CI: arch-ci-linux-gcc-complex-opt.py -> arch-ci-linux-gcc-complex-opt-32bit.py

Fix bug in ex4.c Code assumed that floating point product would produce exact result

/spend 3h
Reported-by: Gitlab CI

Create affine correction for total solution projection in plex

squashme: fix DMInterpolationSolution_Plex

squashme: fix DMInterpolateSolution_Plex()

squashme whitespace

Use fields rather than ds's to get discretizations when interpolating

Interpolation shouldn't depend on the equations just the fields

squashme: fix interpolator for unrefined

DMPlex: Some fixes to the CGNS reader

Merge remote-tracking branch 'origin/release' into main

DMPlex: missing PetscSFSetUp()

SF: rename SFCreateEmbeddedSF to SFCreateEmbeddedRootSF

SF: add an MPI_Op argument to SFBcast

Merge branch 'barry/2021-02-02/petscsf-communication-specific' into 'main'

PetscSFView() never called viewer for the specific type (bug), hence many...

See merge request petsc/petsc!3606

PetscSFView() never called viewer for the specific type (bug), hence many output files were incorrect.

Added PetscSFView() support for basic MPI implementation and binary viewer saves communication

PetscSFView() never called viewer for the specific type (bug), hence many output files were incorrect.

Added PetscSFView() support for basic MPI implementation and binary viewer saves communication data to binary file.
This new viewer shows exactly how much data is sent from each rank to the other ranks
allowing fiting various performanace models to the data.

This uses MatView() to save the PetscSF data, but since Mat is above PetscSF in the software stack
the code will not work with multiple PETSc libraries. How should this be managed if one wishes to preserve the
traditionl PETSc algorithm stack?

Commit-type: bug-fix, feature
Funded-by: brain project
Project: brain project
spend 1h

Update sfbasic.c save the transpose of the matrix since that indicates the sends for each rank which is easier to understand

Enable PetscSFView_Basic_PatternAndSizes only with single library

show more ...

git: switch default branch from 'master' to 'main'

checkbadSource: new rule for wrongfully placed "{"

Merge remote-tracking branch 'origin/release'

CHKERRQ() -> CHKERRMPI()

Fixed doc reference

DM: Document internal labels - Propagate Plex printing flags

Merge remote-tracking branch 'origin/release'

KSP: Reorganized monitors - Now the residual can be viewed using -ksp_monitor draw and a plot can be made using -ksp_monitor draw::draw_lg - This also works for the true residual and error - We no

KSP: Reorganized monitors - Now the residual can be viewed using -ksp_monitor draw and a plot can be made using -ksp_monitor draw::draw_lg - This also works for the true residual and error - We noe register monitors so that dispatch based on viewer and format works - Add KSPMonitorLGCreate(), and remove KSPMonitorLGResidualNorm*() and KSPMonitorLGTrueResidualNorm*() - Add KSPMonitorError(), used by -ksp_monitor_error - Add arguments to KSPMonitorSetFromOptions() to allow line graphs to be configured - Remove SNESMonitorLGCreate() and SNESMonitorLGResidualNorm() which are now handled by the default monitor - PetscViewerAndFormat now allows a payload - VecView_Plex() and VecView_Plex_Local() must propagate the flag indicating zero BC

show more ...

DMPlexVTKWriteAll_VTU: a couple of bugfixes

- Allow calling VecView on a local Vec (requires passing the comm to TransferWrite)
- Fix the case where some processes have zero cells

Plex: Add PETSc git version to HDF5 as attribute

plex_hdf5: add output files for plex/tutorials/ex5

plex_hdf5: add framework to test load on N / save on M

ComponentName: modification of DMPlexCreateGlobalToNaturalSF

The modification of DMPlexCreateGlobaltoNaturalSF allows to pass NULL as an argument
for the sequential section when calling the routine.

ComponentName: modification of DMPlexCreateGlobalToNaturalSF

The modification of DMPlexCreateGlobaltoNaturalSF allows to pass NULL as an argument
for the sequential section when calling the routine. The interface hasn't change. It takes
into account that a DM can be distributed with DMPlexDistribute just after its creation.
In that case, sfNatural cannot be computed in DMPlexDistribute since no section is defined on the DM
at that moment. The user should then call DMPlexCreateGlobaltoNaturalSF after a section is set to the
parallel DM in order to compute the sfNatural. Instead of computing the sequential section needed for
DMPlexCreateGlobaltoNaturalSF, he can now pass NULL as an argument for the section and it is reconstructed
from the parallel section in the routine.
Error messages are slightly modified in all routines of plexnatural.c when the naturalSF is not present and
useNatural is set to TRUE. It mentioned now that a section has to be defined for the naturalSF to be computed.
TODO: since DMPlexCreateGlobaltoNaturalSF is not typically called by the user, automatically call it as soon
as useNatural is TRUE, the DM has been distributed and a section is present.
No failed test when running all dm tests locally after the modification.

Reported-by: Blaise Bourdin <bourdin@lsu.edu>
Thanks-to: Blaise Bourdin <bourdin@lsu.edu>

show more ...

Plex+FE: Now we can turn on second derivatives in an FEM calculation - Now FE interpolation and PetscFEEvaluateFieldJets_Internal() can also do Hessians - Add jet degree to DS - Add PTAP operations f

Plex+FE: Now we can turn on second derivatives in an FEM calculation - Now FE interpolation and PetscFEEvaluateFieldJets_Internal() can also do Hessians - Add jet degree to DS - Add PTAP operations for small matrices

show more ...

Plex: Fixed serial fallback for GlobalToNatural