Ilya Zakharevich on Sun, 23 May 2004 21:33:47 +0200


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: gp: anal.c more bugs


On Sun, May 23, 2004 at 12:48:25AM +0200, Bill Allombert wrote:
> > This is a build option now.  And anyway, this is chicken and egg: when
> > the high-level stuff is available through an installed library, this
> > option has a chance to become a default.  Right now such a build is
> > very crippled.

> You can build a gp.so which is linked against the libpari.so.
> We cannot provide a rigid interface for GP internals anyway.

I do not care about GP at all (at least when Math::Pari is
considered).  What is required is the part of the library which is (by
unfathomable reasons) not installed (the graphing and highlevel stuff).

> > I think that Math::Pari is about 3 orders of magnitude quickier that
> 
> Are you sure GMP support was enabled ? (Math::BigInt is part of the
> base perl install, but the GMP support is optionnal and need to
> be installed separately).

I saw internals of the (new) Math::BigInt, and it has very little
chance to be quick - no matter what the backend is.

> Does they use more than modular arithmetic ?

I could not find anything else used...

> > Math::BigInt.  [However, what is a better option should be determined
> > not only by speed, but by other factors too.  Myself, I'm puzzled by
> > the fact that Math::Pari is made a prerequisite...]

> I agree, I always found that was strange.

Well, this is a good testbed for PARI.  ;-)

Yours,
Ilya