2022/09/12    Nakata Maho <maho@riken.jp> 2.0.1
            * tag 2.0.1
            * Cuda support for Rgemm dd and Rsyrk dd (Tesla V100 and A100); V100 and A100 support revived by Nakazato (aizu-u)
            * Add benchmark results for Ryzen 3970X, Core5-8500B, Xeon-E5-2623, Raspberry Pi4, NVIDIA Tesla V100 and A100.
            * provide DLLs for MINGW mplapack.

2022/07/26    Nakata Maho <maho@riken.jp> 2.0.0
            * tag 2.0.0
            * Update test results for aarch64-unknown-linux-gnu, x86_64-apple-darwin20.6.0,
              x86_64-pc-linux-gnu, x86_64-pc-linux-gnu_inteloneapi and x86_64-w64-mingw32.
            * Update Dockerfiles accordingly.

2022/07/24    Nakata Maho <maho@riken.jp> 2.0
            * Fix double-double CLS (Ctest.in) by disabling gcc optimization mplapack/test/matgen/Claror.cpp. (#30)
            * Correct fixes for Rgejsv, Cgejsv, Rgesvj and Cgesvj using dd/qd. Rlamch("O")s are replaced by
	      "one / sfmin", indicated by the comments in the source codes.
            * Fix qd/dd division overflow bug.

2022/07/22    Nakata Maho <maho@riken.jp> 2.0 branch
            * Fix iMparam2stage.
            * Fix issues for test programs; stack overflows, possible memory leaks and memory accesses.
            * Extensive tests and fixes for mingw64; we provide DLLs for libqd and mpblas.
            * Correct exponent ranges for GMP, MPFR and each OS.
            * Update test results for aarch64-unknown-linux-gnu, x86_64-apple-darwin20.6.0,
              x86_64-pc-linux-gnu, x86_64-pc-linux-gnu_inteloneapi and x86_64-w64-mingw32.

2022/06/21    Nakata Maho <maho@riken.jp> 2.0 branch
            * Enable Rgejsv, Cgejsv, Rgesvj and Cgesvj for dd/qd by changing Rlamch("O") to approx.
              1e-16 * Rlamch("O").

2022/06/14    Nakata Maho <maho@riken.jp> 2.0.0 alpha
            * support all complex routines for all precisions except for the mixed-precision version.
            * support all Rectangular Full Packed (RFP) routines
            * on macOS, dylibs are now correctly linked
            * on Linux, we set absolute rpath so that mp libraries are correctly linked.
            * remove debugging symbols by default
            * improvements in the installation process
            * overflow bug in sqrt is fixed for libqd

2022/05/25    Nakata Maho <maho@riken.jp> 2.0 branch
            * 2.0.0 alpha.

2022/05/16    Nakata Maho <maho@riken.jp> 2.0 branch
            * Disable Rgejsv, Cgejsv, Rgesvj and Cgesvj by dd/qd, since we cannot
              gurantee the precisions.

2022/05/16    Nakata Maho <maho@riken.jp> 2.0 branch
            * Rtest_rfp.in, Ctest_rfp.in: Test for linear equation routines RFP format
              have passed for all precisions.

2022/05/10    Nakata Maho <maho@riken.jp> 2.0 branch
            * csd.in: Test for CS decomposition routines
              have passed for all precisions (except for GMP; it doesn't support csd).

2022/05/09    Nakata Maho <maho@riken.jp> 2.0 branch
            * gsv.in: Test for Generalized SVD routines
              have passed for all precisions.
            * Csg.in: MPFR tests have also passed. Thus, tests of the Generalized
              Nonsymmetric Eigenvalue Problem routines have passed for all precisions.

2022/05/08    Nakata Maho <maho@riken.jp> 2.0 branch
            * Cgd.in: Tests for complex Nonsymmetric generalized Eigenvalue/Schur Form Driver/Expert Driver
              have passed for all precisions.

2022/05/06    Nakata Maho <maho@riken.jp> 2.0 branch
            * svd.in: Tests of the complex Singular Value Decomposition routines
              have passed for all precisions except for dd and qd.

2022/05/04    Nakata Maho <maho@riken.jp> 2.0 branch
            * Cgg.in: Tests of the complex Generalized Nonsymmetric Eigenvalue Problem routines
              have passed for all precisions.

2022/05/03    Nakata Maho <maho@riken.jp> 2.0 branch
            * Cgg.in: Tests of the Generalized Nonsymmetric Eigenvalue Problem routines
              have passed for all precisions expect for _Float128 and _Float64x.
            * Csg.in: Tests of the Generalized Nonsymmetric Eigenvalue Problem routines
              have passed for all precisions expect for mpfr.

