Tzanio
6d2530775b
Extended copyright to 2022
2022-02-25 10:24:50 -08:00
Stefano Zampini
c31ded06c2
BDDC: attach RBMs to the global matrix
2022-02-14 12:46:02 +03:00
Tzanio
0cc104331c
Merge branch 'master' into yohann/mem/min-default-constr
2021-11-10 12:52:15 -08:00
blaz
4fe8308527
Small fix for petsc vectors
2021-09-22 15:01:54 -07:00
Yohann Dudouit
3d399d8eb3
Remove unnecessary Reset.
2021-06-25 11:16:09 -07:00
Stefano Zampini
30da55f38f
Pass dummy matrix to SNES at creation time
2021-06-12 11:58:45 +03:00
Aaron Fisher
212ce737bb
Removed preceding underscore variables in the examples and miniapps, and caught a couple I missed before.
2021-05-14 14:12:37 -07:00
Aaron Fisher
e758c37691
Removed nearly all leading underscores for variable names. Every once in a while one of these will confilct with standard libraries in an oddball compiler and it's best to just avoid this problem alltogether.
2021-05-10 16:06:51 -07:00
Tzanio
048b57dd9e
Merge branch 'master' into hypre-mixedint
...
Conflicts:
linalg/petsc.cpp
2021-04-21 11:25:44 -07:00
Stefano Zampini
b8afc860b2
Checked backward compatibility up to 3.8.4
2021-04-17 13:37:56 +03:00
Stefano Zampini
b041768098
Fix backward compatibility with 3.8
2021-04-11 14:23:22 +03:00
Julian Andrej
a9ea9cba15
Merge branch 'master' into hypre-mixedint
2021-04-09 12:00:35 -07:00
Stefano Zampini
41a96be0e5
MatAIJSetGenerateTranspose has been removed in 3.15 in favor of MatSetOption
...
minimum PETSc version for CUDA support is now v3.15
2021-04-01 14:45:31 +03:00
Stefano Zampini
37c80fdb9b
make style
2021-03-18 16:42:45 +03:00
Stefano Zampini
f28280bfee
Forgot to add the fixes from office workstation
2021-03-18 16:42:38 +03:00
Stefano Zampini
488e3f02a3
Support upcoming 3.15 release
2021-03-17 16:59:16 +03:00
camierjs
be0d8f69bc
make style
2021-03-17 16:59:16 +03:00
Veselin Dobrev
063e31521a
Add a method mfem::Device::GetId().
...
this way we can tell PETSc to use the same device.
2021-03-17 16:59:16 +03:00
camierjs
fcf145b699
Revert MatAIJCUSPARSESetGenerateTranspose in sync with PETSc/master
2021-03-17 16:59:16 +03:00
camierjs
444b1b2444
Move CUDA headers of linalg/sparsemat and use backends.hpp instead
...
MatAIJCUSPARSESetGenerateTranspose ⇒ MatSeqAIJCUSPARSESetGenerateTranspose
2021-03-17 16:59:16 +03:00
Tzanio
0bfd5dbaea
make style
2021-03-17 16:59:16 +03:00
Stefano Zampini
74147c743b
Fix warning
2021-03-17 16:59:16 +03:00
Stefano Zampini
7b0aa28753
make style
2021-03-17 16:59:16 +03:00
Stefano Zampini
d8f4fe81bc
Fix PetscODESolver callbacks in the matrix free case
2021-03-17 16:59:16 +03:00
Tzanio
0a392ec336
make style
2021-03-17 16:59:16 +03:00
Stefano Zampini
d571bdbdcf
PetscParMatrix: lazy setup for device (CUDA only for now)
...
