Bump fblaslapack to address valgrind error in dstegr

Add lgamma checks

More numerically stable in quadrature, particularly
where we have to evaluate gamma ratios

Fix LGamma config test

Handle that gamma sometimes means lgamma

Add LAPACK support for STEGR

Merge branch 'barry/fix-superlu_dist-py-for-gpus' into 'master'

SuperLU_DIST: update to version 6.3.0

See merge request petsc/petsc!2048

SuperLU_DIST: update to version 6.3.0

SuperLU_DIST.py GPU code currently requires OpenMP

Remove -download-superlu_dist-gpu option and use --with-cuda=1

Keep backward compatibility with older super

SuperLU_DIST: update to version 6.3.0

SuperLU_DIST.py GPU code currently requires OpenMP

Remove -download-superlu_dist-gpu option and use --with-cuda=1

Keep backward compatibility with older superlu_dist versions.

show more ...

ESSL: reorganize code that handles PETSC_MISSING_LAPACK by using PetscMissingLapack()

And sync the missing function list with missing functions in current ESSL

If we find more LAPACK symbols missin

ESSL: reorganize code that handles PETSC_MISSING_LAPACK by using PetscMissingLapack()

And sync the missing function list with missing functions in current ESSL

If we find more LAPACK symbols missing in commonly used LAPACK impls,
we can add them back in the new format.

src/ksp/pc/impls/bddc/bddcprivate.c has alternate code for
PETSC_MISSING_LAPACK_GESVD. This code is retained for now [in case we
need to add this flag back in]

show more ...

Quick return of configure when --download-xxx but xxx does not work with batch

Merge remote-tracking branch 'origin/maint'

use cleaner @staticmethod syntax

Suggested-by: Jed Brown <jed@jedbrown.org>

Turn off hypre withinstallbatch ONLY if KNL system

Reported-by: Tomas Mondragon <tom.alex.mondragon@gmail.com>

Remove usage of PETSC_HAVE_FORTRAN_STDCALL, PETSC_BLASLAPACK_STDCALL, HAVE_FORTRAN_MIXED_STR_ARG flags - as Compaq f90 compiler is no longer supported

Merge remote-tracking branch 'origin/maint'

MPICH: update backup download URL to be in sync with upstream URL

Merge remote-tracking branch 'origin/balay/ftn-updates'

Balay/ftn updates

See merge request petsc/petsc!2528

Check for user provided --with-blaslapack-dir with basic ESSL libraries

Note: Since Essel does not have all of the BLAS/LAPACK and some of the calling sequences are different this may not
provide al

Check for user provided --with-blaslapack-dir with basic ESSL libraries

Note: Since Essel does not have all of the BLAS/LAPACK and some of the calling sequences are different this may not
provide all the functionality you want.

Commit-type: portability-fix, enhancement, external-packages
Reported-by: Victor Eijkhout <eijkhout@tacc.utexas.edu>

show more ...

eigen: update URL bitbucket -> gitlab and update CI version

Merge remote-tracking branch 'origin/maint'

Fix MinGW toolchain errors

more f90 fixes: remove PETSC_USING_F90 usage

petsc now requires f90 - so remove some of the f77 specific flags/code -i.e remove USE_FORTRANKIND [and special code for unsupported Compaq-F90]

This part of code isn't getting tested anyway.

Fix f90 warning with openblas:

f951: Warning: Nonexistent include directory ‘openblas_config.h’ [-Wmissing-include-dirs]

Update some packages

* Move Eigen to GitLab
* Guard ParMETIS MPI path by double quotes
* Fix SCOTCH Git URL
* Bump HPDDM version
* Fix a -Wchar-subscripts warning

configure: robustify check for files in PATH [and move duplicate code into a function]

Reported-by: Tomas Mondragon <tom.alex.mondragon@gmail.com>
Reported-by: Fande Kong <fdkong.jd@gmail.com>

Add support for users to pass in FFLAGS the option for promoting integers to 8 bytes.

Tested for ifort and gfortran

Tested with MPICH and OpenMPI and MPI-Uni

Since MPI does not provide any mechani

Add support for users to pass in FFLAGS the option for promoting integers to 8 bytes.

Tested for ifort and gfortran

Tested with MPICH and OpenMPI and MPI-Uni

Since MPI does not provide any mechanism for integer promotion the code very crudely adjusts the
values in the mpif.h to mark them as 4 byte integers this is not neccessary portable for all MPI implementations

This is an extremely fragile option that we should probably not be supporting.

Users must insure their Fortran codes passes only 4 byte integers to MPI calls

PETSc Fortran examples were all fixed to explicitly use 4 byte integers for MPI calls.

Commit-type: portability-fix, testing-fix, feature
Time: 20 hours
Reported-by: Dmitry Melnichuk <dmitry.melnichuk@geosteertech.com>

show more ...

Merge branch 'stefanozampini/sfwin-reuse-win' into 'master'

Improve PETSCSFWINDOW

See merge request petsc/petsc!2336