2022/05/02    Nakata Maho <maho@riken.jp> 2.0 branch
            * Ced.in: Tests for Complex Nonsymmetric Schur Form, Eigenvalue expert/expert drivers
              have passed for all precisions.

2022/05/01    Nakata Maho <maho@riken.jp> 2.0 branch
            * se2.in: Tests for Hermitian Eigenvalue Problem routines for complex numbers have passed
              for all precisions.

2022/04/30    Nakata Maho <maho@riken.jp> 2.0 branch
            * sep.in: Tests for Hermitian Eigenvalue Problem routines for complex numbers have passed
              for all precisions.

2022/04/26    Nakata Maho <maho@riken.jp> 2.0 branch
            * nep.in: Test for Nonsymmetric Eigenvalue Problem routines for complex have passed
              for all precisions.
            * lse.in: Test for Constrained Linear Least Squares routines for complex have passed
              for all precisions.
            * Csb.in: Test for Hermitian Eigenvalue Problem routines have passed
            * glm.in: Tests of the Generalized Linear Regression Model routines for complex have passed
            * gqr.in: testing Generalized QR and RQ routines for complex have passed
            * Cbb.in: reduction of a general band matrix to real bidiagonal form tests have passed

2022/04/24    Nakata Maho <maho@riken.jp> 2.0 branch
            * Cec.in test (Testing complex Eigen Condition Routines) have passed for all precisions.

2022/03/21    Nakata Maho <maho@riken.jp> 2.0 branch
            * Ctest.in test have passed for all precisions, except for linear fitting.
              Some tests using SVDs still fails, however, they will be fixed when eig tests
              will be done.

2021/10/29    Nakata Maho <maho@riken.jp> 1.0.1
            * Add -fp-model precise to fix qd and dd arithmetic for Intel one API.

2021/10/01    Nakata Maho <maho@riken.jp> 1.0.0
            * Release 1.0.0

2021/08/05    Nakata Maho <maho@riken.jp> 1.0.0
            * all test programs are built successfully on Ubuntu20.04 (amd64)

2021/7/28    Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of Nonsymmetric Generalized Eigenvalue Problem driver routines
              (Rgd.in) have passed for MPFR version.
            * Tests of Generalized Symmetric Eigenvalue Problem routines (Rsg.in)
              have passed for MPFR version.

2021/7/19    Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of the Rtest.in have passed for MPFR version.
              (includes driver and expert routines)

2021/7/8    Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of the CS Decomposition routines (csd.in)
              have passed for MPFR version.

2021/7/7    Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of Generalized Singular Value Decomposition routines (gsv.in)
              have passed for MPFR version.

2021/7/5    Nakata Maho <maho@riken.jp> 1.0.0
            * Tests Singular Value Decomposition routines (svd.in)
              have passed for MPFR version.

2021/7/1    Nakata Maho <maho@riken.jp> 1.0.0
            * MPLAPACK now use Rlaqr0 in Rhseqr.cpp for big (n>=15) matrices

2021/6/30   Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of the Nonsymmetric Eigenvalue Problem Driver (Red.in)
              Rgeev (eigenvalues and eigevectors)
              have passed for MPFR version.

            * Tests of the Nonsymmetric Eigenvalue Problem Driver (Red.in)
              Rgees (Schur form)
              have passed for MPFR version.

            * Tests of the Nonsymmetric Eigenvalue Problem Expert Driver (Red.in)
              Rgeevx (eigenvalues, eigenvectors and condition numbers)
              have passed for MPFR version.

            * Tests of the Nonsymmetric Eigenvalue Problem Expert Driver (Red.in)
              Rgeesx (Schur form and condition numbers)
              have passed for MPFR version.

            * Note that currently MPLAPACK use Rhseqr.cpp: L.134 use Rlaqr0 instead of
              Rlahqr for big matrices.

            * Testing Generalized QR and RQ factorization routines
              have passed for MPFR version.

2021/6/29   Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of Rsbtrd (Rsb.in)
              (reduction of a symmetric band matrix to tridiagonal form)
              have passed for MPFR version.

            * Tests of the symmetric eigenproblem using
              the 2-stage reduction techniques (se2.in)
              have passed for MPFR version.

            * Tests of Rgbbrd (Rbb.in)
              (reduction of a general band matrix to real bidiagonal form)
              have passed for MPFR version.

2021/6/28   Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of the symmetric eigenproblem (sep.in)
              have passed for MPFR version.

2021/6/25   Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of the nonsymmetric eigenproblem (nep.in)
              have passed for MPFR version.

2021/6/21   Nakata Maho <maho@riken.jp> 1.0.0
            * Tests of the nonsymmetric eigenproblem condition estimation routines
              have passed (Rlaln2, Rlansy2, Rlanv2, Rlaexc, Rtrsyl, Rtrexc, Rtrsna,
              Rtrsen, Rlaqtr and Rtgexc) for MPFR version (Rec.in).

