Support for PRP/GEC w/ proof?

Jolly · 2023-04-20, 05:29

Hey all,

Sorry if it sounds like I'm nagging, but can I expect Mlucas to have support for PRP/GEC w/ proof anytime soon? I've been doing exclusively LLDCs on my M1 macbook air. The reason of the need for PRP/GEC w/ proof comes to me when one of my LLDCs is mismatched and an LLTC is required. If I could do PRP on my double checking efforts it'd be much nicer.

Andrew Usher · 2023-04-20, 12:23

Given how long it has been, I wouldn't count on proof generation being added to mlucas, though no one can presume to speak for Ernst. Mlucas is truly open-source so someone else could add proof generation and as long as it uses prime95's format it will be accepted.

Remember that you will always have some mismatches and, if you're concerned, track the rate of mismatches and (if known) whether it is your residue or the old one that is bad. If yours is almost always the good one, continue doing LL; if not (and there's nothing you can do to fix it) LL on that machine might not be worth it.

E.g. Ryan Propper, who does more LLs than anyone, maintains a 0.5% rate of mismatches, which amounts to less than the overhead of proofs (including the cert that someone else will do). If PRP/proof isn't even an option on your machine, obviously you must tolerate higher rates than that if you want to keep using it for primality tests - it is better than nothing

kriesel · 2023-04-20, 13:48

I think the last paragraph of https://www.mersenneforum.org/showpo...0&postcount=84 means Ernst had PRP proof or the equivalent working for Fermat numbers, in the as-yet-unreleased Mlucas V21. His plan was to finish implementation for Mersennes also before release. I've no idea how far he got on that. Last email I received from him, in early March, indicated he'd been having serious health issues subsequent to this, that interfere with making programming progress. (No response yet to an email I sent him 3 days ago.) His most recent mersenneforum post was in mid December.
Were he able to post for download a version with sufficiently complete implementation of Mersenne number PRP with proof, I think there would be no shortage of interest in alpha testing it.

It would be good if we could find volunteers with the aptitudes to assist and eventually succeed the original authors of the major GIMPS software titles. I've raised the topic of succession planning before and it went nowhere.

2023-04-20, 05:29	#1
Jolly "Felix" Feb 2023 3×7 Posts	Support for PRP/GEC w/ proof? Hey all, Sorry if it sounds like I'm nagging, but can I expect Mlucas to have support for PRP/GEC w/ proof anytime soon? I've been doing exclusively LLDCs on my M1 macbook air. The reason of the need for PRP/GEC w/ proof comes to me when one of my LLDCs is mismatched and an LLTC is required. If I could do PRP on my double checking efforts it'd be much nicer.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
"PRP proof never uploaded", but proof file was generated	CuriousKit	Software	4	2022-12-20 12:42
Can a Micro-ATX support a GPU as well as an ATX MB?	EdH	GPU Computing	7	2022-06-07 17:37
AMD To Support CUDA	Smokingenius	GPU Computing	9	2015-11-18 23:36
5+ GPU support	TheMawn	GPU Computing	3	2014-07-13 02:31
Support AVX	Unregistered	Information & Answers	5	2011-07-05 17:12

2023-04-20, 12:23	#2
Andrew Usher Dec 2022 2×3×7×13 Posts	Given how long it has been, I wouldn't count on proof generation being added to mlucas, though no one can presume to speak for Ernst. Mlucas is truly open-source so someone else could add proof generation and as long as it uses prime95's format it will be accepted. Remember that you will always have some mismatches and, if you're concerned, track the rate of mismatches and (if known) whether it is your residue or the old one that is bad. If yours is almost always the good one, continue doing LL; if not (and there's nothing you can do to fix it) LL on that machine might not be worth it. E.g. Ryan Propper, who does more LLs than anyone, maintains a 0.5% rate of mismatches, which amounts to less than the overhead of proofs (including the cert that someone else will do). If PRP/proof isn't even an option on your machine, obviously you must tolerate higher rates than that if you want to keep using it for primality tests - it is better than nothing

2023-04-20, 13:48	#3
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 17604₈ Posts	I think the last paragraph of https://www.mersenneforum.org/showpo...0&postcount=84 means Ernst had PRP proof or the equivalent working for Fermat numbers, in the as-yet-unreleased Mlucas V21. His plan was to finish implementation for Mersennes also before release. I've no idea how far he got on that. Last email I received from him, in early March, indicated he'd been having serious health issues subsequent to this, that interfere with making programming progress. (No response yet to an email I sent him 3 days ago.) His most recent mersenneforum post was in mid December. Were he able to post for download a version with sufficiently complete implementation of Mersenne number PRP with proof, I think there would be no shortage of interest in alpha testing it. It would be good if we could find volunteers with the aptitudes to assist and eventually succeed the original authors of the major GIMPS software titles. I've raised the topic of succession planning before and it went nowhere.