| 1e0aba96 | 16-Mar-2018 |
Matthew G. Knepley <knepley@gmail.com> |
Plex ex13: Updated test output |
| 92c96254 | 16-Mar-2018 |
Matthew G. Knepley <knepley@gmail.com> |
Plex ex12: Updated test output |
| 8930a371 | 16-Mar-2018 |
Matthew G. Knepley <knepley@gmail.com> |
Plex ex1: Updated test output |
| 199a4585 | 14-Mar-2018 |
Lisandro Dalcin <dalcinl@gmail.com> |
Merged in dalcinl/fix-dmplex-vecview (pull request #884)
DMPlex: Fix VecView
Approved-by: Karl Rupp <me@karlrupp.net> Approved-by: BarryFSmith <bsmith@mcs.anl.gov> |
| d172c84b | 06-Mar-2018 |
Matthew G. Knepley <knepley@gmail.com> |
Plex+FEM: Fixes for reorg of vector spaces which I missed - Fixed DMPlexComputeInterpolatorNested() field offsets - Fixed DMPlexComputeInjector() |
| e9cb465c | 01-Mar-2018 |
Matthew G. Knepley <knepley@gmail.com> |
Plex: Fixes for ASCII viewer |
| b81cf158 | 12-Feb-2018 |
Matthew G. Knepley <knepley@gmail.com> |
Plex: Better error message |
| 741db25d | 13-Mar-2018 |
Toby Isaac <tisaac@cc.gatech.edu> |
DMField: empty processor protections |
| d918548e | 13-Mar-2018 |
Toby Isaac <tisaac@cc.gatech.edu> |
DMField regression test fixes |
| 83e10cb3 | 12-Mar-2018 |
Lisandro Dalcin <dalcinl@gmail.com> |
DMPlexDistributeLabels: Propagate the output flag |
| fa8e8ae5 | 12-Mar-2018 |
Toby Isaac <tisaac@cc.gatech.edu> |
DMLabel: try to keep stride IS as stride IS as much as possible |
| 2dc8f7eb | 12-Mar-2018 |
Karl Rupp <me@karlrupp.net> |
Merge branch 'haplav/fix-PetscSFView-no-graph-set' [PR #878]
* haplav/fix-PetscSFView-no-graph-set: PetscSFView: call PetscSFSetUp only if graph is set. |
| 6d7373e8 | 12-Mar-2018 |
Toby Isaac <tisaac@cc.gatech.edu> |
DMPlexRefine(): create depth before hybrid-bounded label |
| aeadca18 | 26-Feb-2018 |
Toby Isaac <tisaac@cc.gatech.edu> |
DMPlex: rename newly created label |
| 4a3e9fdb | 21-Feb-2018 |
Toby Isaac <tisaac@cc.gatech.edu> |
DMPlexSNES: caching of new geometry constructs
The old paradigm cached one geometry object for all cells or all faces, an approach that doesn't take into account that different quadratures may requi
DMPlexSNES: caching of new geometry constructs
The old paradigm cached one geometry object for all cells or all faces, an approach that doesn't take into account that different quadratures may require different geometry objects. Now the geometry object is composed with the index set over which the geometry ranges, and the composition string include the object id of the quadrature. To take advantage of this caching, we have to be careful not to reconstruct index sets on the fly when they are known a priori. In particular, the hybrid bounds have been encapsulated as a label, which makes them accessible as ISs.
show more ...
|
| 034e9169 | 09-Mar-2018 |
Karl Rupp <me@karlrupp.net> |
Merge branch 'pr874/mapdes/pcmg-inject/master'
* pr874/mapdes/pcmg-inject/master: This hooks everything up so that if the DM provides an injection we remember it in PCMG. * Add DMHasCreateInjection
Merge branch 'pr874/mapdes/pcmg-inject/master'
* pr874/mapdes/pcmg-inject/master: This hooks everything up so that if the DM provides an injection we remember it in PCMG. * Add DMHasCreateInjection: Will be needed in PCMG. * FAS: guard DMCreateInjection with DMHasCreateInjection * PCMG: Use injection if the DM provides it
The mass-lumped scaling of the restriction is complete nonsense for most discretisations. If the DM provides an injection, use it. This injection is now used by DMRestrictHook_SNESVecSol for transferring the current state to the coarse grid.