Support CUDA for SLEPC
2021-03-17 16:59:16 +03:00
Stefano Zampini
1d38d3b286
PetscParMatrix::MatMultKernel do not scale input since it may be locked
2021-03-17 16:59:16 +03:00
Stefano Zampini
80ee3f2005
Remove usage of GetData() from PETSc interface
2021-03-17 16:59:16 +03:00
Tzanio
9b92be6f9b
make style
2021-03-17 16:59:16 +03:00
Stefano Zampini
d141bfbc72
Added H2 matrix solver
2021-03-17 16:59:16 +03:00
Stefano Zampini
da51141d93
Add support for PETSc Vectors and unassembled operators on GPU
2021-03-17 16:59:16 +03:00
Stefano Zampini
3e4ec69a53
PETSc: allow calling MatMultTranspose on a symmetric matrix
...
remowe wrap = false when reset
2021-03-17 16:58:45 +03:00
Jakub Červený
2949fa9563
WIP trying an approach where there is no distinction between element
...
order and degree. FESpace works with actual degrees, even for RT space.
The methods GetXDegree are removed, GetOrder is the same as GetElementOrder.
2021-02-26 16:02:21 +01:00
Jakub Červený
24630a4dcc
Replace GetOrder and GetFaceOrder in petsc.cpp
2021-02-19 19:05:05 +01:00
Tzanio Kolev
9825bda5c5
Merge pull request #1967 from mfem/slepc-eps-type-fix
...
Hide SLEPc's 'EPS' type in a namespace
2021-02-14 18:36:37 -08:00
Veselin Dobrev
4e5fb3ce0c
Fix doxygen warnings: use 'petsc::Vec' and 'petsc::Mat'
...
instead of 'Vec' and 'Mat' where needed for doxygen.
2021-02-08 14:59:25 -08:00
Adrien M. Bernede
04fb572359
Let's forget about 2020
2021-01-26 11:55:36 -08:00
Tzanio
d472fe16d9
Changes in the linalg/ directory before the mfem-4.2 release
2020-10-24 19:06:53 -07:00
Julian Andrej
238afadd05
more changes towards mixedint
2020-07-24 17:40:26 -07:00
Jean-Étienne Tremblay
bc40981ffe
Move PETSc error handling to shared header
2020-05-20 15:51:05 -07:00
Tzanio
419bd7c6d7
Some edits in linalg/
2020-03-09 20:01:11 -07:00
Tzanio
05125e19d5
Removed a space
2020-03-09 15:36:41 -07:00
Tzanio
ca55ef0217
The release number for mfem-4.1 is LLNL-CODE-806117
2020-03-02 17:18:46 -08:00
Tzanio
e689fd5b28
Merge branch 'master' into bsd-relicense
2020-02-27 09:29:25 -08:00
Tzanio Kolev
120f5a81fd
Merge pull request #1278 from mfem/stefanozampini/fix-dmshell
...
Small fixes when running a matrix-free code (MATSHELL)
2020-02-23 15:02:40 -08:00
Tzanio
2caa222edf
A number of changes to switch the code from LGPL-2.1 to BSD-3:
...
- Updated LICENSE, README, CHANGELOG and CONTRIBUTING.md
- Removed COPYRIGHT, added NOTICE
- Updated the headers in all source files, makefiles, etc.
2020-02-23 11:16:46 -08:00
Julian Andrej
73b14609e6
make style so tests can run
2020-02-03 10:49:40 -08:00
Stefano Zampini
2c737e779e
Fix MATSHELL support
...
it is conceptually wrong freeing the context in the MatDestroy routine,
since MatDuplicate and MatCopy can be called.
Revert back to use the operator in the context only
This does not happen with PC, since we do not have Duplicate/Copy operations
Provided support for MatCopy
2020-01-31 19:09:04 +03:00
Tzanio Kolev
e083fb443e
Merge pull request #1268 from mfem/stefanozampini/fix-td-bchandler
...
PetscBCHandler: fix for -snes_test_jacobian
2020-01-30 08:29:05 -08:00
Stefano Zampini
e2425eed55
PETSc and MFEM have a different approach in handling the Jacobians
...
PETSc reuses the objects, while MFEM usually gives back a new operator each time
When reusing the SNES and TS objects for different simulations,
the matrix type of the underlying DM is never update, nor is the representative matrix
for the DMSHELL (default). This is fine when using AIJ (default mattype in PETSc)
but not when using different types.
2020-01-30 17:19:25 +03:00
Tzanio