On Dec 20, 2008, at 12:31 PM, Craig Citro wrote: > That example was with CyclotomicField(12) and CyclotomicField(132) ...
Ah. I bet the time was spent resolving the roots of CyclotomicField (132) to high enough precision to distinguish them. If you don't come up with a patch for this, I'll (probably) do it later tonight. - Robert > > -cc > > On Sat, Dec 20, 2008 at 12:25 PM, Robert Bradshaw > <rober...@math.washington.edu> wrote: >> >> On Dec 20, 2008, at 12:16 PM, mabshoff wrote: >> >>> On Dec 20, 12:11 pm, "Craig Citro" <craigci...@gmail.com> wrote: >>> >>> >>> <SNIP> >>> >>>> So just running this through prun suggests that it's something >>>> related >>>> to coercion: >>>> >>>> ncalls tottime percall cumtime percall filename:lineno >>>> (function) >>>> 17/11 13.884 0.817 14.137 1.285 >>>> number_field.py:1502(_coerce_map_from_) >>>> >>>> I'll have time to start looking into this later this afternoon, >>>> unless >>>> someone beats me to it ... >>> >>> Yep, a likely candidate is #4276, but that is a 100kb diff, so maybe >>> you should ask RoberWB for a hint. >>> >>> Note that in the modular/modform directory that is the only doctest >>> that got measurably slower, but I would not be surprised if >>> something >>> else in number fields would also show some performance differences. >>> This is another case where the timeit framework would come in handy. >>> Since it has been in for a while and basically unchanged >>> hopefully we >>> can run regression tests back a couple of releases. Getting >>> timeit to >>> compare different runs is definitely an excellent project for SD12 >>> should we not have it in place by then. >> >> Hmm... I think I know what causes this. Basically, to find a coercion >> between two cyclotomic fields, it finds all roots of each (by >> factoring the polynomial over the complex numbers) and matches up the >> closest one. For cyclotomic fields it could easily determine right >> away where the root goes. >> >> Do you have examples of the size of fields it's dealing with here? >> >> - Robert >> >> >>> >> > > > --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
