Seeking a proof that 5^7954+7954 is prime

[WraithX]

Originally Posted by Jeff Gilchrist

Nice. Are you also working on an ECPP prover in C and GMP?

Even just the verifier will be useful for people who don't want to run it on Windows. Any plans on publishing your source code?

Hmmm... I hadn't planned on making an ECPP prover. Do you think there is a need for one? There are some algorithms in Crandall and Pomerance, I might be able to implement them if there's interest.

I can definitely publish the source code for the verifier. I'd like to try and clean up some of the code (maybe making it faster in the process?) before I release it though. Right now my program takes about 1.4 times longer than Primo to verify a certificate.

[retina]

Originally Posted by WraithX

Right now my program takes about 1.4 times longer than Primo to verify a certificate.

This is not a problem. Considering the number of verifications that are done you would likely spend more time improving it than you save in waiting for verification runs.

[Jeff Gilchrist]

Originally Posted by WraithX

Hmmm... I hadn't planned on making an ECPP prover. Do you think there is a need for one? There are some algorithms in Crandall and Pomerance, I might be able to implement them if there's interest.

Right now the only easily findable and usable tools AFAIK is Primo which only works under Windows and you have to do lots of manual work to get it to use multiple cores to prove.

So I think it would be very useful if someone wrote a portable one that could be compiled and run on UNIX or Windows that could use multiple cores or even cluster nodes (via MPI or something). That would make proving easier if using multiple cores was automatic and some bigger numbers could be proven if someone throws it at a system with lots of cores or a big cluster.

[WraithX]

Originally Posted by Jeff Gilchrist

Right now the only easily findable and usable tools AFAIK is Primo which only works under Windows and you have to do lots of manual work to get it to use multiple cores to prove.

So I think it would be very useful if someone wrote a portable one that could be compiled and run on UNIX or Windows that could use multiple cores or even cluster nodes (via MPI or something). That would make proving easier if using multiple cores was automatic and some bigger numbers could be proven if someone throws it at a system with lots of cores or a big cluster.

Yeah, I've read a little about what some people have done to run Primo in parallel. Looking at the algorithms in C&P, I'd have to find a good way to run them in parallel. Then, I'd have to learn a multi-threaded api. It'll be a while before I can get around to that. I have another program that I want to make multi-threaded that can easily take advantage of being such. So I'll start there and see how it goes. I'll try to clean up my verifier code this weekend and try to release it soon (maybe in the Software forum?). I'll put a link here to it once I've posted it.

[WraithX]

Okay, I've posted my Primo Verifier C code in the following thread:
http://www.mersenneforum.org/showthread.php?t=14086