BOINC

Has GIMPS ever considered adding itself to the BOINC project, http://boinc.berkeley.edu/index.php ?

That'll open a whole new can of worms:

http://www.mersenneforum.org/showthread.php?t=13901

[Uncwilly]

Please look at these.

http://www.mersenneforum.org/showpost.php?p=67753

http://www.mersenneforum.org/showthread.php?t=2639

[mdettweiler]

Originally Posted by Uncwilly

Please look at these.

http://www.mersenneforum.org/showpost.php?p=67753

http://www.mersenneforum.org/showthread.php?t=2639

BOINC has come a long way since those posts were made. However, the fact does remain that Prime95 is not easily adaptable to BOINC, since it is tightly integrated with the PrimeNet server.

The somewhat closely related LLR has been successfully ported to BOINC by PrimeGrid, who's been using it for the last number of years with great success. Since a BOINC server is a major pain to run, they've served as a central umbrella under which a lot of mersenneforum non-BOINC projects have gone BOINC alongside their other participation options. If GIMPS ever decided to go BOINC, the easiest route would probably be to do so through PrimeGrid.

It might be easiest to write a BOINC LL testing application based on ntprime, instead of Prime95. Just strip out the PrimeNet code and have it accept exponents to test on the command line, have it write % progress to a simple text file instead of to the console, and boom, you're all set. A wrapper program to handle communication with BOINC can be pretty easily written from there.

[frmky]

It would also be quite easy to create a CUDA BOINC client for cards with DP support from msft's code. That would attract significant interest from BOINC users. But, of course, this all requires George's endorsement.

[Prime95]

Originally Posted by mdettweiler

It might be easiest to write a BOINC LL testing application based on ntprime, instead of Prime95. Just strip out the PrimeNet code and have it accept exponents to test on the command line...

The major problem with that is that it violates the license agreement for the gwnum library. You can use it for anything you want -- except finding Mersenne primes.

BOINC is fine but I have little interest in porting prime95 to it. Yes, it would greatly increase our computing power. However, I don't mind keeping the project smaller - it makes it more likely that prime discoverer's are participating due to a love of math.

[moebius]

Furthermore, I could be tempted again to help design and analyze quantum computing algorithms instead of the lengthy searching for primes. LOL.