[petsc-dev] MatDestroySubMatrices_SeqAIJ and reference counting
Stefano Zampini
stefano.zampini at gmail.com
Fri Jan 12 03:56:10 CST 2018
Hong,
can you explain what’s the rationale behind calling explicitly the Layout and header destroy for the submit case in the loop here https://bitbucket.org/petsc/petsc/src/ac3af1b492556bac9856a6aee1c73992bd0b1779/src/mat/impls/aij/seq/aij.c?at=master&fileviewer=file-view-default#aij.c-2531 <https://bitbucket.org/petsc/petsc/src/ac3af1b492556bac9856a6aee1c73992bd0b1779/src/mat/impls/aij/seq/aij.c?at=master&fileviewer=file-view-default#aij.c-2531>
A code like this fails since you don’t take into account reference counting on the submatrices.
MatCreateSubMatrices(A,….&submats)
PetscObjectCompose(B,”_XXXX”,submats[0);
…..
MatDestroySubMatrices(..,submats);
MatDestroy(B); //Error, corrupt argument when trying to destroy composed objects
Here is a representative valgrind stack trace
==90133== Invalid read of size 4
==90133== at 0x100368F63: PetscCheckPointer (/Users/szampini/software/petsc/src/sys/error/checkptr.c:108)
==90133== by 0x1001ACA9F: PetscObjectDereference (/Users/szampini/software/petsc/src/sys/objects/inherit.c:564)
==90133== by 0x10019AC07: PetscObjectListDestroy (/Users/szampini/software/petsc/src/sys/objects/olist.c:154)
==90133== by 0x1001A8176: PetscHeaderDestroy_Private (/Users/szampini/software/petsc/src/sys/objects/inherit.c:115)
==90133== by 0x100626CAD: MatDestroy (/Users/szampini/software/petsc/src/mat/interface/matrix.c:1237)
==90133== Address 0x106c4c320 is 1,632 bytes inside a block of size 4,420 free'd
==90133== at 0x10014C9F3: free (in /usr/local/Cellar/valgrind/3.13.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==90133== by 0x100229B7A: PetscFreeAlign (/Users/szampini/software/petsc/src/sys/memory/mal.c:88)
==90133== by 0x10022D771: PetscTrFreeDefault (/Users/szampini/software/petsc/src/sys/memory/mtr.c:309)
==90133== by 0x100A7034D: MatDestroySubMatrices_SeqAIJ (/Users/szampini/software/petsc/src/mat/impls/aij/seq/aij.c:2549)
==90133== by 0x1006260BD: MatDestroySubMatrices (/Users/szampini/software/petsc/src/mat/interface/matrix.c:6957)
==90133== by 0x1006D82EC: _DestroyContainer
==90133== by 0x1001AFA31: PetscContainerDestroy (/Users/szampini/software/petsc/src/sys/objects/inherit.c:899)
==90133== by 0x1001ACBE5: PetscObjectDereference (/Users/szampini/software/petsc/src/sys/objects/inherit.c:566)
==90133== by 0x10019AC07: PetscObjectListDestroy (/Users/szampini/software/petsc/src/sys/objects/olist.c:154)
==90133== by 0x1001A8176: PetscHeaderDestroy_Private (/Users/szampini/software/petsc/src/sys/objects/inherit.c:115)
==90133== by 0x100626CAD: MatDestroy (/Users/szampini/software/petsc/src/mat/interface/matrix.c:1237)
==90133== Block was alloc'd at
==90133== at 0x10014C616: malloc (in /usr/local/Cellar/valgrind/3.13.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==90133== by 0x1002299AC: PetscMallocAlign (/Users/szampini/software/petsc/src/sys/memory/mal.c:48)
==90133== by 0x10022CA73: PetscTrMallocDefault (/Users/szampini/software/petsc/src/sys/memory/mtr.c:183)
==90133== by 0x10022B328: PetscMallocA (/Users/szampini/software/petsc/src/sys/memory/mal.c:396)
==90133== by 0x100C12E05: MatCreate (/Users/szampini/software/petsc/src/mat/utils/gcreate.c:89)
==90133== by 0x100B839F6: MatCreateSubMatrices_MPIAIJ_Local (/Users/szampini/software/petsc/src/mat/impls/aij/mpi/mpiov.c:2554)
==90133== by 0x100B7DEE7: MatCreateSubMatrices_MPIAIJ (/Users/szampini/software/petsc/src/mat/impls/aij/mpi/mpiov.c:2038)
==90133== by 0x10066703A: MatCreateSubMatrices (/Users/szampini/software/petsc/src/mat/interface/matrix.c:6806)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20180112/2c0112bb/attachment-0001.html>
More information about the petsc-dev
mailing list