Re: Mersenne: Re: Mersenne Digest V1 #575
Am I just wrong in thinking this? I fully agree with your opinions above[1], but you are wrong when you make your own rules based on these opinions, and try to force other people to follow your rules and not the commonly accepted ones. I don't want to play with people who don't follow the rules or change the rules during the game, and I don't have to. This is supposed to be fun, and your behaviour makes it the oposite. I don't want to stand guard over my exponents, sending in false progress reports to make you stay away from them. Follow the rules Aaron, or leave the game. Please. __ [1] Have a look at "Fact.P90 CPU yrs" on the top 100 list at Primenet, where I'm S00113. My factoring to LL ratio is second on the list, after "dodrillm". (Only a fraction of my work pass trough Primenet.) -- Sturle URL: http://www.stud.ifi.uio.no/~sturles/ Er det m}ndag i dag? ~~ MMF: http://www.alladvantage.com/go.asp?refid=BUP399 - St. URLe Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: status of exponents
You will notice that in an email from George he said he does occasionally go through the database and release those exponents that don't look like they will be done by his own criteria and that he discourages "poaching". I would think that you would trust him, and just get new exponents automatically. This will help keep both you and others from "wasting" cpu time, duplicating first time checks. I really don't think that anything will really "fall through the cracks". There isn't really much point in arguing about this further, its discouraged, but we can't do much about people who insist on doing it. Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Windoze joke
Hey, guys, this is not strictly mersenne related, but I think we all deserve a laugh. Microsoft have just released a patch for all versions of Win 9x (_all_ versions of Win 95 _and_ Win 98) which (is claimed to) resolve a problem with systems hanging after 49.7 days continuous operation. Only took them four years to find out that this might be an issue ;-) The patch can be downloaded from http://support.microsoft.com/download/support/msfiles/ - Vtdapi95.exe (The existence of the file is _not_ a joke!) Regards Brian Beesley Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: These go to 11 (WAS: blahblah...)
or, more concisely, (1+1+1)^(1+1) + 1. Can anyone represent that number in fewer than (1+1+1)! ones? This all depends on what operators and notations are accepted and without specifying that, the whole question is useless. What about without any ones at all: (With C++ operators) ((0++)++)*(0++)++)++)++)++) or without any numbers: (With normal algebra) ((a/a)+(a/a))*((a/a)+(a/a)+(a/a)+(a/a)+(a/a)), a [belongs to] N -- Markus Laire [EMAIL PROTECTED] ICQ# 11887013 http://www.nic.fi/~laire/english or http://come.to/markuslaire PGP Key: 4096/1024 DH/DSSID: 0xB93CD277 Fingerprint: 7C6B AE89 C243 F5A4 9702 EDAB 19ED 59B0 B93C D277 Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Windoze joke
The existence of this file is _not_ a joke. (not a direct quote, force of habit, I deleted the message before I could respond). Yes, but it seems like a joke to us Linux users out there, for whom 2-day patches are not uncommon... -Lucas Wiman Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: LL and factoring quitting
lrwiman [EMAIL PROTECTED] wrote: It has been mentioned several times recently that factoring is all integer work, and LL testing is nearly all floating point. It is my understanding that on intel CPU's, these are done on separate parts of the CPU. Would it increase net performance to do factoring and LL assignments at the same time? If we can execute the two codes together, the functional units will be better utilized. But we will need the combined memory bandwidth for both codes, and enough cache for both working sets. On Pentiums, a bigger concern is the number of registers -- most floating point intensive algorithms need several integer registers for loop counters and address computations, so the integer code will lack these. The two algorithms must be designed so the control (i.e., branch) logic is identical -- we don't want the integer code to be calling a subroutine which executes the Euclidean GCD algorithm and repeatedly tests whether the latest remainder is zero, while the floating point code loops over the FFT output to square each element, for example. When the control logic is identical for two algorithms, and the subscripting (i.e., memory access) patterns are similar, it may be feasible to merge integer and floating point algorithms for improved functional unit utilization. For example, Ernst Mayer is writing code which will do an integer FFT and a floating point FFT of the same length concurrently. Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: LL and factoring quitting
lrwiman [EMAIL PROTECTED] writes: It has been mentioned several times recently that factoring is all integer work, and LL testing is nearly all floating point. It is my understanding that on intel CPU's, these are done on separate parts of the CPU. Would it increase net performance to do factoring and LL assignments at the same time? There is _some_ scope here, but I think we need to be _very_ careful. Actually, the Intel CPU is a bit of a nightmare from this point of view. a) The integer multiplier and the floating-point multiplier share common circuitry, therefore a (I)MUL and a FMUL cannot execute in parallel. Since division is (only very partially) pipelined takes "forever", we avoid using it where practical. Therefore, the time- dominant code in (integer) factorization is actually multiplication. b) Some CPU types - including the Intel P6 family, for which the potential gain would be greatest - already use the FPU (at least to some extent) to do factorization - because can be faster to use the FPU to do integer multiplies than it is to do them in the integer ALU, despite the conversions - doing this reduces the demand on the (very limited number of) (named) integer registers, too. c) If you add extra integer mode instructions into the opcode stream, you'd have to be very careful that these do not cause opcode fetch, decoder or data bottlenecks, even when they don't depend on shared execution units. We really _don't_ want to slow down the LL code _at all_! Looking at George's macros yesterday, I was struck by how efficient they are in terms of usage of FPU registers L1 data cache lines. Really you couldn't squeeze anything else in there, unless it would run entirely in (integer) CPU registers, or you were prepared to take the "hit" involved in L1 data cache line misses quite frequently. Also, I'm going to quit first time LL testing. Call me impatient, but I don't want to wait until early July for my exponent to finish, thus I'm going to switch to double-checking. Hi, Mr. Impatient! Seriously, though, I wonder how many others will do this now that it looks like the chance of grabbing the $50K has gone? Also, how much latent interest might there be in firing straight at the $100K prize, even with run lengths 1 year? Regards Brian Beesley Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Windoze joke
Microsoft have just released a patch for all versions of Win 9x (_all_ versions of Win 95 _and_ Win 98) which (is claimed to) resolve a problem with systems hanging after 49.7 days continuous operation. Only took them four years to find out that this might be an issue ;-) The reason it took so long is that it wasn't until now that ANYONE had Win9x run that long without rebooting. They might find the same sort of bug in Linux in, oh, about 10 years. Randy Given [EMAIL PROTECTED] http://members.aol.com/GivenRandy public key at http://members.aol.com/GivenRandy/pgpkey.asc Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Windoze joke
Microsoft have just released a patch for all versions of Win 9x (_all_ versions of Win 95 _and_ Win 98) which (is claimed to) resolve a problem with systems hanging after 49.7 days continuous operation. Only took them four years to find out that this might be an issue ;-) The reason it took so long is that it wasn't until now that ANYONE had Win9x run that long without rebooting. They might find the same sort of bug in Linux in, oh, about 10 years. Win9x is not run in, for example, serving an ISP, where Uptime is 6 months or more, like Linux is. Such a bug would probably be caught, in about 49.7 days. With a patch in a couple of weeks at the outside. Also, Win95 (in my experience) locks up so frequently, and requires reboots so often, that you would be lucky to keep it up 49.7 days, and even then, who is to say which bug caused the reboot. Farting in Win95's general direction, :-) -Lucas Wiman P.S. If these views offended you, I remind you of two rights: The right to ignore me completly, and the right to use a technologically inferior operating system if you so desire Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Factoring
I was just wondering, could anyone give me any info on how factoring is done, is there a preliminary factoring before numbers send out, how high we factor, what possible factors are, etc. and also, I would really like to see the maths behind it as well. I need something to study over summmer vac :) Chris Jefferson, Girton College, Cambridge, [EMAIL PROTECTED] Someone may have beaten me to Fermat's Last Theorem, but I've done Riemann's general equation. However it won't fit in my signature file... Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Re: Mersenne Digest V1 #575
On Sun, Jun 13, 1999 at 11:26:56PM -0600, Aaron Blosser [EMAIL PROTECTED] wrote: But hey, this is just my opinion. After I test this little teeny tiny group of numbers, I won't poach anymore and you can all do whatever, but I still think it's a good idea to "clean house" every now and then. Agreed. I'm quite comfortable trusting George and Scott will take care of it, though. Bryan -- Bryan Fullertonhttp://www.samurai.com/ Core Competency Samurai Consulting "No, we don't do seppuku." Can you feel the Ohmu call? Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: status of exponents