show more ...
|
| 80153354 | 06-Mar-2018 |
Vaclav Hapla <vaclav.hapla@erdw.ethz.ch> |
PetscSFView: call PetscSFSetUp only if graph is set |
| c2f0e8ad | 06-Mar-2018 |
Vaclav Hapla <vaclav.hapla@erdw.ethz.ch> |
DMPlex tutorial ex5 activated and reproduce the bug |
| 684b87d9 | 06-Mar-2018 |
Lisandro Dalcin <dalcinl@gmail.com> |
DMPlex: Fix VecView
* Fix VTK viewer requiring exclusive ownership of the vector. * Add fast path for HDF5 with global vectors and essential BCs. |
| 713d6c57 | 07-Mar-2018 |
Karl Rupp <me@karlrupp.net> |
Merge branch 'pr838/haplav/haplav/feature-petscpartitionermatpartitioning-rb/master'
* pr838/haplav/haplav/feature-petscpartitionermatpartitioning-rb/master: Implemented wrapper of MatPartitioning i
Merge branch 'pr838/haplav/haplav/feature-petscpartitionermatpartitioning-rb/master'
* pr838/haplav/haplav/feature-petscpartitionermatpartitioning-rb/master: Implemented wrapper of MatPartitioning into PetscPartitioner.
show more ...
|
| 4a7a4c06 | 06-Mar-2018 |
Lawrence Mitchell <lawrence.mitchell@imperial.ac.uk> |
Make DMHasCreateInjection overrideable by subclasses
Default to PETSC_FALSE. It's only provided by Plex, Pforest, and DA in (if using DMDA_Q1 as interpolation type). |
| c8dd51e9 | 05-Mar-2018 |
Barry Smith <bsmith@mcs.anl.gov> |
Revert "Bug in viewing multiple vectors with VecView_Plex_Local_Draw() each overwrote the previous"
This reverts commit 29c95491dc1710544f73d25ac98dc64225f2d7e1. |
| 632d4749 | 05-Mar-2018 |
Barry Smith <bsmith@mcs.anl.gov> |
Revert "Follow fix for commit 29c95491dc1710544f73d25ac98dc64225f2d7e1 since object is already referenced now"
This reverts commit 7762cb5a66803d54d0f43ea8d2d5b956639305c8. |
| 6783bb9d | 05-Mar-2018 |
Karl Rupp <me@karlrupp.net> |
Merge branch 'pr871/acroucher/acroucher/dmplex-hdf5-vecview/master'
* pr871/acroucher/acroucher/dmplex-hdf5-vecview/master: Two fixes to prevent HDF5 DMPlex VecView crashing when one process has no
Merge branch 'pr871/acroucher/acroucher/dmplex-hdf5-vecview/master'
* pr871/acroucher/acroucher/dmplex-hdf5-vecview/master: Two fixes to prevent HDF5 DMPlex VecView crashing when one process has no cells.
show more ...
|
| 72de04e6 | 05-Mar-2018 |
Jed Brown <jed@jedbrown.org> |
Merge branch 'bourdin/dmplex-exodusiiView' of bitbucket:bourdin/petsc-dmplex-exodusiiview [PR #863]
* 'bourdin/dmplex-exodusiiView' of bitbucket:bourdin/petsc-dmplex-exodusiiview: BuildSystem: bug
Merge branch 'bourdin/dmplex-exodusiiView' of bitbucket:bourdin/petsc-dmplex-exodusiiview [PR #863]
* 'bourdin/dmplex-exodusiiView' of bitbucket:bourdin/petsc-dmplex-exodusiiview: BuildSystem: bug fixes in pnetcdf dm: changed DMGetLabelIdIS to return an empty IS instead of NULL if the label is empty BuildSystem: upgraded exodusII, removed fortran library dmplex: bug fixes in exodus viewer, write side sets
show more ...
|