[petsc-dev] MatMatMult crashes with MATMPIAIJ using one processor

[Ce Qin]

Dear all,

I encountered a problem that MatMatMult crashes using only one processor.
The matrix is MATMPIAIJ. I am using the master branch. The same program can
run in parallel without problem.

Valgrind gives the following stack trace:

==5543== Invalid read of size 8
==5543==    at 0xDC094CE: MatGetBrowsOfAoCols_MPIAIJ (mpiaij.c:4509)
==5543==    by 0xDC3FB95: MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable
(mpimatmatmult.c:218)
==5543==    by 0xDC45193: MatMatMult_MPIAIJ_MPIAIJ (mpimatmatmult.c:61)
==5543==    by 0xD9C657D: MatMatMult (matrix.c:9450)

==5543==
==5543== Invalid read of size 4
==5543==    at 0xDC094D7: MatGetBrowsOfAoCols_MPIAIJ (mpiaij.c:4510)
==5543==    by 0xDC3FB95: MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable
(mpimatmatmult.c:218)
==5543==    by 0xDC45193: MatMatMult_MPIAIJ_MPIAIJ (mpimatmatmult.c:61)
==5543==    by 0xD9C657D: MatMatMult (matrix.c:9450)

==5543==
==5543== Invalid read of size 4
==5543==    at 0xDC0793B: MatGetBrowsOfAoCols_MPIAIJ (mpiaij.c:4513)
==5543==    by 0xDC3FB95: MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable
(mpimatmatmult.c:218)
==5543==    by 0xDC45193: MatMatMult_MPIAIJ_MPIAIJ (mpimatmatmult.c:61)
==5543==    by 0xD9C657D: MatMatMult (matrix.c:9450)

==5543== Invalid read of size 4
==5543==    at 0xDC0794E: MatGetBrowsOfAoCols_MPIAIJ (mpiaij.c:4520)
==5543==    by 0xDC3FB95: MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable
(mpimatmatmult.c:218)
==5543==    by 0xDC45193: MatMatMult_MPIAIJ_MPIAIJ (mpimatmatmult.c:61)
==5543==    by 0xD9C657D: MatMatMult (matrix.c:9450)

This problem may be related to this thread
http://lists.mcs.anl.gov/pipermail/petsc-users/2014-September/022864.html.
If you need an example that reproduces this problem, please let me know.

Best,
Ce Qin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20170311/a8d22ef1/attachment.html>