Fix null output from PCFactorGetMatSolverPackage();Reported-by: Jose E. Roman <jroman@dsic.upv.es>
Fixes for issues found with clang static analyzer,mostly set but unused variablesTime: 1 hour
update to Superlu_dist with different struct name than superlurename superlu_dist_options_tFunded-by: IDEASProject: xSDKTime: .1 hours
Merge branch 'master' into barry/fix-private-functions
More decreasing visibility of symbols from outside librariesAlso fix a few bugs related to having multiple PETSc libraries
Fix for freeing just what needs to be free when SuperLU_Dist factorizations reusedFunded-by: IDEASProject: xSDKTime: .5 hours
set 'ReplaceTinyPivot=NO' as default. See Sherry's email Re: zeropivot in superlu_dist on Jan 7, 2016 at 12:03 AM
support !A->erroriffailure
replace default FactPattern = SamePattern_SameRowPerm to SamePattern;see Sherry's email Re: SuperLU convergence problem (More test) on 12/8/2015
bugfix when mpiaij matrix has empty diagonal block -- see bug report from Anthony Haas, Aug 10, 2015
add warning/error flag for '-mat_superlu_dist_parsymbfact' when MatInputMode=GLOBAL
updated manual pages for external direct solvers to explicitly state how to configure PETSc for them and to run with them with KSP
updated PETSc directory layout to match standard packaging strategiesinclude/petsc finclude,private,mpiunilib/petsc confbin/petsc*
superlu_dist: initial port to superlu_dist_4.0- requires c99 compiler to build - so add a configure check for c99- update sources with mc64id_ -> mc64id_dist, and LUstructInit() changes.* still ha
superlu_dist: initial port to superlu_dist_4.0- requires c99 compiler to build - so add a configure check for c99- update sources with mc64id_ -> mc64id_dist, and LUstructInit() changes.* still has conflict with superlu [mc64dd_,mc64ed_,mc64fd_]
show more ...
manual merge of conflict
add error flags for MatGetDiagonal when user calls it for matrix factor of mumps, superlu and superlu_dist
properly extern "C"ing MatSolverPackage registration functions
Moved registration of MatSolverPackages out of MatCreate_XXX routines into standalone system, this allows1) more precise error message when a requested solver is not found2) easier to add new solve
Moved registration of MatSolverPackages out of MatCreate_XXX routines into standalone system, this allows1) more precise error message when a requested solver is not found2) easier to add new solvers without touching basic PETSc matrix implementation files and the possibility of adding new solvers to the library after PETSc is installed or even at run time
update use of PetscOptionsBool() to not rely on default setting automatically
Sys: drop explicit type arguments from PetscNew() and PetscNewLog()git grep -l 'PetscNew\(Log\)\?(' -- $1 | xargs perl -pi -e 's@PetscNew\([^,;()]+ *, *@PetscNew(@;s@PetscNewLog\(([^,;()]+) *,[^,
Sys: drop explicit type arguments from PetscNew() and PetscNewLog()git grep -l 'PetscNew\(Log\)\?(' -- $1 | xargs perl -pi -e 's@PetscNew\([^,;()]+ *, *@PetscNew(@;s@PetscNewLog\(([^,;()]+) *,[^,;()]+, *@PetscNewLog($1,@'
since external packages use a variety of different types for 64 bit indices(though all have the same bit values) casts are needed for some compilers toprevent warning messages when using 64 bit ind
since external packages use a variety of different types for 64 bit indices(though all have the same bit values) casts are needed for some compilers toprevent warning messages when using 64 bit indices
Merge remote-tracking branch 'origin/barry/remove-timelogging'
fix incorrect caste to superlu_dist int_t type when using 64 bit indices
no we don't put calls to PetscTime() randomly in the code, that is what the logging is for
-citations option causes the print out of approriate bibtex citation indices for all packages USED in the runMost of the appropriate citations still need to be added
12345678910>>...13