View Single Post
Old 2012-08-10, 00:43   #622
P90 years forever!
Prime95's Avatar
Aug 2002
Yeehaw, FL

23·1,021 Posts

Originally Posted by rcv
I believe GPU272'ers are hard-working, dedicated, technically advanced users who want what's best for GIMPS. But you GPU272'ers just can't take criticism.
This should be amended to some or many can't take criticism. Many see a critical rcv post and have a knee-jerk reaction to bash back.

For those that do not know, rcv has contributed more to GIMPS than most. He does not toot his own horn, but his initial efforts at GPU sieving code is the basis for the code I'm now working on incorporating into mfaktc. I probably learned more about good CUDA programming techniques from his program and documentation than I've learned from reading Nvidia manuals.

So, before launching a knee-jerk counter-attack ask yourself have I contributed that much to GIMPS?

It's really George's decision, as he is the one to make announcements, and it's his credibility on the line. [I suspect there's a little cross-checking that goes on behind the scenes just prior to an announcement that M(x) is the nth Mersenne Prime.] And the language that is used in the announcement can vary with the circumstances. "All candidates below x have been independently verified by at least two independent users at least two years apart" versus "All candidates below x have been tested twice" versus "At least two reports have been received for all candidates below x".
Sorry to disappoint. There is no behind-the-scenes cross-checking. Brian Beesley used to do triple-checking if the 2 LL tests were done by the same user. However, he stopped at somewhere below M2000000.

When the milestones page says "GIMPS proves Mxxxxxxxx is the x-th Mersenne Prime", there really should be a giant asterisk next to "proves". With a community-based project it is impossible, or nearly-impossible, to eliminate the possibility of a malicious user submitting bogus double-checks just for the fun of it.

The core question is does GPU72/73 increase those risks. I personally don't think so. The dedicated users of GPU72/73 are the least likely to engage in malicious activities. Furthermore, if GPU72/73 did not exist, could we remove the giant asterisk? Nope.

I think the true independent proof of "Mx is n-th Mersenne prime" is many years down-the-road when computers are so fast that somebody decides to test all the exponents below x and compares them to our database.

My biggest concern was that long-time, dedicated, trustworthy PrimeNet users no longer have an equal chance to obtain some of the best assignments.
This is undeniably true, but I think in somewhat different way. The number of users that knew exactly when the plum assignments were made available and were willing to make the effort to get them was very, very small. (I was one. Once a month or two I'd open prime95 at the appointed hour and unreserve big exponents to get the plum smaller assignments). These few users are likely dedicated enough to learn of GPU72/73 and get their plum assignments that way.

Where GPU72 hurts is the average user no longer has even a small chance of getting a plum assignment (if prime95 happened to contact the server within a few hours of the golden hour) they now have virtually no chance as the bot got them all within minutes. Most users probably don't care, but some might resent being reduced to "second-class" GIMPSers.

I had hoped that once GPU72 "caught up" with its TF efforts, it would end up keeping a small number of the plum assignments (say a 100 or so) to satisfy GPU72ers day-to-day needs. To the average PrimeNet user there is little difference between the old system and a system where GPU72 has pre-reserved the lowest 100.

I'm not sure how GPU72 is currently configured. Do they keep the 100 smallest? or is it a 1000? or would it be fairer if it made half of the smallest exponents available to average PrimeNet users by say only holding exponents that are 1 mod 4? This is a fair topic for debate.

On the plus side for the average user, the GPU72 project -- necessitated because PrimeNet doesn't handle GPU clients -- now gets exponents that are factored 4 bits deeper or approximately a 4% better chance of resulting in a new prime.
Prime95 is online now   Reply With Quote