2021/6/3    Nakata Maho <maho@riken.jp> 1.0.0
            * rtests.in for driver routines passed for MPFR version except for
              RGB and RPO.
              Some test in RLS using SVDs failed, it will be fixed.

2021/6/1    Nakata Maho <maho@riken.jp> 1.0.0
            * ztest.in passed for MPFR version (without test for driver routines).
              Some tests using SVDs fail, however, may not be
              a big problem and will be fixed.

2021/5/26   Nakata Maho <maho@riken.jp> 1.0.0
            * dtest.in passed for MPFR version (without test for driver routines)
            * add examples for _Float64x

2021/5/12   Nakata Maho <maho@riken.jp> 1.0.0
            * Import all test program and libraries from LAPACK.

2021/4/22   Nakata Maho <maho@riken.jp> 1.0.0
            * MPBLAS passed all tests for all precisions on Ubuntu 18.04, amd64.

2021/4/20   Nakata Maho <maho@riken.jp> 1.0.0
            * MPLAPACK 1.0.0 build successfully on Ubuntu 18.04, amd64.

2021/4/19   Nakata Maho <maho@riken.jp> 1.0.0
            * 680 LAPACK subroutines are rewriteen using fable.
              * 359 routines are rewritten using fable without modifications.
            * remove fortranwrapper and blaswrapper
            * remove optimized for mplapack

2021/4/18   Nakata Maho <maho@riken.jp> 1.0.0
            * rename debug to test
            * BLAS part is rewritten using fable completely automatically.
            * 653 LAPACK subroutines are rewriteen using fable.
               * approx. 359 routines are rewritten using fable without modifications.

2021/4/11   Nakata Maho <maho@riken.jp> 0.9.3
            * build fix for CentOS7 (AArch64)
            * Rlamch fixes for win64

2021/4/5    Nakata Maho <maho@riken.jp> 0.9.1
            * build fix for CentOS7 (x86_64, _Float128 via libquadmath)
            * build verified for CentOS8 (AArm64, _Float128 via glibc)

2021/4/1    Nakata Maho <maho@riken.jp> 0.9.0
            * Release

2021/3/16   Nakata Maho <maho@riken.jp> 0.9.0(alpha)
            * Migrate from __float128 to _Float128

2021/3/8    Nakata Maho <maho@riken.jp> 0.9.0(alpha)
            * rename from MPACK to MPLAPACK
            * rename from MBLAS to MPBLAS
            * rename from MPLAPACK to MPLAPACK

2012/12/18  Nakata Maho <maho@riken.jp> 0.8.0.
	    * Intel Xeon Phi support
		This build is supported by
		very unusual way.
	      Please see
		misc/prepare_mpacklibs_for_mic.sh
	    and misc/build_mpack_for_mic.sh for details.
	    * build fixes.

2012/11/21  Nakata Maho <maho@riken.jp> 0.8.0.
            * Integration of CUDA code.
            * Far better benchmarking.

2012/03/25  Nakata Maho <maho@riken.jp> 0.7.0
            * Add __float128 support for gcc.
            * Many cleanups.

2012/02/14  Nakata Maho <maho@riken.jp> 0.7.0
            * We can choose which libraries to be built.
            * We build BLAS and LAPACK as well to avoid version mismatch
              or small unnecessary fixes by distributions.
            * Change file/directory layout.

2010/08/19  Nakata Maho <maho@riken.jp> 0.6.7
            * Add Rgecon etc. condition estimation.

2010/08/07  Nakata Maho <maho@riken.jp> 0.6.7
            * All files are now 2 clause BSD style license.

2010/08/05  Nakata Maho <maho@riken.jp> 0.6.6
            * build fixes on various environments just using internal
              MP libraries.
            * Now 92 mlapack routines.
            * mpfr has been update to 3.0.0.

2010/05/20  Nakata Maho <maho@riken.jp> 0.6.5
            * working towards 0.6.5 release

2010/05/04  Nakata Maho <maho@riken.jp> 0.6.5
            * Rewrite testing (quality assurance) part
            * Include GMP/MPFR/MPC/QD under external directory

2010/03/15  Nakata Maho <maho@riken.jp> 0.6.5
            * license has been changed to two clause BSD license.
            * using std::complex<double> instead of double_complex

2010/02/03  Nakata Maho <maho@riken.jp> 0.6.5
            * Environment variable MPACK_GMP_PRECISION, and MPACK_MPFR_PRECISION
              are available.

2010/01/28  Nakata Maho <maho@riken.jp> 0.6.5
            * Add mpfr and double (in C; de-facto IEEE 754) version.
            * Various cosmetic fixes.