### Are we better off with or without these machines? ### This is the question people have been asking you and that you have been avoiding. It seems to me you are of the opinion that if a machine doesn't meet *your* standards of processor power or network connectivity, we should just tell the guy "hey, thanks, but no thanks, we've got better things to do than wait around for you." I haven't avoided the question. In fact, I think I've been quite emphatic that we need to find the work that is best suited to the computer we have. Let's face it, some computers just SHOULD NOT be running first time LL tests, even in the 4M-5M range, and here I'm talking specifically about 486 class machines. Even a P-60 could finish one of those up in less than a year. But for a 486, I'm thinking "factoring...factoring". I am in no way opposed to reclaiming abandoned exponents, but to simply start working on them without contacting the current "owner" when you have contact information available is not only irrational but unethical, regardless of the nobility of your intentions. But I don't have contact information available. I have their primenet ID which means nothing to anyone but Scott, and I don't feel like bugging him in the form of "Scott, could you email this list of 13 people and ask them "Whassup wit dat?" Who gets credit for the work is irrelevant; the fact is by doing this you are taking away their contribution (regardless of whether the number tested turns out to be prime or not) because *you* didn't think their machine was good enough for this project. Not really, because even if they finish, well hey, they just did a double-check. Again just *ask*. It's the polite thing to do. It's the RIGHT thing to do. I feel like I have to repeat this because you have not attempted to explain at all why you went ahead and started working on the exponents without asking. Okay, well I just did explain...I don't know who these people are. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: Computer speeds factoring
The other (like with Peltier junction coolers) is that they often generate as much heat as they dissipate. Often? _Always_! (Second law of thermodynamics. If you find this is violated, get a patent immediately, you have a prototype pertpetual-motion machine). Zing! Ya got me! :-) What I meant was that some people, with thermal sensors for their ambient case temp. besides one for their CPU, will see CPU temps drop a bit, but the ambient temp in the case will just skyrocket, and since Peltiers cool a certain amount below ambient temps, at some point even the CPU temp will be higher than before. In many cases, they have to overclock *less* than they did because other components can't take the increased speed (memory, peripherals, etc). Thus the need for even more case fans to rid the excess temps. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: Re: Mersenne Digest V1 #575
---Aaron Blosser [EMAIL PROTECTED] wrote: How about another option Aaron? You touch anyone's exponents...especially mine and I report you to the FBI for stealing? DAMNITI sure as hell hope the above message was in jest. You better not play around with my exponents (or anyone else's for that matter) or I'll raise hell on this list. Wow! Well, I wasn't kidding, I did grab some exponents and am testing them. I guess you'll have to call the FBI? Obviously, opinions on this matter are VERY extreme! As for you Ashton, which of those numbers in my message belonged to you? Are you still testing it and, if so, why no updates in over a year? Let me know and I'll leave you at it, but for goodness sake, if you have a machine that slow, please consider factoring or double-checking smaller exponents. It's just as important to do those tests and a slower machine is much better suited. I don't use a hammer to cut wood, I don't use a screwdriver to paint my walls, and I don't use a pocket calculator to do FFT work, but for adding 2+2, it's just fine. Get the drift? Certain tools are better suited to certain jobs. Find the job that your slower computer is best suited for and go for it. Am I just wrong in thinking this? First of all, no, none of those exponents are mine. I have tons of machines running Prime95 and I'm pretty high up on the list of producers It's just that I think you're way off base on this...and I'm sure other people on this list think the same too. whine, complain, etc. George, (or someone else!) could you please explain to Aaron why he is off base of this one.something along the lines of "Ashton has the right to set his machine to do whatever kind of work he wants it to do. It's a fun project. Don't go around annoying people, etc., etc. blah, blah." It's just not fair to take exponents that other people are working on (in the hopes of being the co-discoverer of M39) and have them find a "Exponent already tested" when their machine checks in two monthsespecially after some of the exponents you listed seemed to be more that 50% complete. Is it just me or are you just not thinking straight this past week? (Compared with your regular postings, they seem to be far more illogical!) Thanks! Ashton _ DO YOU YAHOO!? Get your free @yahoo.com address at http://mail.yahoo.com Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Windoze joke
At 06:09 AM 6/14/99 -0400, [EMAIL PROTECTED] wrote: The reason it took so long is that it wasn't until now that ANYONE had Win9x run that long without rebooting. I might have actually hit that problem and not realized it. Until recently, for many months I had my old P-120 running in another room doing essentially nothing but Mersenne (double checks most recently). It was on a UPS, so unless there was a long-term power failure, it was on all of the time. I remember at least one time when it locked up for no apparent reason. +--+ | Jud "program first and think later" McCranie | +--+ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: Re: Mersenne Digest V1 #575
First of all, no, none of those exponents are mine. I have tons of machines running Prime95 and I'm pretty high up on the list of producers It's just that I think you're way off base on this...and I'm sure other people on this list think the same too. whine, complain, etc. George, (or someone else!) could you please explain to Aaron why he is off base of this one.something along the lines of "Ashton has the right to set his machine to do whatever kind of work he wants it to do. It's a fun project. Don't go around annoying people, etc., etc. blah, blah." Well, I knew opinions would be varied on this matter...I've seen many "for" and many "against". I think what people react to the most is that I am taking it upon myself to grab certain exponents, whereas if Scott or George were to do it, they'd be more in favor of such "housecleaning". I suppose it's an indication that *I* am perhaps not as trustworthy as George and/or Scott, and there are certainly reasons why people would think that! :-) It's just not fair to take exponents that other people are working on (in the hopes of being the co-discoverer of M39) and have them find a "Exponent already tested" when their machine checks in two monthsespecially after some of the exponents you listed seemed to be more that 50% complete. Is it just me or are you just not thinking straight this past week? (Compared with your regular postings, they seem to be far more illogical!) Sheesh, well it has been a rough past week... :-) Call it "testing the waters". I'm good at that. To my knowledge, the issue of poaching numbers has never been discussed (on the list anyway), so at least we got to talk about it. We now know that some people are a bit bothered by numbers that will take 2 years to finish, and some people could care less. What we do with that knowledge is to find a happy balance, I guess. Let's see if the "ever trustworthy" Scott and George have any comments on the matter...maybe something along the lines of the automated emails sent to the REALLY slow (over a year for one number) computers. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Digest V1 #576
Mersenne Digest Monday, June 14 1999 Volume 01 : Number 576 -- Date: Sun, 13 Jun 1999 19:30:17 -0400 From: George Woltman [EMAIL PROTECTED] Subject: Re: Mersenne: status of exponents Hi all, At 11:46 AM 6/13/99 -0500, several people wrote: The expiration policy is such-and-so... Prime95 reports in with new expected completion dates as well as next expected checkin date every N days where N is by default 28 days. Your exponents are reassigned if you miss your next expected checkin date by 60 days. However, prime95 version 15 used a different algorithm. The program reported an expected completion date when an exponent was reserved. The exponent expired 60 days after the expected completion date is missed. Exponents that are checked out by email do not follow a rigid formula. Generally, progress must be reported every 4 or 5 months and the range completed within a year. People I know such as Mr. Burge and Mr. Sunde are given more slack than a name I do not recognize. These exponents are usually recycled to people who cannot upgrade from version 14 or are using Macs (which work a lot better on exponents below 4.8M). Exponents above 5.2M are given to Primenet. The "problem" exponents that started this thread almost assuredly came from version 15 clients using the old expiration policy. Is "poaching" OK? No. There's nothing I can do about it, but I certainly do not encourage it. For the past two years it seems there has always been two or three people testing the lowest available exponents. I don't think they have improved their chances at all as they often report double-checks rather than first-time checks. Oh great, I reported a result and got "exponent already tested" error. Remember, you will get this error message when you retest an exponent that was originally tested by buggy version 17. It is very likely that you just completed the first CORRECT LL test. That said, accidents happen. A manual tester can enter the wrong range, an expired exponent can all of a sudden have its result reported, etc. The system will never be perfect, but fortunately works as expected 99% of the time. You should email the original person that reserved the exponent. Well, my policy is not to give out email addresses. I'm not upgrading because I think my old P-whatever should run first time checks. GIMPS does not mandate your computer do certain work. You can override the defult behavior in the Test/Primenet dialog box. Remember, you are here to have fun and if you don't mind waiting 6 months for an LL test, that's fine by me. Your checkin