Re: dev version of pari-gp (was: Bug#272886: libpari1: wrong error message for large primelimit)
On Wed, Sep 22, 2004 at 09:41:44PM +0200, Stephan Niemz wrote:
> On Wed, Sep 22, 2004 at 18:35:18 +0200, Bill Allombert wrote:
> > This implementation store differences between consecutive prime numbers
> > in a unsigned char. Unfortunately the gap between 436273009 and
> I see. Well, the difference is always even (except for the first
> one), so with storing half of the difference, the limit could be raised
> to 512. But I suppose the dev version took a better approach.
Yes, we store a number A*255+B as A bytes '255' and one byte B.
> > 436273291 is larger than 255. This limitation was removed in the
> > developement version. (Which I suggest you try out if you have
> > heavy duty computations to perform.)
> I really should do that. Is there a Debian binary package somewhere? Or
> a diff.gz file that could help me build my own one from the source?
Unfortunately not. We changed quite a bit the configuration system to
make for easier packaging, but the diff.gz needs to be updated.
> Other packages have different versions as Debian packages, too, e.g.
> gimp-1.2 vs. gimp-2.0 or mozilla vs. mozilla-snapshot. Wouldn't it be
> possible to have both version in Debian unstable?
I could package the CVS version but I am not sure there is a sufficient
number of interested users for a PARI/GP CVS package.
Also since the ABI/API are not stable in the CVS I could not provide
shared libraries package for the CVS version.
On the other hand the CVS version has some nice goodies like ia64, ppc
and amd64 assembly support , GMP support, much improved core algorithms,
and better gp2c support.
Imagine a large red swirl here.
 and maybe even hppa asm if some one know how to convert HPUX asm
to GNU as format.