On 16 May, 02:41, Roman Pearce <rpear...@gmail.com> wrote: > On May 15, 6:21 pm, Bill Hart <goodwillh...@googlemail.com> wrote: > > > I have the right number of terms, but not quite the right coefficient, > > as of yet. This is a good test to help me dig out the bug. :-) > > Do you have a division routine? I divided f^100 by f to check the > result. This is one way I test sdmp. You can also plug in numbers.
I don't even have an addition function yet, so I basically can't test it very well until the module in flint grows a bit. This bug lay in a bit of the code that had basically never been run, so it wasn't surprising that it was broken. It'll be a while before I have a decent test function for that module. > > > By the way, is your computation running on more than one core? > > No, the polynomial f is too small to split up. We need at least 64 > terms to run two threads or the communication overhead is wasteful. > We try not to waste cpu cycles. > Then the times you are getting are once again extraordinary. That machine shouldn't be much faster than sage.math, and on that machine I get about 125s for the computation. But this time I really am relying on FLINT's fmpz module to do the arithmetic, and presumably there are some inefficiencies there. I could probably optimise for this case, e.g. everything is unsigned here. But there is probably some more fundamental reason it is running slow. Bill. > -- > To post to this group, send an email to sage-devel@googlegroups.com > To unsubscribe from this group, send an email to > sage-devel+unsubscr...@googlegroups.com > For more options, visit this group athttp://groups.google.com/group/sage-devel > URL:http://www.sagemath.org -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