every 28 days will let the server know you are busy working on your exponent. (However, be reasonable. As this thread shows if you grab an exponent that will take 4 years to test, you can expect a "poacher" to finish it before you do. I would think one year completion time would be OK though). Why do we care that these smaller exponents get tested in a timely manner? The obvious answer is we want to make a relatively orderly determination on the primality of every Mersenne number. There is no sure-fire formula for detecting exponents that are no longer being worked on. Once in a great while, I analyze the database and find exponents that have "slipped through the cracks" or I think have been abandoned and release them for reassignment. I do *not* email the affected persons (I used to, but it was *way* too much work). The good news is that the current prime95 expiration policy seems to work very well. Thus, this problem should occur much less frequently in the future. Have fun, George Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Sun, 13 Jun 1999 21:00:48 -0400 From: "Rick Pali" [EMAIL PROTECTED] Subject: Mersenne: Double-checking credits on Primenet? I just started my second machine on double-checking and am curious on whether I would expect to see my LL P90 year total increase when a double-checking assignment is turned in. Also, I notice that one can disable "Request whatever type of work makes the most sense" and then proceed to select *two* types of work that you want. How does the server decide which to send? I assume that it'll choose factoring over double-checking, double-checking over primality testing, and factoring over primality testing. Is this correct? If so, the windows versions might better be served wait radio-buttons so that only one can be selected. Rick. - - [EMAIL PROTECTED] http://www.alienshore.com/ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Sun, 13 Jun 1999 18:54:43 -0700 (PDT) From: Ashton Vaz [EMAIL PROTECTED] Subject: Mersenne: Re: Mersenne Digest V1 #575 Hi Aaron, Or this one: 4465127
No Subject
At 11:26 PM 6/13/99 -0600, Aaron Blosser wrote: This is supposed to be fun, and your behaviour makes it the oposite. I don't want to stand guard over my exponents, sending in false progress reports to make you stay away from them. Hey, whoa. I'm not asking anyone to send in false status reports. I *real* status report every now and then would do. In fact, a status report at least every 6 months seems quite prudent, don't you think? And personally, I think that if a test will take over a year to complete, you're probably better off doing factoring tests or double-checks, or maybe some other (integer based) distributed computing project altogether. I like GIMPS personally, but I'm not about to run Prime95 on my 486-75 laptop except maybe for factoring assignments. I know that factoring means you won't find the next record breaking prime, but so what? I try to run factoring assignments on my computers on the "recommended" 10:1 ratio... 1 factoring assignment for every 10 LL tests. Factoring is every bit as important to GIMPS as anything else. We're now doing first time LL tests in the 7M range...I can recall not too long ago when I would get factoring assignments in that same range, and I like knowing that I could use some of my slower machines to "pave the way" as it were. I'm not out to do all this just to get in the top-100 list...you could take away all my accumulated CPU time (though the others in my team madpoo might not like that) and that'd be fine because I'd still know that I'm contributing. Some people post to this list being upset that the work they turned in hasn't shown up in the primenet status lists yet. I know that this is a valid motivation for some people, but I do think they're missing a bigger picture. We have thousands of people all tied together into one huge, very well organized system. Scott and George have done wonders with putting this altogether. I merely suggest that we try to clean up some of the bits that ultimately will fall through the cracks. As I said before...exponents like the ones that were pointed out earlier are a very rare exception to the rule...but those exceptions must be dealt with to keep the coherency of GIMPS intact. But hey, this is just my opinion. After I test this little teeny tiny group of numbers, I won't poach anymore and you can all do whatever, but I still think it's a good idea to "clean house" every now and then. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm X-Sender: [EMAIL PROTECTED] (Unverified) X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 Date: Mon, 14 Jun 1999 09:25:53 -0400 To: "Aaron Blosser" [EMAIL PROTECTED] From: Jud McCranie [EMAIL PROTECTED] Subject: RE: Mersenne: Re: Mersenne Digest V1 #575 Cc: "Mersenne@Base. Com" [EMAIL PROTECTED] In-Reply-To: [EMAIL PROTECTED] References: [EMAIL PROTECTED] Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" At 11:26 PM 6/13/99 -0600, Aaron Blosser wrote: +--+ | Jud "program first and think later" McCranie | +--+ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: Re: Mersenne Digest V1 #575
So basically, you are on some sort of drugs or something? I hate to jump to my brother's defence here, but I think that the point was to prove 100% that M37 was really M37 and not M38 or M39 or whatever, thus being able to get rid of the '?' on the GIMPs page. :) I think that in the spirit of a "team" effort, assigning an LL test to a 486 is the stupidest and most incosiderate thing a person could do. As far as the GIMPS effort goes, if your LL test is going to take a year, then perhaps you should do a factoring or double check assignment. As far as other efforts go, sure, we could form our own GIMPs like project and have competing projects, but I don't think thats a good idea. For example, when I was coding for distributed.net (RC5/DES), and were going for the DES-II stuff, and the EFF built their super-duper DES cracking machine, it seemed kinda unfair that they got the exponent before we did. Then for DES-III, they figured out a way to have their machine coordinate with the dist.net servers. So the whole effort was organized, and the whole range of keys went by much faster. Now, lets say for example, that the EFF decides they want to build a "Mersenne finding machine", which has 56,000 processors all doing LL tests, and they go right on past your exponent you have had checked out for over a year... would you feel slighted by the EFF? Give me a break folks, you are being way too anal about your assignments. I think it is quite clear to me that George and Scott missed a lot of exponents that slipped thru the cracks and my bro just decided to point that out in his own weird way. I might also add that when I saw my UltraEnterprise 4000 was doing something like .9s/iter, I thought, "Hmmm... thats no good". So I recompiled using Suns C compiler and increased my performance to a respectable .5s/iter so, I suggest that for the "Good of the team", it would be better to use the right tool for the right job... Or should I go out and write my LL tester for my HP calculator? Another note: You may have noticed that I worked on a Java LL tester in the past, and I decided not to release it until it had respectable speeds for this exact reason. Having 1s/iter wasn't good enough for me. Sorry if you don't like that. Lastly, what is the deal with reporting my bro to the FBI for "stealing"? What, he stole your assignment? You know, there is a law against making false charges against someone. Best regards, Jeremy P.S. This whole argument is stupid, so why not just drop it. You can't really stop anyone from poaching, so are gonna go cry to George and Scott and say "Aaron stole my exponent which I've been working on for 3 yrs". I would say, "Do some factoring next time stupid!" But, really, I could care less if George and Scott decided to take team madpoo off the list for poaching or whatever (We're almost to 33). The whole idea is to go forward in the name of science, not "I want to reserve these 100 exponents cuz their small and I might find M37 and get my name in the history books". As a matter of fact, if Aaron finds a prime in his poached exponents, I'm sure he'll be glad to share credit with whoever had it checked out or whatever. Or for that matter, he would probably just give that person the full credit and not want any credit at all really. So if its your "Place in history" you are worried about, I don't really care. -Original Message- From: Ashton Vaz [mailto:[EMAIL PROTECTED]] Sent: Sunday, June 13, 1999 8:55 PM To: [EMAIL PROTECTED] Subject: Mersenne: Re: Mersenne Digest V1 #575 Hi Aaron, Or this one: 4465127 60 472.3 311.8 371.8 26-Feb-98 09:23 koma magek072 Checked out 2/26/98, *NEVER* checked in at all, over a YEAR until it will expire. There are quite a few like that, so I'm gonna play God and take care of 'em. :-) Here's one I just *love*: 4787599 61 376.0 662.0 722.0 02-Jun-98 16:42 andres We could wait around 2 years to finally get around to testing this obviously abandoned one, or I'll just do it now. How about another option Aaron? You touch anyone's exponents...especially mine and I report you to the FBI for stealing? DAMNITI sure as hell hope the above message was in jest. You better not play around with my exponents (or anyone else's for that matter) or I'll raise hell on this list. And now back to George's regular GIMPS programming on my only computer, my beloved P100! Later all. Ashton _ DO YOU YAHOO!? Get your free @yahoo.com address at http://mail.yahoo.com Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: Re: Mersenne Digest V1 #573
Since the agner.org server seems to be down, if you want a copy of this doc, e-mail me and I'll e-mail it to you (Don't want to waste massive bandwidth). I'd mirror it, but my stupid internet provider only allows me to upload via their dial-in (doh!). Oh well. ---snip--- There's the good doc at: http://www.agner.org/assem/pentopt.htm which explains all this stuff better than I could ever hope to. You gave me this link too; it gives a 404. :-) Time for a mirror? ---snip--- mov al, 0 Would it be a big problem replacing this with eax, etc.? Generally, 16-bit stuff isn't good for P6, although I can't see a direct partial stall in your code. ---snip--- Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: All the stuff about slow computers etc.
