binding can be found in the PARPACK/TESTS/MPI/icb_parpack_c.c file.
PARPACK: example of C++/Fortran binding can be found in the PARPACK/TESTS/MPI/icb_parpack_cpp.cpp file.
DEBUG: add support for debug.
STAT: add support for statistics (timers, nb operations, ...).
* Provide tarball generation using cmake (cpack).
* Provide find_package for (cmake) users to find arpack-ng.
[ Denis Davydov ]
* Rename pslamch to pslamch10 to avoid symbol collision with Scalapack 2.0.2 in MPI context.
[ Kyle Guinn ]
* Autoconf cleanup; move generated files to the build-aux subdirectory.
[ Marco Caliari ]
* Force the initial residual to be in the range of the operator OP in the standard case, too (Closes: #79).
[ Sylvestre Ledru ]
* Add coverage information to improve testing: https://coveralls.io/github/opencollab/arpack-ng
[ Darcy Beurle]
* Add C++11 interface through arpack.hpp and parpack.hpp
* Rewrite C++ examples / tests demonstrating new C++11 interface
* Pre-C++11 interface available through arpack.h and parpack.h
-- Sylvestre Ledru <sylvestre@debian.org> Mon, 30 Oct 2017 14:21:48 +0200
arpack-ng - 3.5.0
[ Julien Schueller ]
* Improve cmake build system: disable C++ detection, set default build type.
[ Marco Atzeri]
* Use AC_PROG_FC instead of AC_PROG_F77 for proper initialization
for the usage of AC_FC_LINE_LENGTH. Noted on Cygwin.
[ Denis Davydov ]
* Improve cmake build system: add make install and fix shared libraries.
[ Zhang Z ]
* fix usages of DLACPY to not alias inputs
(patch from https://software.intel.com/en-us/articles/how-to-resolve-arpack-issues-with-intel-mkl-110-update-3)
[ Iskakov Sergei ]
* Fix possible deadlock when PARPACK call uses communicator with a larger
number of CPUs than previous call
[ Kyle Guinn ]
* Portability improvements to the autotools build system.
* Let cmake guess the default installation directories. Can be
overridden by changing CMAKE_INSTALL_LIBDIR and CMAKE_INSTALL_BINDIR.
* Shared libraries built by cmake now have their SONAME set identical to
those built by autotools.
[ Marco Caliari ]
* Avoid purification stage in [d,s]neupd.f if it requires division
by zero (Closes: #58)
-- Sylvestre Ledru <sylvestre@debian.org> Mon, 15 May 2017 14:21:48 +0200
arpack-ng - 3.4.0
[ Milan Bouchet-Valat ]
* Allow adding suffixes to symbols and library names to build ILP64 version
based on ILP64 BLAS/LAPACK with suffixes. This avoids conflicts when loading
libraries with different integer sizes in the same program.
[ Martin Reuter ]
* Add the support of cmake build system
-- Sylvestre Ledru <sylvestre@debian.org> Sat, 02 Jul 2016 21:51:52 +0200
arpack-ng - 3.3.0
[ Denis Davydov ]
* Rename pdlamch to pdlamch10 to avoid symbol collision with Scalapack 2.0.2 in MPI context.
[ Kyle Guinn ]
* General improvements on the build system
* libparpack links against libarpack (instead of doing a static link)
[ Guillaume Horel ]
* reverts using {d,s}lahqr from lapack 2
* use dlahqr from lapack 3 instead of dlaqrb (credit to Marco Caliari)
-- Sylvestre Ledru <sylvestre@debian.org> Mon, 12 October 2015 08:40:51 +0200
arpack-ng - 3.2.0
* Switch to github - https://github.com/opencollab/arpack-ng/
* Fix dsneupd select/calculate wrong eigenpairs if rvec = true
by using dlahqr and slahqr from lapack2 (Closes: #3)
-- Sylvestre Ledru <sylvestre@debian.org> Sat, 14 Nov 2014 16:25:36 +0200
arpack-ng - 3.1.5
* Build all examples and run them as tests
* Fix the version of arpack-ng itself
* Switch to automake 1.14.1
[ Ruediger Meier ]
* Do not install test binaries (Closes: #1348)
[ Nikita Styopin ]
* Fix the diagonal matrix example (dndrv5) (Closes: #1397)
-- Sylvestre Ledru <sylvestre@debian.org> Sat, 15 Feb 2014 14:24:42 +0200
arpack-ng - 3.1.4
* libparpack2: missing dependency on MPI:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=718790
* Replace LAPACK second function with ARPACK's own arscnd in PARPACK
* Fix issue #1259 in DSEUPD and SSEUPD
The Ritz vector purification step assumes workl(iq) still contains the
original Q matrix. This is however overwritten by the call to xGEQR2
earlier.
.
