Hello, On 11/29/2012 01:03 PM, Frantisek Kluknavsky wrote:
I stumbled upon the same problem as http://savannah.gnu.org/bugs/?35367 . It affects only some architectures (32-bit i686, 64-bit s390x and PPC). Disabling FMA (%configure CFLAGS="$CFLAGS -ffp-contract=off") does not help.
OK, that's news for me.
Does this error indicate a serious bug/miscompilation or just an equation too difficult for the solver?
Algorithmically speaking, the order of the method should never change more than by one, except when the method is reinitialized. If it does, it is a sign of something being seriously wrong.
I would love to help with debugging.
It would be awesome if somebody could help sort this out, thanks! Here are the previous relevant discussion threads: http://lists.gnu.org/archive/html/bug-gsl/2011-11/msg00022.html http://lists.gnu.org/archive/html/bug-gsl/2011-12/msg00003.html -- [email protected] http://iki.fi/tuomo.keskitalo
