added MatMeshToVertexGraph() per Glenn Hammond request.Would it be possible to add a wrapper for:call ParMETIS_V3_Mesh2Dual(idxtype *elmdist, idxtype *eptr, idxtype*eind, int *numflag, int *ncom
added MatMeshToVertexGraph() per Glenn Hammond request.Would it be possible to add a wrapper for:call ParMETIS_V3_Mesh2Dual(idxtype *elmdist, idxtype *eptr, idxtype*eind, int *numflag, int *ncommonnodes, idxtype **xadj, idxtype**adjncy, MPI Comm *comm)I am using this routine to compute the graph for my unstructured gridprior to passing the graph to the partitioning algorithm. However, fromthe Fortran side, the arrays **xadj and **adjncy are kicking my tailsince they are allocated within ParMETIS through C. There are all kindsof kludges to employ as workarounds, but I figure it would be best tojust access the routine through PETSc as you have your own solutions forFortran/C addressing. What do you think?Hg-commit: 4f79adfeae65c2215d726c3eede11b29934ca09d
show more ...
Removed MatCompress completelyHg-commit: 4724a9ef7c2da8d62781764b00487fecf0216211
fixes for row/column sizes and maps for Mat MPIAdj, users have to pass number of local rows and global columnsHg-commit: 196f893bfa09d9d4a8621a4b2d97d138081a2ae9
bugfix for umfpack interface with complex precisiontrivial update comment of mpiaij.hHg-commit: 1d8de1a87dd559afd5229b88e9727c0a57b73f99
removed -I$PETSC_DIR from make system and changed the corresponding#include "src/dsdsdsds" to "../src/sdsdsds" in the source codeHg-commit: 6e3c4cc4b4bfcb366bae9b71afcb5009c1f0429d
fixed some memory bleedingchanged PetscMap in Vec and Mat to be an allocated shared structure with referencing counting. Shared by Vecs obtained with VecDuplicate() and MatGetVecs()Hg-commit
fixed some memory bleedingchanged PetscMap in Vec and Mat to be an allocated shared structure with referencing counting. Shared by Vecs obtained with VecDuplicate() and MatGetVecs()Hg-commit: 9cebf4ba1b4632fc12cb8afbc01d420ee612ad72
commit after most of merge of petsc-dev with petsc-dev-new-solvers with new way of handling external solversHg-commit: 7c86fa28bc8468b955cc6a62edd1de55d6c3d700
basic transistion to new model for external direct solvers all tests passHg-commit: 709a5887e4f17b074188a2f60d4f00cb8895dd52
moved the const out of XXType definitions and into the Set and Get MethodsHg-commit: d4d9a237045a151847ef3121b00df73d952d0999
removed no longer needed converts of Fortran to C MPI_Comm (since MPI 1.1 has a standard for this)Hg-commit: 263aca0fcfe7efcaed30a4abb46e648de3bb24ab
mv zpetsc.h (a meaningless name) to fortranimpl.h (a slightly more meaningful name)Hg-commit: d2687f12f92bff49ceff434100d018f6898829c3
mergeHg-commit: d507e9558c653b847e4e8b2a1687018f692e3610
petsc-maint 16799 ==6657== Conditional jump or move depends on uninitialised value(s) ==6657== at 0x8735AD1: petscoptionsgetstring_ (in /home/go
petsc-maint 16799 ==6657== Conditional jump or move depends on uninitialised value(s) ==6657== at 0x8735AD1: petscoptionsgetstring_ (in /home/gormo/fluidity/trunk/bin/petsc_readnsolve) ==6657== by 0x80F426D: petsc_readnsolve_options.878 (petsc_readnsolve.F90:264) ==6657== by 0x80F197C: petsc_readnsolve_ (petsc_readnsolve.F90:73) ==6657== by 0x80F18AD: main (petsc_readnsolve_main.cpp:28)moved zpetsc.h to private/zpetsc.hchanged Fortran interface function PetscOptionsGetXXX functions to allow passing PETSC_NULL_TRUTH as flag argumentHg-commit: a08dc4d2e15e97c0d3849b1cb02e6c1b0214b0c8
changes for solving the strict aliasing issue in PetscObject and derivedHg-commit: 278faecc91564f5495fa9f7d48380d547995fa45
new argument to MatSetOption() true or falseHg-commit: 229dc5e9b5c63fac6f09dcb7c27b358d7c5f1bf4
initial changes for better memory loggin on objectsHg-commit: 0ef7914e13d41d173ef1ba2552c26bd9f41bc807
changed bmake/common everywhere to conffixed a little more for new install systemHg-commit: 1aa781ff09e652ebafd6f016d84e88c15ea650f1
Update MatGetRowIJ() and MatGetColumnIJ() etc routines totake an additional argument blockcompressed. If this flagis set then the reduced matrix [baij/inode] is returned.This fixes hypre breakage
Update MatGetRowIJ() and MatGetColumnIJ() etc routines totake an additional argument blockcompressed. If this flagis set then the reduced matrix [baij/inode] is returned.This fixes hypre breakage with inode & MatGetRowIJ()Hg-commit: df52a525486d9327b0309b55f9268b8b74d8b888
add missing calls to PetscMapSetBlockSize in MatCreate_MPIAdjHg-commit: 3c7cb63c7d5d435d2f678fb8765018e1f950a635
revised the design of PetscMap....should fix petsc-maint 15893Hg-commit: c45a79cbc7d588b27e55b535c0cbb18777095c43
moved many implementation headers to include/privateHg-commit: aa99826eaac01bf46b6b3b0207d0ecd69d3393cb
bugfix: set matrix type as 0 after MatDestroy_xxx() is calledHg-commit: 863552fe268e86601b2450701133dbe07dd08765
added DMMGSetMatType() and -dmmg_mat_type <mtype>started symmetric version of BlockMat.added MatOptions[] to string match the options and renumbered the options from 0.Hg-commit: 9884893e09ba39
added DMMGSetMatType() and -dmmg_mat_type <mtype>started symmetric version of BlockMat.added MatOptions[] to string match the options and renumbered the options from 0.Hg-commit: 9884893e09ba39ad38482aeb4138851bf91cb680
Cleanup of MatConvert....Removal of special AIJ matrix typeMoved setting the matrix type_name from MatSetType() to each MatCreate_Pretty messy stuff, not tested for external packages probably l
Cleanup of MatConvert....Removal of special AIJ matrix typeMoved setting the matrix type_name from MatSetType() to each MatCreate_Pretty messy stuff, not tested for external packages probably lots of broken stuffHg-commit: 4d6c1df13f9fa0d93bc9d5bb964240852547fe7b
fix MatSetDiagonal_SeqAIJ for matrix missing (some) diagonal entriesget rid of unneeded calls to MatMarkDiagonal_xx should only be needed in assemblyor when creating a factored matrixHg-commit: b
fix MatSetDiagonal_SeqAIJ for matrix missing (some) diagonal entriesget rid of unneeded calls to MatMarkDiagonal_xx should only be needed in assemblyor when creating a factored matrixHg-commit: b8588dc3d1c171daa90229be953d2b7fd710d008
12345678910>>...15