Re: Mersenne: Re: Mersenne Digest V1 #912
At 07:57 PM 12/2/2001 +, Gordon Spence wrote: A particularly sore point. If we maintained a top savers list whereby for every factor found you were credited with the time an LL test would have taken, then I and the other Lone Mersenne Hunters would pulverise these big university teams. Should George get credit for eliminating all those composite exponents when he made his initial list in the mid-1990's? He probably found over a dozen times as many composites in (at a guess) under two minutes than we'll EVER find, at least until the project is extended past 80M. Nathan _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
SV: Mersenne: Re: Mersenne Digest V1 #912
No you wouldn't because they would like yourself go to do only factoring work, and as George said to me when I propose this a year ago: It would make the focus of GIMPS towards factoring and not like it is now on primenumber finding. We will (again in George-words) have to eliminate candidates by using a small amount of time in the factoring field; but only so much/less that we still have power to run LL-tests to really prove primes. Regards and happy hunting tsc A particularly sore point. If we maintained a top savers list whereby for every factor found you were credited with the time an LL test would have taken, then I and the other Lone Mersenne Hunters would pulverise these big university teams. 150,000 factors in the 60-69m range, at an average of 27.2 P-90 years each - h just over 4,000,000 years saved regards Gordon _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Re: need access to an SGI
Whoops - when I wrote the original message, I forgot that the source tarball for Mlucas 2.7c is still in the build area of the ftp archive - it hasn't yet been promoted to the release area. You can get it at ftp://hogranch.com/pub/mayer/junk/Mlucas_2.7c.tar.gz -Ernst _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Re: SF Bay area GIMPS party
OK, the San Francisco Bay area GIMPS get-together will take place this coming Friday, 7. December, in the south bay area (precise venue to be decided soon - see below.) Any GIMPS participant or Mersenne prime fan is welcome to join us, along with anyone else you'd like to bring - this is not a formal party.For those not coming from the San Jose area, the venue will be close to a Caltrain station, and hence should be reachable from e.g. SFO airport or other outlying areas. I'm also going to try to arrange what ridesharing I can, so if you can provide a ride or need a ride, please let me know. I looks like several folks from far-flung places will be attending, so should be a quite interesting company. Alas, Luke Welsh can't attend, but he suggested several venues, two of which form my short list. 1) Tied House, Mountain View 2) Faultline Brewery, Sunnyvale 1) is very close to a Caltrain station; 2) supposedly has very good food. Luke also mentioned the Tied House in downtown San Jose, but I've heard it's often very crowded and not as easy in terms of public transit access as the one in Mountain View. So, if you plan to attend and have been to both places (I've not been to (2)), your preference would be most valuable. I hope to finalize the location this afternoon, so out-of-towners will have a reasonable amount of time to make their plans. Since the proverbial M#39 cat is out of the proverbial bag, the guess-the-new-Mersenne-prime-exponent contest is officially cancelled. I was hoping to have a prime poster signed by all the attendees to give to the winner, but I don't think they'll be ready by then, anyway. Cheers, -Ernst _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Re: SF Bay area GIMPS party
1) Tied House, Mountain View 2) Faultline Brewery, Sunnyvale either, orther, as far as I care... heh. re: ride sharing, I have a new 7 passenger van, and might be able to carpool folks from the Santa Cruz area... assuming I can go (I've got to check with SWMBO to make sure we don't have a prior engagement on that date). -jrp _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: M#39 news!
