PCBDDC: adaptive selection of constraints now works without explicitly changing the basis on the velocities

PCBDDC: change the basis during scaling/average operations

PCBDDC: start supporting no-net-flux condition as a constraint

PCBDDC: construct a fake change of basis that will be needed if we want to use deluxe scaling withouth change the basis on the pressures

The added code also goes in the direction of supporting a mix

PCBDDC: construct a fake change of basis that will be needed if we want to use deluxe scaling withouth change the basis on the pressures

The added code also goes in the direction of supporting a mix of user-defined constraints and adaptively generated primal spaces.

show more ...

PCBDDC: add runtime error. I still have to implement this.

PCBDDC: assign enough space for nonzeros in B_0

PCBDDC: if zerodiag_subs is not available from the reusage of the local solvers, compute it

PCBDDC: include primal vertices (if any) in the computation of the local Schur complements even if we are not doing adaption

Remove use of non-C89 features

Time: .05
Reported-by: nightly tests

A fix for fieldsplit pcs when using snes virs.

When reset pc fieldsplit saves a copy of the original IS. SNES VIRS then calls a new function, PCFieldSplitRestrictIS, and supplies
the inactive set t

A fix for fieldsplit pcs when using snes virs.

When reset pc fieldsplit saves a copy of the original IS. SNES VIRS then calls a new function, PCFieldSplitRestrictIS, and supplies
the inactive set to create new ISs for the fieldsplit. This is then reindexed and passed down to any nested fieldsplits.

show more ...

Make error message specific to exactly what matrix is supported

Reported-by: Rajeev Das <rajeevdas25@gmail.com>

Fix for PCReset_HYPRE to make it reentrant.

Reported-by: nightly testing

fixes for merge from maint patch

Merge branch 'master' into barry/add-pcreset-hypre-for-master

Needed to move PCDestroy_HYPRE() added to master to PCReset_HYPRE() since that was introduced
as a bug fix for maint

Added PCReset_HYPRE() so it may be used with SNESVI solvers

Funded-by: maintenance
Time: .3 hours
Reported-by: Cian Wilson (cwilson on bitbucket)

Merge branch 'barry/fix-log-summary-format'

Merge branch 'hzhang/inode-erroriffpe'

When converting from PetscViewerSetFormat() to PetscViewerPushFormat() I missed adding the PetscViewerPopFormat() in several places

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

Deprecate PetscViewerSetFormat() since it better to use PetscViewerPush/PopFormat() and using
PetscViewerSetFormat() can lead to confusing situations for users when later viewers do not recognize the

Deprecate PetscViewerSetFormat() since it better to use PetscViewerPush/PopFormat() and using
PetscViewerSetFormat() can lead to confusing situations for users when later viewers do not recognize the set
format and thus do not produce any information.

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

show more ...

Merge branch 'barry/Sylvain-gmres-symmetric'

fixed up formatting of references in manualpages to consistent style

References:
+ 1. - ref1
. 2. - ref2
- 3. - ref3

previously the formatting was varied and ugly

broken nonlinear solver table links fixed

Reported-by: Andrew McRae <A.T.T.McRae@bath.ac.uk>

Also made References: formatting consistent by always listing authors first

Add support for symmetric preconditioning with GMRES and PCSHELL

Contributed-by: Sylvain Mercier <sylvainmercier85@gmail.com>

enable zeropivot for pbjacobi, bs=3, baij matrix

PCBDDC: prevent from unneeded computations in PCPreSolve_BDDC with benign trick