My policy is to put all slower computers and those which aren't on 24/7 (such as computers at school) to work for distributed.net. I find they can do a lot more useful work cracking rc5 than they would do sitting on Mersenne exponent. Of course this is just my opinion, as I'm not so fond of maths as I'm of distributed computing, and I realize there are a lots of people who want to contribute all of their power to research instead of "useless" code cracking. But this really is an option to consider, as the exponents are getting larger and larger. --Jukkis Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: These go to 11 (WAS: blahblah...)
Actually, your a need not be a positive integer. Any non-zero Real, Complex, quaternion, Hamiltonian, Cayley or Sylvester number will do... among others... JT - Original Message - From: Markus Laire [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Monday, June 14, 1999 1:03 AM Subject: Mersenne: These go to 11 (WAS: blahblah...) or, more concisely, (1+1+1)^(1+1) + 1. Can anyone represent that number in fewer than (1+1+1)! ones? This all depends on what operators and notations are accepted and without specifying that, the whole question is useless. What about without any ones at all: (With C++ operators) ((0++)++)*(0++)++)++)++)++) or without any numbers: (With normal algebra) ((a/a)+(a/a))*((a/a)+(a/a)+(a/a)+(a/a)+(a/a)), a [belongs to] N -- Markus Laire [EMAIL PROTECTED] ICQ# 11887013 http://www.nic.fi/~laire/english or http://come.to/markuslaire PGP Key: 4096/1024 DH/DSSID: 0xB93CD277 Fingerprint: 7C6B AE89 C243 F5A4 9702 EDAB 19ED 59B0 B93C D277 Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Ground Rules for Poaching ?
Original Subject: RE: Mersenne: Re: Mersenne Digest V1 #575 This strand is getting kinda emotional. So we ask the director, "What is my motivation?" Hmm. I do GIMPS because it is fun. I am not a big contributor but I have a total of 5 machines running. Two months ago, I had two machines doing LL tests and this month I find that only one machine is doing an LL test. The rest are doing double-checks... at least while windoze is up and crawling. We just got hot weather (for us, anyway) and two of my slowest machines seem to be suffering heat-stroke but that is hardware rather than software. Even double-checking 3MM exponents is taking about 2 weeks on the fastest of these machines. I would say I am disappointed to see my 2nd fastest machine double-checking. On the other hand, I do not expect to draw a Mersenne prime in this lottery. I do want to see more primes discovered. Sure, I would like to say "(co-)discoverer of Mnn" in my digital signature, but the fact is that I am not THE discoverer, I just ran a program that George and company put together. When there is money or fame on the table, then that raises all sorts of value questions for people. I might even call it greed in myself. For the collective progress of GIMPS, I really like the idea of having some targets for 12/31/99 that can be stated fairly firmly, like tested up to 7.5MM and double-checked up to 5.0MM (or whatever the numbers are). The problem is that there may be a lot of machines out there (including some of mine) that will start something by or before September that will not be expected to finish by year-end. This could include some first-testing of exponents that yield a prime or even double checking that turns up a prime missed on the first pass (clearly only possible by a highly improbable concatenation of errors). So how about some administrative ground rules to make us all "happy" or at least aware of risks? Here is one thought for "poaching rules" for your consideration: first it only really matters if a prime turns up. So, for purposes of achieving a target for year-end, someone -- perhaps Scott -- will check the database for several sorts of holes. The highest priority will be exponents that have not been assigned, of course. Second priority is exponents assigned but not expected to be completed before some date like 10/1/99. These late-bloomers are a risky collection: they may or may not finish in time depending on hardware performance. Third is exponents with a really long time since assignment and last report. (A long-running exponent for someone that is expected to complete by 10/1 would not concern me at this time at all, even if it has been running for two years.) First, the GIMPer originally assigned an exponent "owns" the exponent. Someone who "poaches" that exponent -- starts testing the exponent before the first run completes -- is in fact double-checking the exponent even if this double-check finishes before the first LL test finishes. This may make lots of problems for the database. The real trick is in notifying the exponent owners. Suppose this notification starts in August. They should be notified that double-checking their exponents is underway and asked to respond, say within a month to verify that they are still running the first LL test (if this is not known). Second notices should go out about a month later to update those who responded and to ask again of the non-responders. If no new Mersenne primes turn up during this organized "poaching" then no harm will be done to anyone that I can see. However, there can be a real problem if the double-checking turns up a prime. Here, the owner of the exponent must complete the LL test without being told that there is a prime waiting!! If the owner's LL test somehow fails, then the double-checker gets the credit (and may have to wait a LONG time to see anything public). If the owner appears to drop out of GIMPS without completing the LL test, we face a crisis of conscience: do "we" tell this person that they need to complete this test or let them walk away? If they really want to drop out, do we tell them that they "owned" a winning exponent and must complete the LL test to get public credit for it? I suspect that the only way to preserve any confidentiality is to treat all of the slow cases and potential drop-outs the same and make them take a positive action to be active or to drop-out. The alternative is a passive acceptance of being dropped out. ...And I definitely do NOT like the idea of telling someone that they are dropped by GIMPS. That is really raw. But we still have a problem with an "owner" of a Mersenne prime who fails to respond to anything. At some point, that person must lose title perhaps after GIMPS makes a good faith effort to track them down. I think that this is the only real problem: does GIMPS have an obligation to someone who has been out of touch for over a year (or whatever period seems
Re: Mersenne: Poaching Exponents...
"Blosser, Jeremy" wrote: Okay, so apparently there are some people out there crying because someone might "steal" their exponents (and turn people into the FBI) I think my brother brought up a valid point, that being that there are exponents that people are "sitting" on, and it is holding up the verification of M37. And in his own way decided to bring the whole argument to a head by "poaching" some exponents that have been checking out for over a year for PRIMARY LL testing... My thought is that I don't really mind someone sitting on an exponent, but if you have some 386 doing a primary LL test on a number and its going to take over a year to accomplish, and its not connected to the internet but once a year... then maybe you should assign it a double check or a factoring test. It is the most *considerate* thing to do as far as the whole team approach is concerned. Really, the thought that someone wants to sit on an exponent for over a year because "I've checked it out, and I'm running it on my 386 that is connected to the internet once a year" is more inconsiderate than someone poaching your exponent. As far as poaching is concerned, I unlike my brother Aaron, think that George and Scott need to release whatever exponents, since there are only a dozen or so that have seemed to have slipped thru the cracks... Notice in the email from George on Sunday that he does do this, also it isn't really a few exponents holding up proving that m37? is m37. There are a lot on numbers still being tested, at least wait until it really is down to a few exponents holding things up(ones that look abandoned). Any checking of exponents that are not actually abandoned wastes time, you could be testing higher exponents instead of duplicating work someone else is doing, that will be double checked anyway. I guess I'm just annoyed at the "?" next to M37... Lastly, I think that threatening to falsify primenet reports, or report someone to the FBI for "stealing" is really sophomoric. And I've noticed this has come from the anti-poaching camp... I suggest Aaron stop his poaching, that Scott and George work on getting all the exponents up to M37 at least have a primary LL check, and for those who are whining about someone poaching your exponents, try and think beyond the scope of yourself and think about the GIMPS effort as a whole. Try putting the machine to its best use (Factoring as opposed to LL testing). If you are so concerned about becoming "famous" for finding a new Mersenne Prime, spend the extra $200 or so and get a P233 which at least would finish its testing in 4 days (in the 3-4M range). Otherwise I suppose that it would be okay for me to run my GIMPS client on my toaster oven, in which case, I think it might finish its 700,000 range LL test in 2103... :) Later, Jeremy Blosser (Note: I am Aaron's BROTHER, not Aaron. Please try not to get confused by the fact that we have the same last name (I know thats a foriegn concept to some people) and send me nasty e-mails, it just makes you look stupid, and annoys me...) Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Re: Mersenne Digest V1 #575
