[Ecm-discuss] Issues with ecm-6.1.1.

Laurent Fousse laurent at komite.net
Lun 31 Juil 12:09:43 CEST 2006


* Paul Zimmermann [Tue, Jul 25, 2006 at 11:56:03AM +0200]:
> > I don't know what's causing this problem ("not enough primes in
> > interval") on this architecture.
> 
> This error message comes from NTT, file mpzspm.c. A solution would be to
> disable NTT on this architecture.

The first test that fails is 

    echo 328006342451 | ecm -pp1 -x0 5 120 7043

The first prime that should be selected is 4294966657. Tracing
incorrect values up to the root of the problem leads to umul_ppmm,
already unable to square 2:

Breakpoint 3, sp_sqr (x=2, m=4294967265, d=31) at sp.h:279
279       umul_ppmm (u, v, x, x);
(gdb) print x
$14 = 2
(gdb) n
280       udiv_qrnnd_preinv2norm (w, z, u, v, m, d);
(gdb) print u
$15 = 0
(gdb) print v
$16 = 3223094856

If I disable umul_ppmm (as I did for s390) ecm compiles correctly.




Plus d'informations sur la liste de diffusion Ecm-discuss