PCBDDC/KSPFETIDP: don't scale the interface pressure solution

PCBDDC/KSPFETIDP: added code for interface pressure preconditioning

PCBDDC/KSPFETIDP: move extraction of off-diagonal saddle point matrices during PCISSetUp

This allows to use the change of basis in the inner BDDC

PCBDDC/KSPFETIDP: don't restrict to alldofs\setminus interfacepressure

just zero rows and columns of Amat relative to interface pressure
fix a couple of bugs with Pre/PostSolve flags

It works with

PCBDDC/KSPFETIDP: don't restrict to alldofs\setminus interfacepressure

just zero rows and columns of Amat relative to interface pressure
fix a couple of bugs with Pre/PostSolve flags

It works with incompressible Stokes with Taylor-Hood; there are still problems when using the change of basis

show more ...

PCBDDC: fix bug for nonsymmetric problems

Updated and simplified Fortran interface

Reduced from 4 distinct approaches to using PETSc from Fortran to 1
No more .h90 files (Fortran 90 functionality is now always available)
Parameters and func

Updated and simplified Fortran interface

Reduced from 4 distinct approaches to using PETSc from Fortran to 1
No more .h90 files (Fortran 90 functionality is now always available)
Parameters and function prototypes are now provide through Fortran modules, not includes;
only #define are provided in the petscXXXdef.h files
PETSc objects can be declared as either type(tXXX) or XXX. For example type(tVec) or Vec
Moved now private Fortran include files out of public include directory since uses don't use them;
they are only used to generate the Fortran modules
Easier to support and test; we can now easily add more function prototypes
Also had to rework generation of manual pages to consistently use MANSEC and SUBMANSEC
since these variables also handle how the Fortran modules are generated.

Time: 57 hours

show more ...

PCBDDC: added a couple of tests for the scatters of the interface pressure

PCBDDC: added support for saddle point problems when getting the RHS and computing the solution of the FETIDP problem

PCBDDC: fix from bad merge

PCBDDC: fix additional scatters for saddle point FETIDP

PCBDDC: fix creation of the FETIDP matrix

KSPFETIDP: add construction of relevant objects for saddlepoint problems in SetupFETIDPMatContext

PCBDDC: fix for unsymmetric problems

PCBDDC: added citation for PCBDDC

PCBDDC: allow specification of field id for pressures at command line

PCBDDC: unlock vecs of nullspace for no-net-flux and lock them again after setting the values

PCBDDC: check benign trick when dbgflag > 1

PCBDDC: remove consistency checks from PCBDDCGraph code

Check Index sets in PCBDDC code

PCBDDC: fix deadlock

PCBDDC: add robustness to the correct detection of malign/benign subdomains when we don't have enough information from the user

MatIS: attach local fields to the local matrix when converting from MatNest
PCBDDC: use the information to setup the dofs splitting

PCBDDC: first check validity of IS

PCBDDC: allow setting 1 when the matrix is missing the diagonal

PCBDDC: yet another deadlock fixed

PCBDDC: fix deadlock