"Aaron" == Aaron Blosser [EMAIL PROTECTED] wrote the following on Sun, 13 Jun 1999 22:16:28 -0600 Aaron Certain tools are better suited to certain jobs. Find the Aaron job that your slower computer is best suited for and go for Aaron it. I think this argument also applies to your computers...why not put your faster computers on the bigger exponents where they are needed, instead of poaching smaller exponents? Aaron Am I just wrong in thinking this? I think that most people's concern is that they will have their exponent "stolen" from under them. It's not clear exactly how you will go about determining which exponents need to be reclaimed, since you're sort of doing it unofficially. I know that you have the best of intentions in mind, and it's clear that the exponents you picked are indeed dropped, but that's just because I happen to recognize your name. However, it's not TOO big a step to imagine someone else getting even more anxious and 'reclaiming' exponents that are still actively being worked on. This is the slippery slope that most people fear. If there's no official policy on exponent reclaiming, and just many little ad hoc ones, the whole concept of reserving exponents goes out the window. If I were you, I'd let George and Scott determine which exponents need to be reclaimed. (Perhaps you could offer them suggestions :) Getting their official blessing before doing work would make everybody more at ease. Mike Gebis Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: LL and factoring quitting
On 14 Jun 99, at 6:41, lrwiman wrote: We will of course have to check factors considerably further than we are doing on our current exponent range (due to the increased LL iteration time.) Yes - on the principle that it's worthwhile to spend 5% to 10% of the LL testing time attemptimg to find at least one factor before we run a LL test, the pre-factoring should be run for 3 or 4 weeks per exponent first (assuming PIII-500 class power). At a rough guess, that's up to 2^66 or 2^67, but we don't have any benchmarks yet. think that someone (Brian...?) did test for all 10,000,000 digit primes 36,000,000 for factors 2^40. Maybe George can add these to his database, or something. George advised "don't bother" but I did it anyway. Just for fun. But I checked the results carefully enough to be sure I'm not making an idiot of myself. You can find the results in standard format in the /gimps/DecaMega directory on my anon ftp server lettuce.edsc.ulst.ac.uk Of course, trial factoring to 2^40 is very quick - I spent only about 2 (PII-350) CPU hours spread over all 159,975 candidate exponents in the range I selected. But that's enough to eliminate a third of them. If anyone's _really_ keen I could send them the source of the program I used (needs MS VC++). It's reasonably efficient will go to 2^63. Exponents in the 30 millions are not acceptable to Prime95 v18 and its derivatives. The factors found were verified against a much simpler program run on an Alpha I have lots of confidence in them. Regards Brian Beesley Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
RE: Mersenne: Re: Mersenne Digest V1 #575
At 07:40 AM 6/14/99 -0600, you wrote: Sheesh, well it has been a rough past week... :-) Call it "testing the waters". I'm good at that. To my knowledge, the issue of poaching numbers has never been discussed (on the list anyway), so at least we got to talk about it. We now know that some people are a bit bothered by numbers that will take 2 years to finish, and some people could care less. What we do with that knowledge is to find a happy balance, I guess. Let's see if the "ever trustworthy" Scott and George have any comments on the matter...maybe something along the lines of the automated emails sent to the REALLY slow (over a year for one number) computers. That was exactly my first suggestion: send an email to those who have not show up in the last year and ask them if they are still working on them. Personnaly, I will 'poaches' those number only if they their progression is a way behind the tail of the distribution. By the way, I order to share the negative reaction with Aaron, I checked again for the snail (a snail is a calculation with an to go date larger than 365). Last time I did it (29 sept 98) there was 80 snails, now there is 135. However, since we check about twice the number of exponent than we were checking at that time, this should be consider as a progress. This is 0.3% of the overall assingment. Alos of the 80 'snails' of the last year, only 20 survive up to now the others have be checked by someone else. In those survivor, only 7 are expected to survive another four months. There is also four silent runner in this list. One weird bahavior: 5292757. Which as appear to have been resetted recently by its 'owner' to an duration as long as the initial one! Maybe it is the imfamous NSA which try to hide their own prime :) Surviving 'snail' prime fact current days exponentbits iteration run / to go / exp date updated date assigned account ID computer ID -- - - --- --- -- 4369949 61393215 476.5 15.1 75.1 14-Jun-99 11:53 23-Feb-98 02:21 madpoo cosy2k-1 4465127 61458752 473.2 14.2 74.2 13-Jun-99 19:01 26-Feb-98 09:23 madpoo cdry2k-2 4580201 61405780 481.5 15.1 75.1 13-Jun-99 17:11 18-Feb-98 01:42 madpoo cosy2k-4 4664917 60 424.3 349.6 409.6 11-Jul-98 05:09 16-Apr-98 07:26 markb P-90 4771433 61 48.5 9.5 64.5 12-Jun-99 03:37 27-Apr-99 03:45 sgrupp ocap 4774129 61 49.3 -11.3 48.7 26-Apr-99 06:03 Ri auweia 4787599 61 376.9 661.1 721.1 02-Jun-98 16:42 andres 4833901 61 404.9 404.1 464.1 05-May-98 16:35 andrewftom1 4933207 61 427.8 -55.0 5.0 24-Jan-99 12:49 12-Apr-98 18:53 kampelakampela 5016679 61 388.5 380.5 440.5 22-May-98 02:09 redwine 5103829 61 2292622 396.0 127.7 17.7 05-Apr-99 07:13 14-May-98 14:27 Vesa 5292757 62 71.5 -55.5 4.5 04-Apr-99 01:16 S06537 5293451 6231059028.8 1775 57.0 10-Jun-99 15:50 16-May-99 20:38 DOCKING 5362403 62 1725634 345.6 399.3 72.3 26-May-99 20:27 04-Jul-98 00:30 tp68spa 5635247 62 17.3 6.7 66.7 28-May-99 06:15 glcross 5834261 62 64.3 -2.5 37.5 25-Apr-99 02:04 11-Apr-99 07:54 kosh jobb_2 5920043 62 249036760.3 -13.0 47.0 13-May-99 13:25 15-Apr-99 06:29 kevinrosenberg Ferret 5963843 62 340787233.7 12.4 72.4 08-Jun-99 23:29 11-May-99 22:11 GScharfSAI 5977801 62 2196233 264.3 373.2 84.2 10-Jun-99 18:12 23-Sep-98 08:21 S02408 5996143 62 17.3 42.7 102.7 28-May-99 06:17 glcross Situation in September 1998 4369949 60 218.7 480.8 98-May-22 12:13 98-Feb-23 02:21 Walkabout BamBam 4465127 60 215.4 568.698-Feb-26 09:23 koma magek072 4580201 60 223.7 826.6 98-Feb-25 09:06 98-Feb-18 01:42 MITrhansen2 4664917 60 166.5 607.4 98-Jul-11 04:09 98-Apr-16 06:26 markb P-90 4771433 61 169.7 835.398-Apr-13 01:30 x69-irc2 IRC-USER 4774129 61 196608 64.4 506.2 98-Aug-10 22:41 98-Jul-27 07:45 josephg 4787599 61 119.1 918.998-Jun-02 15:42 andres 4833901 61 147.1 661.998-May-05 15:35 andrewftom1 4933207 61 170.0 385.0 98-Sep-29 17:41 98-Apr-12 17:53 kampelakampela 5016679 61 130.7 638.398-May-22 01:09 redwine 5103829
Re: Mersenne: new frost free computers
Aaron Blosser wrote: One nut is working on total immersion of his system in oil, with an air-conditioner coil submersed as well. This would solve the problem of condensate, but there is concern that the mineral oil will break some of the components on the board. I like the idea, but instead of mineral oil, some inert water. Pure nitrogen would not react to anything and would carry that heat around in the case pretty much the same as dry air For that matter if you're just using dry air the only condensation will be on the cooling coils, because everything else will be too hot. Maybe several sets of cooling coils that can swap in and out of the insulated chamber and chip the ice off on the outside. Anyone want to search www.patents.ibm.com for "refrigeration" and hmm, what other terms -- condensation will get most anything that refrigerates -- same with "moisture" David Nicol 816.235.1187 UMKC Network Operations [EMAIL PROTECTED] Proud to use and endorse the "last used on top" filing system Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: To poach or not to poach, that is the question.
