On Saturday 14 March 2009 15:04:09 Bill Hart wrote: > On 14/03/2009, Jason Moxham <ja...@njkfrudils.plus.com> wrote: > > On Saturday 14 March 2009 13:46:51 Bill Hart wrote: > >> I've been working on converting Paul Zimmermann's implementation of > >> the Bodrato-Zanoni Toom-4 algorithm to the mpn_level. > >> > >> I've uploaded the current code to the files section. I'm pretty much > >> just going to convert the relevant functions from FLINT's fmpz module > >> (in the file fmpz.c - also uploaded) into tc4_blah functions as > >> required, much as I have done with fmpz_add->tc4_add. > >> > >> I thought I would upload it in case anyone felt like working on it > >> today, as I will not be able to work on it for most of the rest of the > >> day. > > > > When are we releasing MPIR 1.0.0 ? > > It's always better to release on a weekday, thus I plan to update > documentation tomorrow and release Monday morning. Feel free to do the > docs and tarball for a release candidate while I am away today, but > don't forget to touch .c and .h files in demos/calc and remove .svn > dirs. Actually there is a script in my sage.math home directory called > makempir. >
make dist nearly works , I'll see if I can get it to go > >> The tc4_addlsh1 function has no analogue in FLINT. I think in every > >> case it is used in Toom-4 it only deals with unsigned integers, which > >> makes it easier to implement. > > > > Is this the usual mpn_addlsh_1 ? > > Not quite. It must also deal with operands of different lengths. > > >> Also, the mpn_divexacts, by 12, 18, 30, etc could probably be done > >> efficiently in assembly. > >> > >> for example div_exact by 30 is just mul by (2^B - 1)/15, then divide > >> by (2^B - 1) and shift right by 1. > > > > I allready have divexact by any factor of 2^B-1 , runs at 2.2c/l on K8 , > > just > > have to squeeze some shifts in too, or somehow combine it with the mul > > Cool! > > >> Paul's original implementation is here: > >> > >> http://www.loria.fr/~zimmerma/software/toom4.c > >> > >> Note it is LGPL v2+. Thanks Paul!! > >> > >> I now believe a very efficient implementation of this could make quite > >> a difference. > > > > Nice one :) > > > >> Bill. > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en -~----------~----~----~----~------~----~------~--~---
