Fix error in parallel loading of dense format matrix

Reported-by: Jose E. Roman <jroman@dsic.upv.es>

MatDENSE: fix string for blas calls

Infrastructure that allows failures in PCSetUp(), PCApply(), MatMult() etc due to, for example a zero pivot or function evaluation outside its domain, to propagate up and become a KSP_DIVERGED inste

Infrastructure that allows failures in PCSetUp(), PCApply(), MatMult() etc due to, for example a zero pivot or function evaluation outside its domain, to propagate up and become a KSP_DIVERGED instead of generating an error that stops the program. In response to Issue 96. This includes failures in MatCreateSNESMF() applications due to domain errors. The mechanism to propagate some errors is by setting Info or Nan into the output vector and using the norm or inner product reductions in SNES or KSP to propagate the error condition to all processes and then handling them immediately after the norm or inner product.

This allows, for example, ODE integrators to try again with a smaller time-step if the PCSetUp failed instead of requiring a complete restart of the run with other options.

Currently some error conditions, such as function domain error in a line search may not get progated up using the correct SNESConvergedReason

See src/snes/examples/tests/ex69.c for the handling of several conditions

show more ...

MatDENSE: this fix the case when MAT_REUSE_MATRIX is specified with a LDA different from the number of rows of the reused matrix.

Merge branch 'barry/fix-matload-mpidense-denseinfile'

Merge branch 'jose/mpidense-with-array'

should be max of all processors not sum

updated PETSc directory layout to match standard packaging strategies

include/petsc finclude,private,mpiuni
lib/petsc conf
bin/petsc*

Buffer space used for loading dense matrix from file was too small when first process didn't have largest part of the matrix
also simplified the MatLoad_MPIDense() code slightly

Mat: in MPIdense matrices avoid having to assemble in case of user-provided array

Merge branch 'barry/optimize-matshift'

Cannot assume that MatSetUp() has been called on dense matrix which has been passed
into MatLoad()

Optimized MatShift() for a completely empty matrix since this is commonly used with SLEPc to produce diagonal matrices

Reported-by: Jose Roman <jroman@dsic.upv.es>

MatLoad_MPIDense() always allocated array space even if it was previously provided

Contributed-by: Alejandro Lamas <alejandro.lamas@dsic.upv.es>

Merge branch 'maint'

MatLoad_SeqDense() always allocated new space for the data even if the user already provided it

Contributed-by: Alejandro Lamas Davina <alejandro.lamas@dsic.upv.es>

Merge branch 'stefano_zampini/seqaijtodense'

Merge branch 'barry/patch-flopcounts'

Conflicts:
src/mat/impls/dense/seq/dense.c

MatLoad_ and VecLoad_ routines that use options database information from the binary file .info files need to PetscViewerSetUp() before accessing the options database because the "lazy" setup of bina

MatLoad_ and VecLoad_ routines that use options database information from the binary file .info files need to PetscViewerSetUp() before accessing the options database because the "lazy" setup of binary viewers defered the opening of the files until binary reading was first done which was after the calls to determine matrix and vector block sizes.

show more ...

Prevent integer multiplication overflow for some flop logging

Reported-by: Stefano Zampini <stefano.zampini@gmail.com>

MatAssembly: move check for InsertMode consistency into MatStashScatter impls

The BTS implementation now encodes InsertMode in-band to avoid another
set of messages for MAT_SUBSET_OFF_PROC_ENTRIES.

MatAssembly: move check for InsertMode consistency into MatStashScatter impls

The BTS implementation now encodes InsertMode in-band to avoid another
set of messages for MAT_SUBSET_OFF_PROC_ENTRIES. It does the reduction
in debug mode for better error reporting.

show more ...

PetscViewerBinaryGetDescriptor() is now a collective and must be treated as such
fixed some calls to PetscViewerBinarySetUseMPIIO()
removed some test runs that take too long