From the GIMPS page: "You could be the first person to find a 1,000,000 digit prime number! However, these exponents take quite a while to test. A 200 MHz Pentium computer will take 4 weeks to test a single exponent! These ranges are definitely not recommended for 486 or Cyrix 6x86 owners." "Slower Pentiums (120MHz and below) can be used in double-checking previous Lucas-Lehmer tests. You could find a new Mersenne prime if the previous Lucas-Lehmer test was incorrect. You must use version 18 of the program and use the PrimeNet server's web pages to get double-checking assignments." "486/100s or better and Cyrix 6x86s can be used in factoring Mersenne numbers. You won't find any Mersenne primes this way, but your results will be used to update the database to help speed up the Lucas-Lehmer testing. You must use the PrimeNet server's web pages to get factoring assignments." And there IS a reason why, left to it's own devices, Primenet will hand out double-checks to machines P166 and under, and why 486 machines will get factoring assignments. Scott must have considered the "suitability to task" when deciding on these limits. And as we can see, George has expressed his opinion also on what machines are suitable for what type of work. We're all free to check out first time LL tests in the 7M range and run it on our 486DX-40, but personally I think that is a ridiculous concept. While every machine *can* be useful for GIMPS, they are useful for different things...486's are fine for factoring, P166 and below are fine for double-checking the smaller numbers. On the other hand, there is a point where a machine is no longer useful, and I would certainly lump 386 class machines into this category. At some point, factoring will take to long for 486's under 120MHz or so, and even eventually, double-checking assignments will exceed the reasonable limits of Pentium class machines. I certainly think that if ANY assignment takes longer than 12 months, you'd be better off on a codebreaking contest or something of the sort. Something to consider...maybe even ECM or the like. Wouldn't it be nice to go out of 1999 having done first time LL tests on all exponents under 5.26M? C'mon folks, set goals for yourself! Currently, there's only about 210 exponents under this (which will finish off the 256K FFT size numbers). And I'd like to be able to prove M37 and M38 are actually 37 and 38! There's just under 7,000 exponents to double-check (according to the GIMPS status page) to prove this...so get those slower machines cracking on the double-checks! It's just as important as doing first time LL tests if you ask me. :-) And I vow not to poach more numbers, though I'd hope that George and Scott would take the matter of "exponent hoarders" into consideration. Consider this nasty response my brother and I got from a name withheld: "I'm starting to realise that you probably are the most stupid and ignorant person I have ever discussed anything with. What makes you think someone are using a 486 for testing, even if Primenet tells you that it would take two years to finish one exponent? You don't _know_, and you do not intend to ask, because you have no repsect for any ideas other than those you find in your own little world." Well, he/she is right...I don't know what kind of machine it's running on, just because it'll take 2 years to test oen exponent. In fact, I'd be quite wrong if I said it was a 486 since a 486 would finish it up MUCH faster in all likelihood (I mean, we are talking about numbers in the 4M-5M range). Maybe it is a 386, or a pocket calculator. :-) And, FWIW, ad hominem attacks are never useful...consider cutting out the words like "stupid" and "ignorant" if you really want to make a point. Or these fun quotes: "Well, since you are stupid, I'll keep two of my assignments below M37, which are reserved directly from George, outside Primenet. I might even return bogus results for them sometime next year. They will not be tested properly until they had a third check sometime in perhaps 2002. Don't that make you feel smart, while you take the fun out of the project? What the heck, if you want to kill the fun for other people, I'm free to take your fun away, right? I can play really dirty too, if I want to. "I have a 486DX40, and it needs half a year to complete a factoring assignment, so I find it hard to belive that it can do a LL-test in only four times longer, but I'll reserve one and try just to nag you. Nah, thinking about it, I think I'll use my old 386DX20 instead. I can tell it to pretend to be a PIII and see how long it takes until someone figures." Well, obviously this person is trying to make their point by stooping even lower than he/she accuses me of being. Again, not a great way to make a point, but very helpful in understanding this person's mindset. That's *all* we'd need is for this person to start sending in bogus reports on the residue of his/her numbers. Very adult,
RE: Mersenne: Re: Mersenne Digest V1 #575
Aaron Certain tools are better suited to certain jobs. Find the Aaron job that your slower computer is best suited for and go for Aaron it. I think this argument also applies to your computers...why not put your faster computers on the bigger exponents where they are needed, instead of poaching smaller exponents? Indeed, I usually do this...I just set every machine to automatically get whatever type of job from Primenet. I'm not kidding myself...these PPro 200's I have will only take longer and longer to do LL tests and at some point Scott will set the limit for first time LL tests to maybe a 233MHz machine, and at that point I'll be happy to let my PPro 200's get double-check assignments. As for these numbers, it just so happens that I have some test machines for a limited time, and whaddya know, these 4-5M exponents are small enough that I can finish them up in time before these machines go bye-bye. Otherwise, they'd be doing first time LL tests also. Aaron Am I just wrong in thinking this? I think that most people's concern is that they will have their exponent "stolen" from under them. It's not clear exactly how you will go about determining which exponents need to be reclaimed, since you're sort of doing it unofficially. I know that you have the best of intentions in mind, and it's clear that the exponents you picked are indeed dropped, but that's just because I happen to recognize your name. For the official record: I got the assignments report from Primenet, popped it into a spreadsheet, sorted by when the exponent was assigned, picked the very oldest assignments then checked for when they had last checked in. If they hadn't checked in for over a year, then I further looked at how long before Primenet would expire that number. What I ended up with was a handful of maybe a dozen or so exponents that were checked out over a year ago, had not checked in for nearly a year (about half had NEVER checked in) and still had over 300 days before Primenet would have expired them (one would not have expired for nearly 2 more years!!). Again, I must stress that Primenet will currently expire numbers on a much more rapid schedule, so this problem has been addressed and will go away, but these are the worst numbers I could find that even the most vocal opponents of poaching would probably admit are abandoned numbers. I can only say that I won't be poaching any more numbers *because* after these are done, there really won't be any more "bad" ones. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Digest V1 #577
Mersenne Digest Monday, June 14 1999 Volume 01 : Number 577 -- Date: Mon, 14 Jun 1999 08:51:12 -0400 From: Bryan Fullerton [EMAIL PROTECTED] Subject: Re: Mersenne: Re: Mersenne Digest V1 #575 On Sun, Jun 13, 1999 at 11:26:56PM -0600, Aaron Blosser [EMAIL PROTECTED] wrote: But hey, this is just my opinion. After I test this little teeny tiny group of numbers, I won't poach anymore and you can all do whatever, but I still think it's a good idea to "clean house" every now and then. Agreed. I'm quite comfortable trusting George and Scott will take care of it, though. Bryan - -- Bryan Fullertonhttp://www.samurai.com/ Core Competency Samurai Consulting "No, we don't do seppuku." Can you feel the Ohmu call? Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Mon, 14 Jun 1999 07:22:49 -0600 From: "Aaron Blosser" [EMAIL PROTECTED] Subject: RE: Mersenne: status of exponents ### Are we better off with or without these machines? ### This is the question people have been asking you and that you have been avoiding. It seems to me you are of the opinion that if a machine doesn't meet *your* standards of processor power or network connectivity, we should just tell the guy "hey, thanks, but no thanks, we've got better things to do than wait around for you." I haven't avoided the question. In fact, I think I've been quite emphatic that we need to find the work that is best suited to the computer we have. Let's face it, some computers just SHOULD NOT be running first time LL tests, even in the 4M-5M range, and here I'm talking specifically about 486 class machines. Even a P-60 could finish one of those up in less than a year. But for a 486, I'm thinking "factoring...factoring". I am in no way opposed to reclaiming abandoned exponents, but to simply start working on them without contacting the current "owner" when you have contact information available is not only irrational but unethical, regardless of the nobility of your intentions. But I don't have contact information available. I have their primenet ID which means nothing to anyone but Scott, and I don't feel like bugging him in the form of "Scott, could you email this list of 13 people and ask them "Whassup wit dat?" Who gets credit for the work is irrelevant; the fact is by doing this you are taking away their contribution (regardless of whether the number tested turns out to be prime or not) because *you* didn't think their machine was good enough for this project. Not really, because even if they finish, well hey, they just did a double-check. Again just *ask*. It's the polite thing to do. It's the RIGHT thing to do. I feel like I have to repeat this because you have not attempted to explain at all why you went ahead and started working on the exponents without asking. Okay, well I just did explain...I don't know who these people are. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Mon, 14 Jun 1999 07:26:59 -0600 From: "Aaron Blosser" [EMAIL PROTECTED] Subject: RE: Mersenne: Computer speeds factoring The other (like with Peltier junction coolers) is that they often generate as much heat as they dissipate. Often? _Always_! (Second law of thermodynamics. If you find this is violated, get a patent immediately, you have a prototype pertpetual-motion machine). Zing! Ya got me! :-) What I meant was that some people, with thermal sensors for their ambient case temp. besides one for their CPU, will see CPU temps drop a bit, but the ambient temp in the case will just skyrocket, and since Peltiers cool a certain amount below ambient temps, at some point even the CPU temp will be higher than before. In many cases, they have to overclock *less* than they did because other components can't take the increased speed (memory, peripherals, etc). Thus the need for even more case fans to rid the excess temps. Aaron Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Mon, 14 Jun 1999 06:11:03 -0700 (PDT) From: Ashton Vaz [EMAIL PROTECTED] Subject: RE: Mersenne: Re: Mersenne Digest V1 #575 - ---Aaron Blosser [EMAIL PROTECTED] wrote: How about another option Aaron? You touch anyone's exponents...especially mine and I report you to the FBI for stealing? DAMNITI sure as hell hope the above message was in jest. You better not play around with my exponents (or anyone else's for that matter) or I'll raise hell on this list. Wow! Well, I wasn't kidding, I did grab some exponents and am testing them. I guess you'll have to call the
