On Thu, Jan 26, 2012 at 20:24, Mark F. Adams <mark.adams at columbia.edu>wrote:
> No, I can't imagine how, but there are several moving parts here. I'm not > sure about this issue. > > I did start getting an error on a malloc in MatSetValues - this is new > right? > Yes, this is an experiment, we are setting MAT_NEW_NONZERO_ALLOCATION_ERR by default if you preallocate. You can set the option to false if you are intentionally not preallocating correctly. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120126/315a7677/attachment.html>