- Original Message - From: [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Saturday, December 01, 2001 8:02 PM Subject: Re: Mersenne: M#39 news! ...I would have expected George's initial announcement to say over 4 million digits rather than well over 3.5 million digits (which would nevertheless be true!). Or he might have said nearly 4 million digits which would also have been true. Of course I could simply be misreading George's mind - possibly if he had said over 4 million it would have narrowed the field sufficiently to make identification easy. I fear that this may be the case, given what various people have been able to dig out of what has been said. [...] I would strongly suggest that procedures are changed so that the next time a Mersenne prime is discovered, no information at all is released except to prior discoverers of Mersenne primes... As a matter of interest, why should prior discoverers be so privileged? [...] Irritated Yes, I rather feel the same. It's like enjoying the build-up to Christmas only to find that someone's open your present five days ahead of time. Brian Beesley Daran G. _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: Factoring benefit/cost ratio
- Original Message - From: George Woltman [EMAIL PROTECTED] To: Gerry Snyder [EMAIL PROTECTED]; [EMAIL PROTECTED] Sent: Saturday, December 01, 2001 10:39 PM Subject: Re: Mersenne: Re: Factoring benefit/cost ratio I prefer a factor to a double-check. But it is hard to quantify prefer in a mathematical formula for computing trial factoring limits. Prime95 uses the formula: cost_of_factoring must be less than chance_of_finding_a_factor times 2.03 * the cost_of_an_LL_test. Shouldn't that be 1.015 for double-checking assignments? Also does the cost part of the calculation recognise the increased cost of trial-factorisation after 2^64? I've noticed on occasion that I've had to do an extra round of trial factoring before proceeding with an doublecheck. This indicates that previous factorisation has been non-optimal, or have the estimates for the relative costs of factoring vs. LL testing changed with the introduction of new hardware? Finally if P-1 factorisation were to be spun off into a separate work unit, then the optimal arangement would be to trial factor while cost_of_trial_factoring * chance_of_P-1_factoring is less than cost_of_P-1_factoring * chance_of_trial_factoring. Then P-1 factorise. Then complete trial factorisation according to the above formula. -- George Daran G. _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: Factoring benefit/cost ratio
On 3 Dec 2001, at 20:45, Daran wrote: Shouldn't that be 1.015 for double-checking assignments? I think 1.03. However you do have a point. P-1 limits do depend on the trial factoring depth, and are much smaller for DC assignments than for first tests, so there is already something built in. Also does the cost part of the calculation recognise the increased cost of trial-factorisation after 2^64? Yes. The trial factoring depth is constant at 64 bits from ~8.5 million to ~13 million. Don't forget that the number of candidates which need to be checked is _inversely_ proportional to the exponent. I've noticed on occasion that I've had to do an extra round of trial factoring before proceeding with an doublecheck. This indicates that previous factorisation has been non-optimal, or have the estimates for the relative costs of factoring vs. LL testing changed with the introduction of new hardware? I think the latter has something to do with it - PPro is about twice as efficient at factoring compared with LL testing as a plain Pentium. This is because of much improved pipeline organization including the provision of spare registers enabling speculative execution, which greatly increased the throughput of the floating point unit in particular. Another factor is that early versions of the program were unable to factor as deep as current versions. Finally if P-1 factorisation were to be spun off into a separate work unit, then the optimal arangement would be to trial factor while cost_of_trial_factoring * chance_of_P-1_factoring is less than cost_of_P-1_factoring * chance_of_trial_factoring. Then P-1 factorise. Then complete trial factorisation according to the above formula. Interesting - but I think the effect would be small. What about factoring to a fractional depth? With a roughly logarithmic distribution of factors, surely about half the factors between 2^n and 2^(n+1) would be smaller than 2^(n+0.5), whilst searching to 2^(n+0.5) would take only about 41% of the time taken to search the whole interval. Regards Brian Beesley _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: M#39 news!
At 08:45 PM 12/3/01 -, Daran wrote: At 08:02 PM 12/1/01 -, Brian Beesley wrote: I would strongly suggest that procedures are changed so that the next time a Mersenne prime is discovered, no information at all is released except to prior discoverers of Mersenne primes... As a matter of interest, why should prior discoverers be so privileged? Various reasons, but elitism not one of them. One of the reasons is to provide an opportunity to offer advice to the new discoverer on how to handle the publicity... what kinds of, umm, insightful questions the press might ask... and which answers are likely to make it into print. --Luke, who *still* manages to look foolish _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: Factoring benefit/cost ratio