2010/01/13  Nakata Maho <maho@riken.jp> 0.6.4
            * 90 routines.
            * RE on MacOSX, PPC/Intel Tiger, Cent OS (Linux x86-64)
               and mingw Windows.
            * Total 666 routines.
              LAPACK 3.1.1 : d*f, z*f, izmax, ilaenv, dlamch : 674
                             not to be implemented              14
                                                (Rlacon, Clacon -2)
2009/11/24  Nakata Maho <maho@riken.jp> 0.6.0
            * All LAPACK routines are translated into C++.
              Total 668 routines.
              LAPACK 3.1.1 : d*f, z*f, izmax, ilaenv, dlamch : 674
                             not to be implemented              12

2009/10/06  Nakata Maho <maho@riken.jp> 0.6.0
            * Start OpenMP support
            * BLAS wrapper support

2009/10/05  Nakata Maho <maho@riken.jp> 0.5.1
            * -Wall -Werror support 
            * clean up with gcc 4.4

2009/09/24  Nakata Maho <maho@riken.jp> 0.5.0
            * add examples
            * using namespace std cleanup
            * builds fine on fbsd7/amd64
            * lost compatibility between sdpa-gmp, qd and dd but will be added
              in 0.6.0.

2009/09/19  Nakata Maho <maho@riken.jp> 0.1.0
            * mlapack GMP/QD/DD, and testing is 
            * <inttypes.h> to use 32bit int or 64bit int

2009/09/12  Nakata Maho <maho@riken.jp> 0.1.0
            * GMP/QD/DD support by shared source
            * abondon template support

2009/05/18  Nakata Maho <maho@riken.jp> 0.1.0
            * Start working as 0.1.0
            * QD/DD support
            * template support

2009/02/09  Nakata Maho <maho@riken.jp> 0.0.9
            * Fix small misc stuffs.
            * Upload as 0.0.9 mpack package.

2009/02/06  Nakata Maho <maho@riken.jp> 0.0.9
            * Fix bug in Msign and Rlaswp.
            * MPACK supports sdpa-gmp!
            * 355 MLAPACK routines are implemented, and 37 are tested.
            * 76 MBLAS routines are implemented and tested. 

2009/01/07  Nakata Maho <maho@riken.jp> 0.0.8
            * Tested MLAPACK routines 
              Rpotrf Rsteqr Rsterf Rsytd2 Rsytrd Rsyev
            * 355 MLAPACK routines are implemented, and 29 are tested.
            * 76 MBLAS routines are implemented and tested. 
            * Moved to sourceforge.net.
 
2009/01/01  Nakata Maho <maho@riken.jp> 0.0.7
	    * Workaround expornential bug (?) in 64bit/GMP.
            * Bug fix in Rlarft. dlarft.f should also mention N<=K
            * Tested MLAPACK routines 
              Rorg2l Rorg2r 
              Rorgqr Rorgql Rorgtr Rpotrf Rsteqr
            * 355 MLAPACK routines are implemented, and 25 are tested.
            * 76 MBLAS routines are implemented and tested. 

2008/08/22  Nakata Maho <maho@riken.jp> 0.0.6
            * Build fix.

2008/08/14  Nakata Maho <maho@riken.jp> 0.0.5
	    * Implement some of LAPACK functions.
                Rlamch Rlae2  Rlaev2 Rlassq Rlanst
                Rlansy Rlapy2 Rlarf  Rlarfg Rpotf2
                Rlarfb Rlaset Rlarft Rlartg Rlascl
                Rlasrt Rlasr  Rlatrd
            * Generating of prototypes from the BLAS/LAPACK code and f2c.
            
2008/07/23  Nakata Maho <maho@riken.jp> 0.0.4
	    All test cases have been imported. Currently,
	    Ctrsv, Rsyrk, Rtrsv, Rtrsm, Rtbsv and Rtpsv

2008/07/18  Nakata Maho <maho@riken.jp> 0.0.3
	    "OpenOffice.org" is inappropriately used in headers. 

2008/07/17  Nakata Maho <maho@riken.jp> 0.0.2
	    Build fix on gcc-4.3 (Fedora 9).

2008/07/15  Nakata Maho <maho@riken.jp>
	    Configurable on MacOSX and FreeBSD. 

2008/07/11  Nakata Maho <maho@riken.jp>
	    Add configure script.

2008/07/02  Nakata Maho <maho@riken.jp>
	    License has been changed to LGPLv3.
            Prepare configure stuffs.

2008/06/27  Nakata Maho <maho@riken.jp>
            Add modifid BSD liceence headers to all files,
            From http://www.netlib.org/lapack/COPYING .

2008/06/26  Nakata Maho <maho@riken.jp>
            Initial release. No configure, no Makefile, no
            installation instruction. Only the sources.

