[RC5]V3...Lets beat the GIMPS effort

gindrup at okway.okstate.edu gindrup at okway.okstate.edu
Tue Aug 18 18:08:13 EDT 1998


     GIMPS (apparently) has two kinds of operations performed by their 
     clients.
     
     The first is the Lucas-Lehmer test.  By itself, this test applied to 
     an odd number is pretty quick.  However, up around where GIMPS is 
     currently searching, there are a lot of composite odd numbers.
     
     The second is factorization of known "Mersenne composites".  I see a 
     reference to an Elliptic Curve factorization method on the GIMPS 
     site.  I *thought* they were using Number Field Sieve methods up 
     until very recently.  Anyway...
     
     These two operations could in a new client be broken up.  In the 
     current GIMPS client this does not appear to be the case.
     
     Finding a new prime can be easily attributed to the one person who 
     tested it.
     
     Factorization using the above methods is usually nondeterministic.  
     This means (in particular) that only the person whose machine picks 
     a good "trial value" and completes the trial will find a factor.  
     Everyone else will just find out that the number is divisible by 
     one.
     
     The *slow* part is the division afterward to find out what's left 
     after the new factor is removed.
            -- Eric Gindrup ! gindrup at okway.okstate.edu


______________________________ Reply Separator _________________________________
Subject: RE: [RC5]V3...Lets beat the GIMPS effort 
Author:  <rc5 at lists.distributed.net> at SMTP
Date:    8/18/98 11:06 AM


But considering the d.net website and sheer amount of users participating, I 
think that a lot of participants who are currently participating in GIMPS or 
both GIMPS and d.net will switch over to one solid d.net core.  Why worry 
about having two different clients to install(some on some machines and some 
on others) when they could just install one client on all machines.
     
One thing I think would also be a good idea is for d.net's recommendation on 
what kind of work to be done to be included in the client program(determined 
by processor and amount of RAM or something).  I know this may slightly 
increase the file size, but I think many users, especially new ones, may find 
it annoying to have to go back to the website to determine what project would 
be best for their machine.
     
Also, another issue is that with Mersenne primes, factoring will still need 
to be done.  The problem is that there won't be any 'glory' in factoring. 
Who wants to work to make it easier for someone else to find a new prime 
number when they could have just dedicated their machine to finding the 
prime number themselves.  So, I think that perhaps d.net should force this 
on people.  Basically you HAVE to do some factoring, like it won't allocate 
you a new part of a prime to check until you accomplish a certain block of 
factoring.  
     
Another issue is well, how will this work.  My P2-266 can crunch a prime 
number in about 2 weeks.  Now, I know that's going to be unappealing to 
people and hopefully d.net will break that up into smaller chunks.  But, if 
this happens and a new prime is discovered then who would get credit?  Will 
the new stats box keep track of all email addresses that worked on a number 
or will the credit just be given to "The Users of d.Net"?
     
I know I'm kind of just spouting off, I'm sure that d.net is already trying 
to tackle this problem.  It's just that info about v3 has been sparse for 
the last month or so and I know I'm not the only user waiting for something 
to chew on.
     [snip]

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list