Karim BELABAS on Wed, 5 Feb 2003 00:02:01 +0100 (MET)


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

Re: quadclassunit() bug


On Thu, 30 Jan 2003, Igor Schein wrote:
> On Thu, Jan 30, 2003 at 03:40:29PM -0500, Igor Schein wrote:
>> ? setrand(178);quadclassunit(82421)[1]
>> 2
>>
>> This is wrong, should be 1.  It looks like it can be related to the
>> top priority items in section BUGS (general) of TODO file.
>
> Actually, it looks like regression, since it was working fine prior
> to this change:
>
> \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
> revision 1.49
> date: 2002/09/02 21:15:07;  author: karim;  state: Exp;  lines: +365 -450
> major cleanup buchquad [preliminary]
> \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
>
> Or maybe, rather, this change exposed previously hidden bug.

Not really. Before the patch, an approximation to L(chi, 1) was computed
using the first 30000 primes. After, we use the p <= cbach2 * (log D)^2
[ as the generic program bnfinit() does ].

This is not enough for quadratic fields of small discriminants, because we
cheat even more on the default value for cbach than we do in the case of
general fields... [ cbach = 0.1 where it should be 6. For general fields,
it's 0.3 where it should be 12 ]

I have added a sanity check to use at least all primes less than 100 [
arbitrary, tested OK for a while; we'll see if bad examples crop up... ]

    Karim.
-- 
Karim Belabas                    Tel: (+33) (0)1 69 15 57 48
Dép. de Mathématiques, Bât. 425  Fax: (+33) (0)1 69 15 60 19
Université Paris-Sud             Email: Karim.Belabas@math.u-psud.fr
F-91405 Orsay (France)           http://www.math.u-psud.fr/~belabas/
--
PARI/GP Home Page: http://www.parigp-home.de/