[petsc-users] Possible bug associated with '-pc_factor_nonzeros_along_diagonal' option?

[Zou, Ling]

Dear all,

For the same code and same input file, using '-pc_factor_nonzeros_along_diagonal' causing a seg fault, while '-pc_factor_shift_type nonzero -pc_factor_shift_amount 1e-8' works fine.
The seg fault happened after the very first residual vec norm was evaluated, somewhere before stepping into the second residual vec norm evaluation, i.e.,


    NL Step =  0, fnorm =  5.22084E+01

Segmentation fault: 11

I tend to believe that it is not in my `snesFormFunction` because I used print to make sure to see that the ‘sef fault’ is after snesFormFunction was called in the first residual vec norm was shown, and before snesFormFunction was called for the next residual vec norm.

I am behind ‘moose,’ so the debug version did not tell much info, showing the exact error message.
It’s not an urgent issue for me, since the other option worked fine.

Best,

-Ling
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20240313/91c872ff/attachment-0001.html>