Re: Mersenne: Poaching (was Mersenne Digest V1 #573)
lrwiman wrote: *never ever* cheap out on power supplies. This is good advice, but personally I have never seen the point of giving every machine in a rack of computers its own power supply rather than having one big one and just running DC all the way up the rack. The fact that it is not done that way seems to be about politics of having AC wall current rather than engineering efficiency. Why not have a single (redundant) big 24VDC power supply for all the boards instead of supplying them all 120VAC? Would there be a commercial market for such "unicluster" devices, where multiple independent boards are associated with a single power supply transformer --- or maybe an alternate power standard connector which would provide computer voltages to computer equipment instead of 120 VAC? Or is the current state of distributed power supplying really a best practice because it prevents loading problems and line resistance problems that running DC lines around the computer room (using Edison wiring rather than a Tesla wiring) would cause? Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Inane Stuff (Was: Mersenne: M38, SETI, and other random stuff )
Chris Nash wrote: maybe every electronic device in my house will be squaring and subtracting 2 in its idle time. voice character="futurist" aspect="tut-tut" make that every stitch in your clothing voice David Nicol 816.235.1187 UMKC Network Operations [EMAIL PROTECTED] Proud to use and endorse the "last used on top" filing system Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Re: Poaching
At 11:29 AM 6/14/99 -0500, Paul Becker wrote: I've got a slower computer working on this project, and I've been involved for perhaps two years. My machine IS contributing to the project 24/7, and I'd like that to continue. How long is it taking per exponent? What % complete is it for the one it is working on? +--+ | Jud "program first and think later" McCranie | +--+ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Re: factoring 10^7 digits (was LL and factoring quitting)
Brian, We will of course have to check factors considerably further than we are doing on our current exponent range (due to the increased LL iteration time.) Yes - on the principle that it's worthwhile to spend 5% to 10% of the LL testing time attemptimg to find at least one factor before we run a LL test, the pre-factoring should be run for 3 or 4 weeks per exponent first (assuming PIII-500 class power). At a rough guess, that's up to 2^66 or 2^67, but we don't have any benchmarks yet. Sounds about right for a SWAG estimate. Of course, trial factoring to 2^40 is very quick - I spent only about 2 (PII-350) CPU hours spread over all 159,975 candidate exponents in the range I selected. But that's enough to eliminate a third of them. Maybe we should start checking factors in (the range of prime numbers used in your database) in 2^40 segments, between you and me (and others?). If anyone's _really_ keen I could send them the source of the program I used (needs MS VC++). It's reasonably efficient will go to 2^63. Exponents in the 30 millions are not acceptable to Prime95 v18 and its derivatives. Just call me "Mr. Keen." ;) When you send me the source, I'll try to port it GCC (then maybe we can make something useful out of it :-) A few questions: What type of sieving did you do (if any) ? Did you write the modpow routine, or was it included with some math header? How much room for improvment in speed would you say is there in the source? I have limited computational resources, a PII233, a P100, but because of my work, I have practically unlimited resources in the 486 department they could finish a 2^40 section in about a day or so. I hope this 10,000,000 digit thing will go somewhere. It would be keen to have this range checked in maybe 2 years, attracting many new members to GIMPS in the process. -Lucas Wiman Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Digest V1 #578
Mersenne Digest Monday, June 14 1999 Volume 01 : Number 578 -- Date: Mon, 14 Jun 1999 11:21:46 -0700 From: Michael Gebis [EMAIL PROTECTED] Subject: Re: Mersenne: Re: Mersenne Digest V1 #575 "Aaron" == Aaron Blosser [EMAIL PROTECTED] wrote the following on Sun, 13 Jun 1999 22:16:28 -0600 Aaron Certain tools are better suited to certain jobs. Find the Aaron job that your slower computer is best suited for and go for Aaron it. I think this argument also applies to your computers...why not put your faster computers on the bigger exponents where they are needed, instead of poaching smaller exponents? Aaron Am I just wrong in thinking this? I think that most people's concern is that they will have their exponent "stolen" from under them. It's not clear exactly how you will go about determining which exponents need to be reclaimed, since you're sort of doing it unofficially. I know that you have the best of intentions in mind, and it's clear that the exponents you picked are indeed dropped, but that's just because I happen to recognize your name. However, it's not TOO big a step to imagine someone else getting even more anxious and 'reclaiming' exponents that are still actively being worked on. This is the slippery slope that most people fear. If there's no official policy on exponent reclaiming, and just many little ad hoc ones, the whole concept of reserving exponents goes out the window. If I were you, I'd let George and Scott determine which exponents need to be reclaimed. (Perhaps you could offer them suggestions :) Getting their official blessing before doing work would make everybody more at ease. Mike Gebis Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Mon, 14 Jun 1999 19:36:51 +0100 From: "Brian J. Beesley" [EMAIL PROTECTED] Subject: Re: Mersenne: LL and factoring quitting On 14 Jun 99, at 6:41, lrwiman wrote: We will of course have to check factors considerably further than we are doing on our current exponent range (due to the increased LL iteration time.) Yes - on the principle that it's worthwhile to spend 5% to 10% of the LL testing time attemptimg to find at least one factor before we run a LL test, the pre-factoring should be run for 3 or 4 weeks per exponent first (assuming PIII-500 class power). At a rough guess, that's up to 2^66 or 2^67, but we don't have any benchmarks yet. think that someone (Brian...?) did test for all 10,000,000 digit primes 36,000,000 for factors 2^40. Maybe George can add these to his database, or something. George advised "don't bother" but I did it anyway. Just for fun. But I checked the results carefully enough to be sure I'm not making an idiot of myself. You can find the results in standard format in the /gimps/DecaMega directory on my anon ftp server lettuce.edsc.ulst.ac.uk Of course, trial factoring to 2^40 is very quick - I spent only about 2 (PII-350) CPU hours spread over all 159,975 candidate exponents in the range I selected. But that's enough to eliminate a third of them. If anyone's _really_ keen I could send them the source of the program I used (needs MS VC++). It's reasonably efficient will go to 2^63. Exponents in the 30 millions are not acceptable to Prime95 v18 and its derivatives. The factors found were verified against a much simpler program run on an Alpha I have lots of confidence in them. Regards Brian Beesley Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm -- Date: Mon, 14 Jun 1999 14:35:25 -0400 From: Yvan Dutil [EMAIL PROTECTED] Subject: RE: Mersenne: Re: Mersenne Digest V1 #575 At 07:40 AM 6/14/99 -0600, you wrote: Sheesh, well it has been a rough past week... :-) Call it "testing the waters". I'm good at that. To my knowledge, the issue of poaching numbers has never been discussed (on the list anyway), so at least we got to talk about it. We now know that some people are a bit bothered by numbers that will take 2 years to finish, and some people could care less. What we do with that knowledge is to find a happy balance, I guess. Let's see if the "ever trustworthy" Scott and George have any comments on the matter...maybe something along the lines of the automated emails sent to the REALLY slow (over a year for one number) computers. That was exactly my first suggestion: send an email to those who have not show up in the last year and ask them if they are still working on them. Personnaly, I will 'poaches' those number only if they their progression is a way behind the tail of the distribution. By the way, I order to share the negative reaction with Aaron, I checked again for the snail (a snail is a calculation with an to go date larger than 365). Last
Mersenne: Re: factoring 10^7 digits (was LL and factoring quitting)
