That should say mutivariable polynomial multiplication, not arithmetic.
Bill.
--- On Mon, 1/26/09, Bill Hart <goo...@go...> wrote:
> From: Bill Hart <goo...@go...>
> Subject: [flint-devel] Multivariate polynomial arithmetic
> To: "Development list for FLINT" <fli...@li...>
> Date: Monday, January 26, 2009, 6:12 AM
> I've implemented a basic algorithm for doing
> multivariate polynomial
> arithmetic (I'm thinking of including a module for this
> in FLINT 2.0 -
> due out towards the end of the year).
>
> It computes Fateman's bechmark f*g where f = d^20, g =
> d^20 + 1 with d
> = (1+x+y+z+t) in 23.7s on a 2.66GHz Xeon. Pari takes about
> 38.5s and
> it's probably comparable if not better than Magma.
>
> Next I'll work on improving the cache locality of the
> algorithm.
>
> The real times to try and equal are those of Trip and sdmp
> which can
> both do it in about 2.5s. Trip is/will be open source I
> believe. sdmp
> is not as far as I know. We don't really need to
> compete with either
> as they aren't aimed at the same sorts of problems as
> FLINT (Trip is
> used for Celestial Mechanics and sdmp is available only as
> a binary to
> link against Maple and only performs a handful of
> operations). But it
> would be nice to have times vaguely comparable with theirs.
> Another
> nice open source package for doing this sort of thing is
> giac. I don't
> have an idea of times for that yet.
>
> I haven't committed the code yet. But presuming someone
> wanted to
> examine it, I would clean it up and commit it.
>
> Bill.
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> flint-devel mailing list
> fli...@li...
> https://lists.sourceforge.net/lists/listinfo/flint-devel
|
