cleanup

fix c++ comments

cleanup

bugfix for MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(): incorrect estimate of Crmax

fix complains reported by nightly tests

MatMUMPS: added API to set info on Schur complement using MUMPS convention

When A is just symmetric and we eliminate all the dofs accounting for the negative eigenvalues, S is positive
Like P2/P0 St

MatMUMPS: added API to set info on Schur complement using MUMPS convention

When A is just symmetric and we eliminate all the dofs accounting for the negative eigenvalues, S is positive
Like P2/P0 Stokes eliminating the pressures

show more ...

Merge branch 'jed/mat-assembly-perf' of bitbucket:petsc/petsc

VecAssembly and MatAssembly now use a scalable exchange pattern based on
PetscCommBuildTwoSided. This feature can be controlled with th

Merge branch 'jed/mat-assembly-perf' of bitbucket:petsc/petsc

VecAssembly and MatAssembly now use a scalable exchange pattern based on
PetscCommBuildTwoSided. This feature can be controlled with the options

-vec_assembly_bts 0 or 1 (default 0)
-matstash_bts 0 or 1 (default 0)

The rationale is that the new implementation with scalable data
structures can be slightly slower than the old version at small process
counts. The default here could be changed to depend on the process
count (leading to possibly-confusing scaling performance diagnostics) or
the implementation could learn to take a fast path.

* 'jed/mat-assembly-perf' of bitbucket:petsc/petsc: (49 commits)
MatStash: fix -Wsign-compare by using size_t for loop index when max is also size_t
Sys BuildTwoSided test: fix for non-POD std::complex
MatStash: cast to satisfy non-structural MPI type tag check
mpiuni: fix compile error /sandbox/petsc/petsc.clone-2/arch-linux-uni/lib/libpetsc.so: undefined reference to `MPI_Type_create_resized'
Vec: Silence compiler warning
Vec: fix typo in comment
Sys: fix C89 compiler warning
VecStash BTS: fix block stash InsertMode accounting
VecStash BTS: fix indexing bug counting sends to rank 0
MatStash BTS: fix memory leak on MAT_SUBSET_OFF_PROC_ENTRIES
VecAssemblyEnd_MPI_BTS: fix donotstash code path
VecAssemblyEnd_MPI_BTS: fix C++ conversion to InsertMode
Sys: fix datatypes test using MPI_Type_create_resized
MatStash BTS: work around lack of offsetof() for non-POD (std::complex)
MatAssembly: move check for InsertMode consistency into MatStashScatter impls
MatStash BTS: fix memory leak on reassembly with MAT_SUBSET_OFF_PROC_ENTRIES
MatStash BTS: small simplification to row ownership calculation
MatStash BTS: add MAT_SUBSET_OFF_PROC_ENTRIES, impl with neighbor-only comm
MatStash: initial BTS (BuildTwoSided) implementation
MatStash: add extension point for new BTS implementation
...

show more ...

cleanup

MatDuplicate_MPIAIJ_MatPtAP_new() works now

cleanup

Error in MatGetSubMatrix_MPIAIJ() MatGetOwnershipRangeColumn() final argument is one plus last location not the length

Reported-by: Danqing Wu <wuda@mcs.anl.gov>

cleanup and optimization of MatPtAPSymbolic_MPIAIJ_MPIAIJ_new()

incorrectly used stride value instead of local size in check that ISSTRIDE has same layout as columns of matrix

optimize

small optimize

small cleanup

rm redundant computation

continue optimize symbolic PtAP

rm previous AP from MatPtAPSymbolic_MPIAIJ_MPIAIJ_new()

cp old MatPtAPSymbolic into xxx_new for debugging memory leak

check for PetscMPIInt overflow in ISAllGather()

Though this routine is not scalable it is used in multiple places in PETSc and maybe used
on problems large enough to produce overflow of PetscMPIInt(

check for PetscMPIInt overflow in ISAllGather()

Though this routine is not scalable it is used in multiple places in PETSc and maybe used
on problems large enough to produce overflow of PetscMPIInt() when running with 64 bit PETSc integers.

show more ...

made MatGetSubMatrix_MPIAIJ scalable if one is requesting certain rows and ALL columns
This is useful for some fieldsplit preconditioners, for example in Xolotl

We need to product all the CUDA code from the spinlock definitions since the CUDA compiler
crashes on them

MatPtAPSymbolic_MPIAIJ_MPIAIJ_new() works, but has mem leak :-(

optimize