Bug: someone assumed a PetscInt could always fit in the space for a PetscScalar, not true for fp16

Support for half-precision using __fp16.

Requires repository trunk of gcc and ARM systems
Configure with --with-precision=__fp16 --download-f2cblaslapack --with-fc=0 --with-cxx=0
Many standard PETSc

Support for half-precision using __fp16.

Requires repository trunk of gcc and ARM systems
Configure with --with-precision=__fp16 --download-f2cblaslapack --with-fc=0 --with-cxx=0
Many standard PETSc tests fail due to lack of precision

Funded-by: ECP
Project: fp16
Time: 24 hours

show more ...

windows: __FUNCT__ vs __func__ fix mismatch for icl

Merge branch 'stefano_zampini/allow-late-matsetblocksizes'

Remove newline character

Added support for MatMatMatMult_transpose_aij_aij

actually, boomerAMGBuildCoarseOperator needs Rt and not R to perform RAP.
Removed HYPRE support from MatMatMatMult_AIJ_AIJ_AIJ and added the compose

Added support for MatMatMatMult_transpose_aij_aij

actually, boomerAMGBuildCoarseOperator needs Rt and not R to perform RAP.
Removed HYPRE support from MatMatMatMult_AIJ_AIJ_AIJ and added the composed to the AIJ objects

show more ...

MatSeqAIJ and MatMPIAIJ: added support for MatMatMatMult using HYPRE

MatSeqAIJ and MatMPIAIJ: add support for MatMatMult via hypre

MatMPIAIJ: added support for MatPtAP with boomerAMGBuildCoarseOperator

MatSeqAIJ: added support for using boomerAMGBuildCoarseOperator for PtAP

Merge branch 'stefano_zampini/feature-pcbddc-saddlepoint'

Conflicts:
src/mat/impls/aij/mpi/mpiaij.c
src/mat/impls/is/matis.c

Modified ex23.c for the API change of the function CheckMat

Merge branch 'hzhang/opt-matgetsubmatrices-rebased-v2'

fix memory leak as reported by nightly tests

Add additional PETSC_NULL_XXX objects, PETSC_NULL_INTEGER etc need to be arrays, not scalars

Added more /= == comparisons of PETSc types
Problems found when updating PFlotran for latest PETSc Fortra

Add additional PETSC_NULL_XXX objects, PETSC_NULL_INTEGER etc need to be arrays, not scalars

Added more /= == comparisons of PETSc types
Problems found when updating PFlotran for latest PETSc Fortran features
Funded-by: IDEAS
Project: IDEAS/xSDK
Time: 2 hour

show more ...

fix comment

fix typo in makefile reported by nightly tests; rm a loop of size np

use rmax as size of subcols and subvals

replace MatGetRow_MPIAIJ() with direct access to c->A and c->B arrays
final cleanup

replace MatGetRow_MPIAIJ() with direct access to c->A and c->B arrays

rm all while (row >= C->rmap->range[proc+1]) proc++;

use r2pmap to replace while-loop for finding proc for each row

bugfix for rmap_loc

implement MatGetSubMatrices_MPIAIJ_single() to optimize MatGetSubMatrices_MPIAIJ() when ismax = 1

update src/mat/examples/tests/ex4.c for testing MatGetSubMatrices()

introduce ctable to remove nons

implement MatGetSubMatrices_MPIAIJ_single() to optimize MatGetSubMatrices_MPIAIJ() when ismax = 1

update src/mat/examples/tests/ex4.c for testing MatGetSubMatrices()

introduce ctable to remove nonscalable malloc C->rmap->N array for MatIncreaseOverlap_MPIAIJ_Once()

remove nonscalable d_p by using ctable

add MAT_SUBMAT_SINGLEIS to MatOption. Set this flag in PCSetUp_ASM() to
reduce MPI_AllReduce() in MatGetSubMatrices()

show more ...

Merge branch 'barry/fortran-update'

Remove the needed def extension from Fortran include files

Time: .3 hours
Thanks-to: Satish Balay <balay@mcs.anl.gov>