On Wed, Oct 26, 2011 at 22:44, Paul Sicsic <paul.sicsic at polytechnique.edu>wrote:
> Can you show the value of ark->shift and arkimex.c:713, >> > It is inf > (gdb) p ark->shift > $3 = inf > Something must have computed a step size of 0 (or we're doing the wrong thing with a method that has explicit stages; we are changing some things with that). Of the recent methods that we have been playing with, I would recommend starting with "-ts_type rosw -ts_rosw_type ra3pw" or "-ts_type rosw -ts_rosw_type ra34pw2", both of which have nice stability properties, good error estimates, and are more permissive than the ARKs about how you write your equations. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111026/c58c1e74/attachment.html>
