Parallel Prime Search

2007-02-07, 08:10

Does anyone have a parallel prime search written in MPI? Our cluster is being under utilized, so I thought I'd throw a couple hundred thousand hours at something worthwhile.

dtripp@hawaii.edu

BotXXX · 2007-02-17, 17:10

Quote:

Originally Posted by DonaldTripp

Does anyone have a parallel prime search written in MPI? Our cluster is being under utilized, so I thought I'd throw a couple hundred thousand hours at something worthwhile.

Not sure if it is compatible with MPI but to verify finds of Mersenne primes usually a few people 'donate' time on quite large machines to have a fast double check. They use Glucas incombination with PThreads.
Not sure if it might help you in your case.

Glucas -> http://www.oxixares.com/glucas/

ewmayer · 2007-02-17, 19:35

Donald, none of the LL testing codes is parallelized to a very large degree, and even the ones that are suffer a penalty when running on multiple CPUs, so the best way for you to contribute bandwidth is to run one copy of whatever client is best for your setup (Prime95 for x86-ish, Glucas or Mlucas for anything else) on each available CPU.

I realize setting that up may require more overhead than you feel like putting in.

Cheers,
-Ernst

2007-02-07, 08:10	#1
DonaldTripp 67F₁₆ Posts	Parallel Prime Search Does anyone have a parallel prime search written in MPI? Our cluster is being under utilized, so I thought I'd throw a couple hundred thousand hours at something worthwhile. dtripp@hawaii.edu

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Prime Search at School	Unregistered	Information & Answers	5	2009-10-15 22:44
Exponential prime search	Citrix	Prime Sierpinski Project	37	2009-08-17 06:32
Benson's prime search	Kosmaj	Riesel Prime Search	67	2009-01-18 21:59
Parallel version of Prime for HPC?	dtripp	Software	3	2008-09-30 19:52
Prime Search on PS-3?	Kosmaj	Riesel Prime Search	6	2006-11-21 15:19

2007-02-17, 19:35	#3
ewmayer ∂²ω=0 Sep 2002 República de California 10110101011101₂ Posts	Donald, none of the LL testing codes is parallelized to a very large degree, and even the ones that are suffer a penalty when running on multiple CPUs, so the best way for you to contribute bandwidth is to run one copy of whatever client is best for your setup (Prime95 for x86-ish, Glucas or Mlucas for anything else) on each available CPU. I realize setting that up may require more overhead than you feel like putting in. Cheers, -Ernst