Karim BELABAS on Tue, 13 Aug 2002 15:48:32 +0200 (MEST)

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

Re: rnfkummer-induced bug

On Sat, 10 Aug 2002, Igor Schein wrote:
>> I've committed all my remaining rnfkummer patches. In its current state,
>> rnfkummer should be able to treat extensions of arbitrary prime degree (in
>> particular larger than 5...), again provided that bnfinit( K(zeta_ell) )
>> succeeds.
>> I don't intend to modify it further. You can crash-test it now :-)

I found one area which could use improvement: it would be nice to remove the
restriction that the degree has to be prime. It's trivial to extend the code to
squarefree degrees [in fact, to reduce the general case to degree = p^n, p
prime]; and we have the obvious but inefficient algorithm of making
successive extensions of degree p to treat the case p^n.

I'll wait a bit then try to implement this.

> Was it a challenge? :)

If taking it as such improves your already tremendous bug-finding rate ...

> ? setrand(1);rnfkummer(bnrinit(bnfinit(quadpoly(-1752,y)),19,1),matdiagonal([5,1,1]));
>   ***   bug in GP (Segmentation Fault), please report

Unrelated to rnfkummer, quite specific to bnfinit() [ memory reallocation
problem. I had added relatively recently an extra array first_non_zero[] in
buchall() to remove an horrible kludge, and forgot to resize it together with
its companion arrays ]. It's gone now.

Karim Belabas                    Tel: (+33) (0)1 69 15 57 48
Dép. de Mathematiques, Bat. 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/