Responses to a couple of postings -- Steve Harris wrote: Actually, Richard's statement that a 'Factored' status is better for GIMPS than a 'Two LL' status is not quite true. It's better for the mathematical community as a whole, Better for the mathematical community as a whole is actually almost exactly what I had in mind when I started composing that paragraph ... honest! But then I started revising, and lost some stuff in the process. Nevertheless, I stand by my assertion that a 'Factored' status is better for GIMPS than a 'Two LL' status ... insofar as our LL testing remains less reliable than our factor verification. Double-checking certainly improves LL result reliability, but I think our record shows that a verified factor is still slightly (by a minute but nonzero margin) more reliable an indicator of compositeness than two matching nonzero LL residues. (See earlier discussions of potential reasons for incorrect LL results. Some error sources would be more likely to affect the longer LL test than the shorter factor verification.) It matters to those who are attempting to fully factor Mersenne numbers, but that's a different project altogether, Some of us like to participate in both. :-) and one that is decades (at least) behind GIMPS. The only reason we do any factoring at all is to reduce the time spent on LL testing. But if factoring is not really part of GIMPS's purpose (and I agree it isn't), how can a separate factoring effort be behind GIMPS at all? Aren't they measuring their progress in a different, non-comparable, dimension? Besides, if you do manage to find a 75-digit factor of a 2-million-digit Mersenne number, that still leaves a 125-digit remainder. Really not much help :-) A journey of a thousand miles begins with but a single step. Two million three-foot steps total more than one thousand miles. - - - - - In reply to a statement of mine about the extra benefit of finding a specific factor, Daran G. wrote: I can see no way of objectively quantifying this benefit. Well -- if there's no objective quantification of the extra benefit of finding a specific factor, then it seems to me that there's no objectively quantifiable justification for saying that it's not valuable to do a certain amount of extra P-1 factoring on a once-LL'd Mnumber. :) In response to my writing: This reflects the view (with which I agree) that it is more valuable to know a specific factor of a Mnumber than to know that a Mnumber is composite but not to know any specific factor of that Mnumber. So a Factored status is better for GIMPS than a Two LL status, but calculations of factoring benefit that consider only the savings of L-L test elimination are neglecting the difference between those two statuses. If one consciously wants to neglect that difference ... well, okay ... but I prefer to see that explicitly acknowledged. Daran G. wrote: It seems to be implicitely acknowledged in the way the trial factoring depths are determined. But don't forget that this refers to a depth determined by Prime95 _in the context of calculating the factoring tradeoff point for maximizing GIMPS throughput for the Test= worktype_, NOT the context of a Factor= or Pminus1= worktype where the user has explicitly specified factoring limits, possibly for reasons beyond the ken of Prime95. Daran continues: If one places a non-zero value on a known factor, then the utility of extra factoring work on untested, once tested, and verified composites would be increased. Yes. But that doesn't have to affect Prime95's calculation of tradeoff points _for maximizing throughput of the Test= worktype_. I'm not saying that Prime95 should incorporate into its calculations a nonzero value for the benefit of finding a specific factor. I'm saying that it is rational for someone to decide to factor past the Prime95-calculated tradeoff points, and that it is unjustified to criticize extra factoring on the grounds that going past the Prime95-calculated tradeoff points is wasted effort. Richard B. Woods _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Re: M#39 news
Hi all, Mersenne mailing list readers were treated to unexpected sneak-peek at the new Mersenne prime's exponent. Science magazine heard of the discovery and contacted Scott and I. Since Science magazine is one of the two most widely read science-oriented magazines, we felt it would be a good opportunity. Because of longish lead times in the magazine publishing industry, we took a calculated risk and gave them a draft of the press release with the understanding that specific details should be withheld until confirmation is official. Unfortunately, this did not happen. This was an unintentional mistake for which the author has sincerely apologized. The on-line article has been corrected. Fortunately, word has not spread too widely and it seems this slip-up will not interfere with an orderly announcement later this week. Press coverage is always a wild card. The quality and number of news stories depend on many factors including the author's knowledge and interest, editor's pen, space considerations, other news, time since last prime found, whether an interesting angle can be promoted, etc. M#35 got good coverage with the PC bests supercomputer angle. M#36 got swamped by Princess Diana's death. M#37 didn't make too big a splash. M#38 got decent coverage because of the EFF award. M#39 - time will tell. I don't see any great reason to change our announcement policies. The premature article on slashdot may even have been beneficial as several reporters have contacted us rather than us having to go out and drum up interest. Have fun, George _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: Factoring benefit/cost ratio
At 08:38 PM 12/3/2001 -0600, [EMAIL PROTECTED] wrote: Nevertheless, I stand by my assertion that a 'Factored' status is better for GIMPS than a 'Two LL' status I think everyone is in agreement on this. But if factoring is not really part of GIMPS's purpose (and I agree it isn't), I think of GIMPS as two projects. The larger project is let's find the next Mersenne prime. The much smaller project is let's maximize our knowledge of all Mersenne numbers. I'm saying that it is rational for someone to decide to factor past the Prime95-calculated tradeoff points, I agree. I've done several P90 CPU years of that myself. I think more of the discussion has centered around stats and the formula for picking how far to trial factor, rather than whether factoring is of some mathematical value. I'm not inclined to change the trial factoring formula. If I recall correctly, when I last computed the formula I erred on the side of more factoring rather than less. As to the stats, maybe, just maybe, we can upgrade our stats system in 2002. I know from other distributed projects that many users are fanatical about any changes in stats. We will have to tread very carefully! -- George _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