This patch fixes the issue by making a copy of the last row of the
eigenvector matrix, after it is recomputed after QR by xORM2R. The work
space WORKL(IW+NCV:IW+2*NCV) is not used later in the routine, and can
be used for this.
* Use configure supplied blas and lapack in the pkg-config.
Thanks to Ward Poelmans (Closes: #1320)
* Switch to automake 1.14 + libtool 2.4.2.
Thanks to Ward Poelmans (Closes: #1321)
* dseupd routine may lead to a segmentation fault
Thanks to Edouard Canot (Closes: #1323)
* dsaupd and 'BE' option returns wrong eigenvalues for a SPD matrix
Thanks to Edouard Canot (Closes: #1329)
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Mon, 07 Oct 2013 14:24:42 +0200
arpack-ng - 3.1.3
[ Jordi Gutiérrez Hermoso ]
* Replace depcomp symlink with actual file.
* Update libtool usage. Thanks to John W. Eaton <jwe@octave.org>.
* Replace arpack.pc with proper autotooled arpack.pc.in
* Add debug.h to TESTS/Makefile.am sources
* "make dist" is functional
* Also build the library "libparpacksrcblacs" (PARPACK/UTIL/BLACS/)
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Tue, 02 Apr 2013 10:53:08 +0200
arpack-ng - 3.1.2
* Wrong call to pdlamch was causing segfaults
Thanks to Kyrre Sjøbæk for finding the bug and the fix.
* Get rid of the mpif.h occurrences in the source code (Closes: #782)
* Compile also PARPACK / MPI example (Closes: #783)
* Configure detected built-in LAPACK and BLAS, but refused to use them
(Closes: #784)
* Fixed division by zero in smlnum by using p[d,s]lamch instead of the
serial. Thanks to Umberto De Giovannini.
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Fri, 22 Jun 2012 22:05:41 +0200
arpack-ng - 3.1.1
* Option --enable-maintainer-mode added to the configure
* --disable-mpi disables the build of parpack (Closes: #714)
* Switch to automake 1.11.3
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Mon, 21 May 2012 09:08:41 +0200
arpack-ng - 3.1.0
* Many bug fixes in the parpack lib. It is an old patch from upstream.
Thanks to Viral Shah for pinging us on this subject.
See the PARPACK_CHANGES file for the details.
* Change the bug report from arpack@caam.rice.edu to
http://forge.scilab.org/index.php/p/arpack-ng/issues/
* Provide a M4 macro (detect_arpack_bug.m4) to check if the underlying
arpack is buggy (ie not arpack-ng). This allows developer applications
to perform the check in their autotools build system (configure).
* Fixed a lack of appropriate bounds check in DNAUP2. Thanks to Pauli Virtanen
for the patch (Closes: #632)
* Update of the doc about TOL in dnaupd.
* Reorder bug fixed when eigenvectors are requested and the resulting
number of converged eigenvalues is less than the number requested.
Patches from Tim Mitchell. (Closes: #664)
* TESTS/ directory added and built.
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Wed, 22 Feb 2012 10:58:39 +0100
arpack-ng - 3.0.2
* Fix a long line in pznaup2.f which was showing some wrong symbols
(Closes: #620)
* README content updated regarding ARPACK-NG
* arpack.pc (pkg-config) file added
* Update the title & version in the configure.ac
* Always search for MPILIBS (in order to have the variable correctly set)
* Explicitly link against MPI fortran libs for parpack
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Wed, 28 Dec 2011 13:45:53 +0100
arpack-ng - 3.0.1
* libtool was missing (Closes: #615)
* Missing license information (Closes: #614)
* TODO added
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Tue, 13 Dec 2011 16:33:25 +0100
arpack-ng - 3.0
* Patches from Scilab
second_NONE used by default (TO DO replace by second in LAPACK)
second_NONE works with all fortrans compilers (used by default with Scilab)
sneupd.f, cneupd.f: modified for scilab add a check on nconv value (Scilab bug fix)
dnaupd.f: modified NEV Integer: INPUT/OUTPUT before only INPUT (Scilab bug fix)
* Patches from Octave: (Thanks to John W. EATON)
dneupd.f: Restore value of nconv
dseupd.f: Restore value of nconv
sseupd.f: Change GOTO target to eliminate warning about landing on end if.
zneupd.f: Restore value of nconv
* Compilation
Apply gentoo patches to use an autotools build system
Build system updated to build with Visual Studio 2010 + Intel fortran 2011 compiles on Windows.
Specify the SONAME to libarpack.so.2 (no API/ABI changes compare to version 2.0)
-- Sylvestre Ledru <sylvestre.ledru@scilab-enterprises.com> Sat, 10 Dec 2011 20:32:45 +0100
Reference in New Issue View Git Blame Copy Permalink