> > Rob looks at the new points every day. With 1/3 of 1.3 millions
> > points, the chance of having found the solution is about 1/20.
>
> While we're at it : do you check that no two "machines" are on the same
> track?
No. Too complicated. People have to check that, when they see that some
machine seems to be stopped, although it is not (the duplicated points
are not registered).
> Chances for that are theoretically infinitesimal, I think, but I
> reckon that the probability is not negligible in practice, since it
> is so easy to start two processes with the same "saved.state"...
If somebody uses the same saved.state in two different machines, the
probability of error is 1. For ecdl8, I asked Rob to find something to
avoid that.
-- Daniel de RAUGLAUDRE daniel.de_rauglaudre@inria.fr http://cristal.inria.fr/~ddr/
This archive was generated by hypermail 2b29 : Tue Jan 18 2000 - 19:17:51 MET