P-1 factoring as distributed computing

jocelynl · 2002-11-22, 16:41

Hi George,

I've finally been able to divide the work on P-1 factoring. Let me know if this would be of any help in prime95.

Deamiter · 2002-11-22, 21:27

p-1 factoring goes by so fast, I don't see how it would be useful. Now if George implimented an option to do p-1 along with the factoring, I'd set up a box or two for that, but when you're talking weeks for a single exponent, a few hours probably isn't a big deal.

philmoore · 2002-11-23, 00:27

For GIMPS, P-1 factoring only makes sense as long as the expected time saved doing Lucas-Lehmer tests is greater than the time spent factoring. As a result, P-1 factoring assignments are usually completed within a day or so, and being able to distribute the job would not be a big savings. However, for Fermat numbers, it would be advantageous to do some deep P-1 work before going on to ECM, and for these cases, being able to do stage 2 P-1 work in a distributed manner could be a big advantage. One machine could run stage 1 to a bound of a few million, and the stage 2 work could then run on several machines simultaneously to a billion or more. For the 25th Fermat number, I am guessing that my Athlon XP (1400 MHz) would take 2 years to get to 1 billion, and for the 26th Fermat number, it would probably take twice as long to get that far. So yes, there are definitely cases where being able to distribute the work would be an advantage.

2002-11-22, 16:41	#1
jocelynl Sep 2002 106₁₆ Posts	P-1 factoring as distributed computing Hi George, I've finally been able to divide the work on P-1 factoring. Let me know if this would be of any help in prime95.

2002-11-23, 00:27	#3
philmoore "Phil" Sep 2002 Tracktown, U.S.A. 3×373 Posts	P-1 distributed work For GIMPS, P-1 factoring only makes sense as long as the expected time saved doing Lucas-Lehmer tests is greater than the time spent factoring. As a result, P-1 factoring assignments are usually completed within a day or so, and being able to distribute the job would not be a big savings. However, for Fermat numbers, it would be advantageous to do some deep P-1 work before going on to ECM, and for these cases, being able to do stage 2 P-1 work in a distributed manner could be a big advantage. One machine could run stage 1 to a bound of a few million, and the stage 2 work could then run on several machines simultaneously to a billion or more. For the 25th Fermat number, I am guessing that my Athlon XP (1400 MHz) would take 2 years to get to 1 billion, and for the 26th Fermat number, it would probably take twice as long to get that far. So yes, there are definitely cases where being able to distribute the work would be an advantage.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
My plan for RSA factoring distributed computing	stathmk	Open Projects	37	2019-12-08 04:36
Etiquettes of Distributed Computing	a1call	Miscellaneous Math	8	2018-05-21 16:25
Considering getting back into distributed computing	jasong	jasong	7	2016-03-28 14:36
Massively distributed computing and factoring...	flouran	Math	2	2009-11-21 05:30
The difference between P2P and distributed computing and grid computing	GP2	Lounge	2	2003-12-03 14:13

2002-11-22, 21:27	#2
Deamiter Sep 2002 117₁₀ Posts	p-1 factoring goes by so fast, I don't see how it would be useful. Now if George implimented an option to do p-1 along with the factoring, I'd set up a box or two for that, but when you're talking weeks for a single exponent, a few hours probably isn't a big deal.