We will of course have to check factors considerably further than we are doing on our current exponent range (due to the increased LL iteration time.) Yup. And don't forget that the larger the exponent, the fewer the possible factors in a given range (e.g., from 0 to 2^40 or 0 to 2^63). Yes - on the principle that it's worthwhile to spend 5% to 10% of the LL testing time attemptimg to find at least one factor before we run a LL test, the pre-factoring should be run for 3 or 4 weeks per exponent first (assuming PIII-500 class power). At a rough guess, that's up to 2^66 or 2^67, but we don't have any benchmarks yet. Oh, but we do. George's factoring is not the only one out there. In particular, I maintain, as part of the mers package, factoring programs that use the freeLIP and gmp libraries to try arbitrarily large factors. Since George's program uses the Intel CPU's 64 bit mantissa floating point operations, it is limited to checking possible factors up to about 2^63. Of course, trial factoring to 2^40 is very quick - I spent only about 2 (PII-350) CPU hours spread over all 159,975 candidate exponents in the range I selected. But that's enough to eliminate a third of them. In fact, there was a problem with a particular version of George's factorer such that it didn't always find factors _smaller_ than 2^32 or so. And this was discovered only when GIMPS was expanding to the current high exponent of 20.5 million or so. So I ran mersfacgmp to find all the factors less than about 2^33 for all prime exponents up to about 21.5 million. It took a couple of days on my (then) 90MHz Pentium. Maybe we should start checking factors in (the range of prime numbers used in your database) in 2^40 segments, between you and me (and others?). Others, including myself, have already done parts of this. The data that I have collected from them all is significantly larger than the web-accessible disk space I have. But feel free to send me more factoring data and I'll send, in suitably sized chunks, whatever parts of the data that I have that you want. When you send me the source, I'll try to port it GCC (then maybe we can make something useful out of it :-) No need; mersfacgmp already supports GCC, since my primary machine is RedHat Linux v5.2. But see below for reasons that another program, independently developed, can be very useful. What type of sieving did you do (if any) ? Mersfacgmp does essentially the same sieving as George Woltman's code, though in a different order so as to test possible factors in order. Mersfacgmp tests possible factors in order so that the "checkpoint" file is just the largest number attempted so far. George's factoring checkpoint files are in binary and include the sieving pass number (out of 16). George's code is organized that way for the speed advantage. Did you write the modpow routine, or was it included with some math header? George got the basic algorithm from Knuth, I believe, and pointed it out to me not long after GIMPS started. It sped up my pre-GIMPS factoring program by something over a factor of three, as I recall. How much room for improvment in speed would you say is there in the source? Quite a bit, probably, but the primary loop is rather small and most of the math is done by libgmp (or freeLIP in the case of mersfaclip). I have limited computational resources, a PII233, a P100, but because of my work, I have practically unlimited resources in the 486 department they could finish a 2^40 section in about a day or so. As mentioned above, feel free to send me any data that you want; I've been saving this sort of data for a lot longer than GIMPS. E.g., I have some pre-GIMPS data for certain exponents up to about 240 million. The results file, containing all the data that I have, is now over 90 MB; the list of exponents alone is over 14 MB. If nothing else, sending me your data will let me compare it to what I already have for exponents and factoring ranges in commonn and check to see if any of the programs have bugs. This sort of comparison has found bugs in both the mers package programs and in George's programs several times and is a big reason for doing double checking with distinct programs on distinct hardware. Will http://www.garlic.com/~wedgingt/mersenne.html Brief web page w/ links mers.tgzMers package, tar'd, gzip'd Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Re: Mersenne Digest V1 #576
Daren Scot Wilson writes: I've switched from Linux to BeOS - entirely, not even dual-booting both. Same hardware as before - PII 400 MHz. BeOS is POSIX compatible, has TCP/IP, but the file system is offbeat, and from what I hear most linux software needs a little bit of tweaking to compile for Be. Is there any Mersenne testing software that can run on BeOS? Or other interesting math crunching software? See the mers package code that I maintain. It is ANSI C source code plus a shell script. There are also pointers to other programs, notably Ernst Mayer's Fortran90 LL tester, on my mersenne.html page. Will http://www.garlic.com/~wedgingt/mersenne.html mers.tgz mers.tar.gz Two different names for the mers source tar file since some web browsers don't realize that '.tgz' is binary and some operating systems don't like two dots in file names. If you need a zip or shar file or other format, just ask. Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Re: Mersenne Digest V1 #575
Yvan Dutil [EMAIL PROTECTED] wrote: One weird bahavior: 5292757. Which as appear to have been resetted recently by its 'owner' to an duration as long as the initial one! Obviously, someone got dinged with the v17-v18 upgrade. I pity them... Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Get free e-mail and a permanent address at http://www.netaddress.com/?N=1 Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Re: Mersenne Digest V1 #575
On Mon, 14 Jun 1999, Steinar H. Gunderson wrote: I thought that if no check-in was done in 60 days, the number was put back in the pool. No, if no check-in has been done in 60 days _after the exponent was expected to complete_, it is put back into the pool. Of course, once in a while, the software will report `new expected completion dates' to the server, and this date will be moved. George did a post earlier telling about this, the policy you're describing is the one used for version 15. In the later versions it's 60 days after last check-in, unless you've reported a vacation. -- Henrik Olsen, Dawn Solutions I/S URL=http://www.iaeste.dk/~henrik/ Do not meddle in the affairs of dragons, for you are crunchy and taste good with ketchup. Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Ideas : maybe no new ones...
All you GIMPS-ers... Today while I was mindlessly working my body on a treadmill to attempt to maintain some kind of decent physical condition, I had a thought (it wasn't intentional... it just happened, honest!). Has anybody tracked the value of "S" in the LL test? When you square S over and over and over in one LL test, does that same value of S come up in a test for another exponent? Is there a pattern of S values that can help us shorten the LL test? If there is a pattern, and it can be used to calculate a certain S value for some other (untested?) exponent at some given iteration, then it could cut weeks off of testing exponents. Maybe just wishful thinking... But seriously, has somebody tracked this kind of data to look for repeating patterns that can be made into formulas? Gary Diehl --If curiosity killed the cat, then M38 brought him back! Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Final (hopefully) comments on poaching
Hi all, 1) The problem is solved for all v16 and later clients. In other words it is 99% solved. Even if you check out a first time LL test on a P-75 the existing system will not time you out as long as you are running the program regularly. The Primenet status reports will show the world that you are indeed working diligently on your exponent. 2) The only good idea to come out of this thread is to allow users to send an I'm-still-working message in the manual web forms, along with a more generous initial timeout of 120 days. It is up to Scott to decide if he is willing to spend the time to implement this. 3) Other ideas for reassigning exponents that were assigned to v15 clients or by email included setting a firm policy, sending emails, waiting for responses, etc. etc. While a good idea it is more time-consuming for me. 4) M#37 was a recycled assignment. BTW, there are 9000 double-checks required before the ? is removed from M#37. 5) One of GIMPS greatest features is the lack of rigid rules. However, as this thread points out, this feature has a downside too. Since my time is limited and I've been entrusted by default with inventing GIMPS' official poaching rules, here it is: 1) Do not poach exponents. 2) I will apply the existing ad hoc policy for reassigning stale v15 exponents and email-reserved exponents. They will be recycled on an infrequent basis when I find the time. The server will then hand them out to random lucky users. I'm more aggresive with smaller exponents than larger exponents so that we will march steadily onwards. I'm more aggresive with people I don't know. If you reserved a range by email and report results every 5 months then you are safe. Have a nice day, George Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: Ideas : maybe no new ones...
At 08:06 PM 6/14/99 -0500, Gary Diehl wrote: When you square S over and over and over in one LL test, does that same value of S come up in a test for another exponent? It could, but I don't think that helps since for each exponent the calculations are done to a different modulus. +--+ | Jud "program first and think later" McCranie | +--+ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne: Power supplies (was poaching was digest whatever)
*never ever* cheap out on power supplies. This is good advice, but personally I have never seen the point of giving every machine in a rack of computers its own power supply rather than having one big one and just running DC all the way up the rack. The fact that it is not done that way seems to be about politics of having AC wall current rather than engineering efficiency. Why not have a single (redundant) big 24VDC power supply for all the boards instead of supplying them all 120VAC? Well, I would guess that this is a catch-22. This is a good idea, which could actually produce cheaper, high quality power. But this is non-standard. The logistics of creating a new case design, and power supply designs are more than most corporations are willing to stomach on something as cheap as electric power, and equipment that is going to be upgraded in 2 years anyway. -Lucas Wiman Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Re: Mersenne: factoring 10^7 digits
Sounds about right for a SWAG estimate. SWAG? SWAG stands for Scientific Wild-Assed Guess - translation: educated guess -Lucas Wiman Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm