Re: Mersenne: Re: 26 exponents
Subject: Re: Mersenne: Re: 26 exponents On 12 May 2001, at 15:26, Nathan Russell wrote: On Sat, 12 May 2001 14:20:36 -0400, Jud McCranie wrote: snip Thirdly most people will neither know nor care about the detail of how allocations are made. That's an excellent point. When I first started I had no idea what exponent I would get or why. A better fix would be to patch PrimeNet so that it can assign an exponent for two LL test runs simultaneously. (Whichever finishes first becomes the LL test, the other is the double-check). While that's basically a good idea, it's important to be honest with participants. A patch would not be enough. People need to be informed about departures from documented practice. That _might_ be a good idea, except in the eventual situation where both participants return results indicating their number is indeed prime. Whoever had the slightly slower machine will not be very happy! Steve Harris _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Pentium 4 owners - pre-beta prime95 release available
Is ordinary factoring code also present (Factor Only) and is there a corresponding speed-up? What a prospect: doing a 33M exponent L-L in about 60 days! Thanks And Best Wishes, Stefanovic xqrpasuper - Original Message - From: George Woltman [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Saturday, May 12, 2001 6:02 PM Subject: Mersenne: Pentium 4 owners - pre-beta prime95 release available Hi, The P4-optimized version of prime95 is ready for adventurous users to try out. Ordinarily I would not release a trial version at this point, but I'm sure P4 owners are itching for a 3x performance boost. -mers _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Re: 22 below M#38 and counting
On Sun, May 13, 2001 at 09:24:00PM -0400, George Woltman wrote: So. the first 6 folks that email me privately can have one exponent each. Yes please -- if there are any left by now :-) /* Steinar */ -- Homepage: http://members.xoom.com/sneeze/ _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
On Mon, 14 May 2001 00:20:47 +0100, Daran wrote: As someone currently running a legacy machine, (It's taking 4-5 months to run double-checks in the range under consideration,) I have some thoughts on this. First of all, as Jud notes, the 'elitism' is already there, in that different machines get treated differently in the assignments that they are given. It also lies at the heart of the 'Top producer' chart. Every distributed computing project larger than a few dozen members has such a chart. It would be difficult to keep people interested without one. I know that I don't feel less a part of the project because my placing is above 4,000. Even readers of this list get opportunities to acquire exponents or prebeta-test software, etc., that are not available to the unwashed masses. Very true. Of course, George can hardly contact all several thousand participants when such an opportunity happens. Additionally, GIMPS, unlike most other projects, has exponents taht are 'better' than others. If I run a range of distributed.net keys, it has an equal chance of containing the correct key compared to any other range of equal size. If I run a work unit for seti@home, some work units may be slightly more likely to contain a message, but since a fast computer does multiple work units in one day, it all averages out fairly quickly. Secondly, if - when I ask the server to give me whatever kind of work makes most sense - it gives me something else, whether out of spurious concern for my feelings or for any other reason, then not only are the programmers betraying my trust in them, they are also indicating that they don't trust me to ask for what I want. The type of work that makes the most sense is chosen based on CPU speed, and is based on George's desire to avoid giving machines work that will keep them busy for more than a certain length of time, increasing the risk of an incorrect result. You can override that setting if you have a strong preference. If I choose to specify what kind of work I want, I still expect the be given the work that makes most sense within that category. I certainly would not expect to be given work that would delay a milestone, given the limitations of my machine. This might be a reasonable change in PrimeNet. Personally, I don't think milestones should be a focus of the project, but it is nice when a new one appears on the page. It makes sense to offer factorisations to newcomers to the project. It probably would make sense to offer at least one double-check before moving on to first time checks. And it makes sense to offer milestone-blocking work to fast machines with a proven track record of reliability. People will understand this. People do not expect to be given jobs that they are not able to do, or positions of trust within days of joining a new club. I still think that this is very debatable. There should not be a certain /assignment type/ reserved for 'veterans', but it may be reasonable to, e.g., only give triple-checks to accounts that request double-checks, and have returned more than a certain number of results. Note that an exponent given out for triple-checking has a microscopic chance of being prime (something like two in one billion), since it must 1. Be prime (once chance in 60,000-70,000) and 2. Have been missed by both previous tests (1 in 100 for each). People need to be informed about departures from documented practice. Are you suggesting that, every time George offers exponents to the members of this mailing list, he should send out a newsletter to every participant - guaranteeing hundreds or thousands of replies for him to deal with? I think there may be no good solution to this. Nathan _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Pentium 4 owners - pre-beta prime95 release available
At 01:02 AM 5/14/2001 -0700, xqrpa wrote: Is ordinary factoring code also present (Factor Only) and is there a corresponding speed-up? The factoring code is present but unchanged. Thus, no speedup. Regards, George _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: missing exponents?
Chris, I don't think he was bashing Outlook PER SE -- just the version number in question. There have been MANY MANY security fixes to OE since that release, which came with Internet Explorer FOUR a few years back. Even the granola OE that comes with IE 5 (v5.00.2314.1300) has been radically security-patched since release At 12:50 PM 5/12/01 +0100, you wrote: You should use another MailClient as Microsoft Outlook Express 4.72.3110.1 8) Didn't you read all the bad things about Mailwurms etc? Sorry, I just felt the urge to say this. There is nothing particualarily bad about Outlook Express, these various viruses require you to excute attachments, and you can do that in any mail client.. If any other mail client gets as popular as OE, then it will start to have viruses aimed at it's address book too! Chris regards, Mohk _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: miscellaneous
Hi all, The 6 available small exponents are long since gone. One had a P-1 factor found this morning. I've had one other LL result reported, so we're at 20 below M#38 and counting Embarrassingly, the P4 version I announced 2 days ago fails the self-test. You would have thought I'd have run that(!), but I ran the QA suite instead. The FFT code is fine, but a change in the way P4 handles carry propagations means you run into trouble if there are too *few* bits per FFT word. Thus, I tweaked the self-test code to not run any small exponents in big FFT lengths. Any LL tests started with the 2 day old version are just fine, the bug was only in the self-test code. Regards, George _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: Primenet exponents (was: missing exponents?)
On Sun, 13 May 2001 20:39:34 -0400, George Woltman wrote: At 07:46 PM 5/13/2001 -0400, Nathan Russell wrote: As it stands, I notice that PrimeNet is given assignments only a few tens of thousand of exponents in advance. Is this done so that you have more flexibility, or is it a technical issue with the number of exponents the server can handle? There is no server limit, nor is there any good reason for my only giving the server a few thousand exponents in advance. some months ago I wanted to test 'my lucky number' :-) with Prime95 so I added it manualy to the worktodo.ini file. Everything was OK - until Prime95 contacted the server and got the 'exponent not assigned to you' error message. After that the newly added line disappeared from the worktodo.ini file. Then I tried it with the Advanced-Test option and with that the line remains in the worktodo.ini but now I'm the only one who knows that I'm testing that exponent. OK, I saw on the status page that only 2 exponents were tested in that range (2040-2533), but what if one of these is exactly the one? Wouldn't it be usefull to allow the server to accept new exponents if they are 'reasonable'? BTW, what happens when 'AdvancedTests' are completed? Can the server handle the results? greetings Siegmar _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Emails and virus (slightly OT) :)
Sorry, I just felt the urge to say this. There is nothing particualarily bad about Outlook Express, these various viruses require you to excute attachments, and you can do that in any mail client.. If any other mail client gets as popular as OE, then it will start to have viruses aimed at it's address book too! Chris, I don't think he was bashing Outlook PER SE -- just the version number in question. There have been MANY MANY security fixes to OE since that release, which came with Internet Explorer FOUR a few years back. Even the granola OE that comes with IE 5 (v5.00.2314.1300) has been radically security-patched since release FWIW, one of the best ways to protect your email against virii is at the server level. My current job involves, among other things, administering a couple hundred email domains, and I'm just appalled to see all the viruses that pass through our system each and every day. There are a LOT of viruses out there, and a LOT of people who have no scanning. Of course, we've got our system set to block all VBS files, and we only grudginly allow EXE's (we do scan them though). And of course, in our own company, it was easy enough to just set VBS files to auto-open with notepad rather than running via wscript, merely by changing the default file association. I'll try to keep this a little bit on-topic... Does this mailing list do any sort of virus scanning? I know you can't post to the list unless you're actually on the list, but any scanning going on? Hopefully the folks on this list are bright enough that they know better than to run any executables they receive via email... Aaron _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Emails and virus (slightly OT) :)
I'll try to keep this a little bit on-topic... Does this mailing list do any sort of virus scanning? I know you can't post to the list unless you're actually on the list, but any scanning going on? No there isn't any virus scanning going on. There is however a message limit that tends to block most attachments. FWIW, I work at Postini, http://www.postini.com , a company that provides real time email spam and virus blocking services. Customers simply point their MX records at us, and then we take care of the rest, filtering stuff out, and passing on just the good mail. I don't however run base.com mail through Postini. base.com provides me a portal on the net without the Postini filters in place. This is useful for collecting spam which can then be used to tune Postini's spam filtering engine. gordon _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: games one can play with genuine composites
On 13 May 2001, at 20:06, [EMAIL PROTECTED] wrote: True, for composite c, not all factors of 2^c - 1 need have form 2.K.c + 1; for example, if c = 15, 2^c - 1 = 7.31.151, and the smallest factor, 7, does not have the form 2.K.15 + 1. However, if we do not know any of the factors of c, the best we can reasonably do is to search for factors of the form 2.K.c + 1, which we can do without knowing any of the factors of c. Yes. For c sufficiently large, even sieving for factors of this form becomes prohibitively expensive - take your own favorite composite lacking known factors, c = 2^33219281-1. Checking if the smallest eligible candidate factor 2.K.c + 1 divides C = 2^c - 1 would need roughly 33219281 squarings modulo the candidate factor, and since that is not of special form (e.g. Mersenne or generalized Fermat) each squaring modulo the trial factor will be more than twice as expensive as a squaring modulo c. Yes, I did point out in my original message on this subject that testing a single factor of M(M(p)) involves about as much effort as running a LL test on M(p). Actually the squaring need not be much more expensive as it is possible to combine squaring and modulo reduction for Proth numbers using DWT in much the same way as is done for Mersenne numbers, and 2kM(p)+1 is going to be a Proth number for all reasonable values of k once p is any size at all. On the other hand, if c itself has a smallest factor so large as to make finding it practically impossible, then indeed it may prove easier to find a small (in the sense that K is small) factor of 2^c - 1 than of c. By its very nature such a factor would not help one in factoring c itself. Do you mean you can prove that last sentence, or do you just mean we're all so stupid that we haven't found out how to, yet? I'm not sure what point you're trying to make with the K k^x condition - do you mean to ensure that if k is large (which it must be if c has undergone a sufficient amount of trial factoring), K also will be? Yes. If we know that e.g. k10^40 and we can prove that e.g. Kk^0.1 then we know K1. Maybe _much_ bigger. Which implies (even after sieveing out those values of K for which 2KM(p)+1 itself has a small factor, or is congruent to 3 or 5 modulo 8) the work involved in finding that factor is at least thousands of times greater than the work involved in LL testing M(p). In other words, given current hardware, it really is a waste of time, for p ~ 10^6. Conversely, if there is no such limit on x, then we _could_ be lucky and find a factor with a very small K. In fact, given the distribution of factors of Mersenne numbers (even ignoring the special case where p is a 3 mod 4 Sophie Germain prime), it is rather more probable that we _will_ find a factor in a fixed size interval when K is small than it is when K is large. My gut feeling is that there is some limit, i.e. if k is large then K will be large too. However, formal proofs and heuristic arguments both seem to be vanishingly thin on the ground. Perhaps one should simply exclude such constructions by defining a genuine composite as a number which has been shown to be composite by direct (nonfactorial) means, e.g. Lucas-Lehmer, Pe'pin, Proth or any other rigorous compositeness test, and which has no known factors. Yes. Otherwise the very question of what is the largest known composite with no known factors has no meaning. Regards Brian Beesley _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: missing exponents?
On 13 May 2001, at 19:41, Nathan Russell wrote: Another nice thing would be if people who have submitted several results that don't match other people's results they could be notified that they may have a hardware problem. I can't help wondering whether some users would find that to be overly invasive or startling; however, if I had such a problem - especially one that could cause slow thermal damage - I would want to know about it. There is already a mechanism where people can opt in or out of being notified if an assignment is due to expire. Is it reasonable to ask that this may also be used to notify the user if/when it is discovered that they have submitted a result which has turned out to be incorrect? There are two practical difficulties with this approach: (1) the server doesn't know: George discovers discrepancies when he processes PrimeNet transactions into his master database; (2) by the time the discrepancy is found, some considerable time may have elapsed (when are exponents in the 12M range going to get routine double-checks?) and so the reliability or otherwise of the hardware involved in the original test may well be a moot point. I deal with altogether too many people on other mailing lists and newsgroups who believe that it is normal for their machine to take two attempts to make it through a 25-minute kernel upgrade. Um, upgrading a kernel is essentially nothing worse than a reboot! But yes, there are some people out there with _very_ dodgy hardware. Granted, a 5-week GIMPS run is a more stringent criterion, but AFAIK most machines /should/ be able to survive without more than 1-2 errors a year (I've had two since I started last January). I agree entirely, though there are causes of errors which are not related to hardware problems per se. Seriously noisy utility power supply and memory corruption due to badly behaved applications running on insecure operating systems probably claim a fair proportion of the victims. Regards Brian Beesley _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Emails and virus (slightly OT) :)
Sorry, I just felt the urge to say this. There is nothing particualarily bad about Outlook Express, these various viruses require you to excute attachments, and you can do that in any mail client.. Not true! With Outlook Express all you have to do is read the message. The virus can be embedded in the message with Outlook Express. The Wscript.KakWorm virus is a particularly bad example of this. For details see: http://www.sarc.com/avcenter/venc/data/wscript.kakworm.html Regards, Herb Savage _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
Nathan Russel wrote: Are you suggesting that, every time George offers exponents to the members of this mailing list, he should send out a newsletter to every participant - guaranteeing hundreds or thousands of replies for him to deal with? I think there may be no good solution to this. Have the e-mails go to an account linked to the server. It would take a little work, but the whole process could be automated. Regards, Joshua Zelinsky [EMAIL PROTECTED] _ Get your FREE download of MSN Explorer at http://explorer.msn.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Pentium 4 owners - pre-beta prime95 release available
On 14 May 2001, at 9:08, George Woltman wrote: At 01:02 AM 5/14/2001 -0700, xqrpa wrote: Is ordinary factoring code also present (Factor Only) and is there a corresponding speed-up? The factoring code is present but unchanged. Thus, no speedup. ... except for the fact that the slowest P4 processor runs about as fast as the fastest Athlon, and is also supposed to be more efficient at executing integer instructions. Trial factoring seems to be well in advance of LL testing, so it probably doesn't matter too much. But is there any scope for using SSE2, or even MMX / 3DNow, extensions in trial factoring? The point being that trial factoring can be implemented reasonably efficiently as almost pure integer code. If we could test two or four factors in parallel, we should get a considerable speedup, even though some of the threads would spend some time stalled whilst exceptions were dealt with in another. Regards Brian Beesley _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Emails and virus (slightly OT) :)
I'll try to keep this a little bit on-topic... Does this mailing list do any sort of virus scanning? I know you can't post to the list unless you're actually on the list, but any scanning going on? No there isn't any virus scanning going on. There is however a message limit that tends to block most attachments. FWIW, I work at Postini, http://www.postini.com , a company that provides real time email spam and virus blocking services. Customers simply point their MX records at us, and then we take care of the rest, filtering stuff out, and passing on just the good mail. I don't however run base.com mail through Postini. base.com provides me a portal on the net without the Postini filters in place. This is useful for collecting spam which can then be used to tune Postini's spam filtering engine. I was just wondering... that Homepage virus that was going around last week was actually relatively small. I don't know what size limits you have on the list, but it did cross my mind: what would happen if someone got a virus and had [EMAIL PROTECTED] in their address book (as I do). As for blocking spam, I created a spam user and try to advertise that as much as possible, hoping it'll get on as many lists as there are. The email software we use, Communigate, will reject any email that has that spamtrap user as one of the recipients. Works fine for those spammers that use to: or cc: fields, but not really much help for the bcc. Oh well. I wonder... could the base.com be setup to block all attachments, or, for that matter, block HTML posts, allowing text only? Aaron (who is currently fighting off some spam of his own... grid.net user bouncing off mindspring.net and then into our machines... argh!) _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Emails and virus (slightly OT) :)
Sorry, I just felt the urge to say this. There is nothing particualarily bad about Outlook Express, these various viruses require you to excute attachments, and you can do that in any mail client.. Not true! With Outlook Express all you have to do is read the message. The virus can be embedded in the message with Outlook Express. The Wscript.KakWorm virus is a particularly bad example of this. For details see: http://www.sarc.com/avcenter/venc/data/wscript.kakworm.html Well, that's true if you haven't upgraded your IE with the latest security patches, and especially true if you're security zones are left at the default which allow all sorts of nonsense to go on. Call me paranoid (Hey, paranoid!), but setting your security zones in IE to be more restrictive is absolutely vital. I'm puzzled as to why older versions of IE were so lax. I haven't tried it, but my co-workers who are more daring have said that the IE6 beta actually does a better job of restricting your sites more, so the Internet site allows less monkey-business. The Kakworm, fortunately, never made it big time, but you're right, that was a fine example of an HTML email actually being able to cause some damage. One can only ponder what would have happened had it spread better. Aaron _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Overriding assigned exponent type (was Re: Mersenne: Re: 26 exponents)
Secondly, if - when I ask the server to give me whatever kind of work makes most sense - it gives me something else, whether out of spurious concern for my feelings or for any other reason, then not only are the programmers betraying my trust in them, they are also indicating that they don't trust me to ask for what I want. I agree. Either you allow people to choose the type of work they want, or you tell people plainly that you will select for them the type of work you will ask them to do. Either works, but a mixture is inconsistent. I'm confused... I know there's the option to have Primenet request whatever type of work makes the most sense, or you can uncheck that and select a particular type. That works fine. However, if you do select multiple options, the server will still select which, from among those options, is most suitable for your machine. Personally, I'd always wondered why it let you select multiple types when the point was to override any server default and get the type you want. It should be a radio box instead of a combo box on that one, near as I can figure. ... Clearly this is ridiculous. I don't have a problem with George offerring a few exponents selectively through this list, because the list does not have a closed membership. The only sane alternative is to wait for deadline critical assignments to complete in the normal way - something which some people have vociferously objected to. Agreed with that, Brian. Besides, people on this list are probably more interested in what's going on than the fellow who just likes knowing that his computer is doing something with all those spare cycles. So offering limited cool things to list subscribers is fine by me. Perhaps there could be an option somewhere in the user information (same place you set your email address) that would allow you to opt in to the mailing list? I fear that many folks may not be aware of the list, or find that subscribing seems too hard (odd as that may sound to us experts :) Next time they're online, it'll update their computer info, generate a subscribe letter for that email address, and they'd still have to respond to the subscribe message, but it's a heckuva lot easier for some people that way. Of course, it would require Scott to do something on Primenet to allow such an option to be trackable, and to generate the subscribe messages to the list. But I reckon you'd see the list membership jump up quite a bit if only it was easier for folks to subscribe. Aaron _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Purpose of the self-test; also, aren't P4s fast!
I've just collected the beta of version 21, to run on my P4 at home. When I run it on my K62/333 laptop (which I was using as a portable disc having downloaded the beta at college), the self-test tries to test M5242881 with a 256k FFT, gives an excess-roundoff-error message and dies; this is presumably a bug. I did set the processor type to K6. When I run it on the P4, the self-test checks a set of thirty or so exponents with a 320k FFT length, and then does it again, and again, until it's run a total of 306 tests. This surprised me slightly; I was expecting it to exercise all the FFT lengths with random exponents to check for edge-conditions in the new P4 code. Is the self-test in fact just to check that there's not something in the CPU which goes glitchy when running flat-out SSE2 code for hours on end? The P4's PSU fan seems to step up a gear when I've been running Prime95 for a few hours, though the CPU temperature reported by Intel Active Monitor on my D850GB board doesn't go about 46C. I suspect the P4 probably runs too hot for it to be possible to build a really quiet solution, but does anyone have suggestions? In its current state I'd get no sleep if I left the machine running overnight, so I'll probably be running Prime95 only if I remember to set it going when I wake up. 33M ticks per iteration for M5171311, which is 25.8 milliseconds on this 1300MHz machine; so one double-check in that range every 36 hours. That's about a factor six faster than the P2/350 I just sold, and I recall that machine as having been at least four times faster than the P90 I had when I started running Prime95 ... I'll have done noticeably more calculation by the end of next month than that P90 did during its lifetime. Tom _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Purpose of the self-test; also, aren't P4s fast!
The P4's PSU fan seems to step up a gear when I've been running Prime95 for a few hours, though the CPU temperature reported by Intel Active Monitor on my D850GB board doesn't go about 46C. I suspect the P4 probably runs too hot for it to be possible to build a really quiet solution, but does anyone have suggestions? In its current state I'd get no sleep if I left the machine running overnight, so I'll probably be running Prime95 only if I remember to set it going when I wake up. Any machine that has a variable speed fan will probably have this same thing happen. On a couple newer Compaq Proliant servers (which have temperature sensitive fans), it's very obvious to tell when I've started the Prime service, because the fans nearly instantly speed up and are noisier. This is, of course, because normally the CPU is pretty idle, not generating much heat at all. But when the service starts, the FPU is suddenly in full time use, and the core generates more heat, which causes the temp switch to notice this, and causes the fans to speed up to dissipate the extra heat. Hearing your fans kick into high gear is a good sign... it means your system is working as intended. :) I'm thankful that most of my servers are off in an air-conditioned room down the hall a bit. It was a real pain when I had a Proliant cluster in my office undergoing some testing, and having those fans spin at full tilt. And if you've ever been around a Proliant server, you'll know what I mean because those things have about a million different fans in them. 33M ticks per iteration for M5171311, which is 25.8 milliseconds on this 1300MHz machine; so one double-check in that range every 36 hours. That's about a factor six faster than the P2/350 I just sold, and I recall that machine as having been at least four times faster than the P90 I had when I started running Prime95 ... I'll have done noticeably more calculation by the end of next month than that P90 did during its lifetime. Hmm... your P4 at 1.3GHz should, on face value, be 3.8 times faster, and according to George, you should be seeing an extra 3 times improvement in the P4 execution for an overall boost (over a P2/350) of over 11 times. Of course, that's assuming similar cache architecture, bus speeds, etc. which we know isn't the case. In fact, without the new code, I'd expect a P4 1300 to be more than 3.8 times faster than P2/350, just on the basis of the memory architecture, FSB (and thus memory) speeds, etc. I assume your beta has been configured with the proper CPU type? Aaron _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: missing exponents?
On 14 May 2001, at 19:04, Brian J. Beesley wrote: There is already a mechanism where people can opt in or out of being notified if an assignment is due to expire. There is? At the risk of looking dim, what is it? In the user information config window, you enter your email address and there's a checkbox to receive email from Primenet server if exponents are about to expire. _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Purpose of the self-test; also, aren't P4s fast!
Hi, At 11:38 PM 5/14/2001 +0100, Thomas Womack wrote: When I run it on my K62/333 laptop, the self-test tries to test M5242881 with a 256k FFT, gives an excess-roundoff-error message and dies; this is presumably a bug. Yes, apparently I broke the old FFT code Another good reason for only P4 users to download this version! When I run it on the P4, the self-test checks a set of thirty or so exponents with a 320k FFT length, and then does it again, and again, until it's run a total of 306 tests. This surprised me slightly; I was expecting it to exercise all the FFT lengths with random exponents to check for edge-conditions in the new P4 code. The SSE2 instructions only support 53 bits of precision. The old Intel FPU supports 64 bits of precision. Consequently, the old FFT code can handle slightly higher exponents. For a 256K FFT, the old code could handle up to 5,250,000, the new code can only handle up to 5,140,000. Is the self-test in fact just to check that there's not something in the CPU which goes glitchy when running flat-out SSE2 code for hours on end? Yes. The QA suite that Ken Kriesel and Brian Beesley worked on does a better job at testing edge conditions. Of course, they'll need to update that suite using the new limits. Regards, George _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Purpose of the self-test; also, aren't P4s fast!
Hi, At 04:41 PM 5/14/2001 -0700, Aaron Blosser wrote: That's about a factor six faster than the P2/350 I just sold, Hmm... your P4 at 1.3GHz should, on face value, be 3.8 times faster, and according to George, you should be seeing an extra 3 times improvement in the P4 execution for an overall boost (over a P2/350) of over 11 times. Not quite, the old FFT code did not run well on the P4. Thus, his 1.3GHz P4 was not 3.8 times faster than his P2/350. -- George _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: missing exponents?
On Mon, 14 May 2001 17:27:19 -0700, Aaron Blosser wrote: On 14 May 2001, at 19:04, Brian J. Beesley wrote: There is already a mechanism where people can opt in or out of being notified if an assignment is due to expire. There is? At the risk of looking dim, what is it? In the user information config window, you enter your email address and there's a checkbox to receive email from Primenet server if exponents are about to expire. I can't find this in my copy (Prime95 20.6.1). Nathan _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: missing exponents?
Hmm... well, then again, I'm looking at the NTPrime. I've only got one machine running Prime95, and it's been so long... I thought it had all the same options though, but I could just be terribly mistaken. Running NTSetup (part of the NT service package), I show version 20.6.5... Aaron - Original Message - On Mon, 14 May 2001 17:27:19 -0700, Aaron Blosser wrote: On 14 May 2001, at 19:04, Brian J. Beesley wrote: There is already a mechanism where people can opt in or out of being notified if an assignment is due to expire. There is? At the risk of looking dim, what is it? In the user information config window, you enter your email address and there's a checkbox to receive email from Primenet server if exponents are about to expire. I can't find this in my copy (Prime95 20.6.1). Nathan _ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
On Sat, 12 May 2001 16:04:17 -0400, Jud McCranie wrote: At 03:26 PM 5/12/2001 -0400, Nathan Russell wrote: I think that's more of a 'quick fix', and might make new participants feel that GIMPS doesn't trust them. Yes, but a new user need not know that they don't get an exponent that has expired until they have finished an assignment. My point is that if an exponent is dropped, it could be reassigned to someone that has shown a willingness to finish it. However, that is still drawing a distinction between new and experienced users. For that matter, if a milestone is delayed by a month or two, it doesn't significantly hurt everyone's overall odds of finding a prime. I've been steadily working on GIMPS for nearly 5 years, always 1 fulltime machine, occasionally 2. I've been doing doublechecks in the 6,000,000 range for a few months because I use a 300 MHz machine. I know doublechecking is important. But then I see these few gaps under M38? and I think I could have done several of those. I can empathize with you here. However, I was a new user only a little over a year ago, and if someone had said on the mailing list at that time that new users should be given assignments chosen so that they couldn't harm milestones, I would have been upset. Yes, it would be nice to say that we know for sure which Mersenne prime is the thirty-eighth, but doing so does not speed us towards discovering the thirty-ninth. For that matter, I am sure that there are users who have run a single exponent and then left, though they may not be as many as those who left without ever finishing any exponents. Nathan _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: missing exponents?
On Mon, 14 May 2001 18:17:10 -0700, Aaron Blosser wrote: Hmm... well, then again, I'm looking at the NTPrime. I've only got one machine running Prime95, and it's been so long... I thought it had all the same options though, but I could just be terribly mistaken. Running NTSetup (part of the NT service package), I show version 20.6.5... Aaron I have version 20.6.1 - and the web page reads that all versions were last updated June 15 2000. Something odd is going on... Nathan _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
On Mon, 14 May 2001 20:23:45 -, Brian J. Beesley wrote: On 14 May 2001, at 8:45, Nathan Russell wrote: First of all, as Jud notes, the 'elitism' is already there, in that different machines get treated differently in the assignments that they are given. (To clarify, I did not write the above - Daran did) Sorry, I don't buy that. Every system has exactly the same chance of picking up any given assignment; it's a matter of the time at which you make the request. And you _can_ override the assignment type which would be the default for your system, if you wish to do so. That is indeed true. I would have to say that having defaults different for different systems is a Good Thing; otherwise, you might have someone using a 486 suddenly realize that their computer was doing a first-time check that would take over a year, get frustrated, and give up. Even readers of this list get opportunities to acquire exponents or prebeta-test software, etc., that are not available to the unwashed masses. AFAIK everyone is entitled to subscribe to this list, whether they participate by running assignments or not. Agreed. Membership to the list indicates a slightly-more-than-casual interest in the project, specifically a willingness to sift through a few dozen messages per month in order to learn more about the project. That interest might well also be a sign of someone who is more likely to faithfully complete 'special' assignments in a relatively timely fashion. Additionally, GIMPS, unlike most other projects, has exponents taht are 'better' than others. In the absence of completed tests, small exponents are more likely to be prime than larger ones, as well as taking less effort to test. However, note that a considerable number of users have voluntarily chosen to run 10 million digit range exponents, thus reducing the probability that they will discover a prime. The increased reward for being successful counterbalances the reduced chance of success. True - and probably by a greater margin now that ordinary first-time testing is getting higher (12,200,000 now as opposed to 9,700,000 when I joined late in January 2000). There is also a theoretical difference between those exponents congruent to 1 modulo 4 and those congruent to 3 modulo 4. However I believe that this is due to the fact that one of these groups has a larger probability of having a small factor; thus this irregularity is removed by the time that LL testing begins. I think I read something similiar. Might it relate to whether the first potential factor itself is prime, specifically whether it is divisible by 3? I can't do the arithmetic in my head, but I have a hunch... Secondly, if - when I ask the server to give me whatever kind of work makes most sense - it gives me something else, whether out of spurious concern for my feelings or for any other reason, then not only are the programmers betraying my trust in them, they are also indicating that they don't trust me to ask for what I want. I agree. Either you allow people to choose the type of work they want, or you tell people plainly that you will select for them the type of work you will ask them to do. Either works, but a mixture is inconsistent. Perhaps clicking the 'give me the work that makes the most sense' box should immediately set the appearance of the others to the work that will be chosen, rather than simply graying them out. Note that an exponent given out for triple-checking has a microscopic chance of being prime (something like two in one billion), since it must 1. Be prime (once chance in 60,000-70,000) and 2. Have been missed by both previous tests (1 in 100 for each). NO! Conditional probability: if we need a third LL test run, it is because at least one of the other two _must_ be in error. So the probability of finding a prime on the third LL test run is (about) one half the probability of finding a prime on the second LL test run - irrespective of the error rate, provided it is small. I stand corrected here. Nathan Russell _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
At 08:23 PM 5/14/2001 +, Brian J. Beesley wrote: On 14 May 2001, at 8:45, Nathan Russell wrote: First of all, as Jud notes, the 'elitism' is already there, in that different machines get treated differently in the assignments that they are given. Sorry, I don't buy that. Every system has exactly the same chance of picking up any given assignment; it's a matter of the time at which you make the request. Under your proposal, exponents would be double assigned. A 200 MHz system and a 1.2 GHz system calling in at about the same time could get the same exponent. But the one with the elite fast system would in effect get the first LL test and the non-elite system would get the DC. Also, if you have a 200 MHz and a 1200 MHz working on the same exponent, if it turns out to be a new prime the 1200 will show that first. Then someone with a fast machine will run a DC before the 200 machine can finish it. I know you said that the 200 still gets credit, but if a prime is reported, we don't want to wait several (possibly many) more months for a DC. Under my proposal, newcomers would get a first time LL test - they just wouldn't get an exponent that someone else had abandoned. That is not to keep newcomers from contributing - they even get to do first time LL tests (and be assured of that) - it is try to keep an exponent from being abandoned more than once. ++ | Jud McCranie | || | former temporary part-time adjunct | | instructor of a minor university | ++ _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
At 09:44 PM 5/14/2001 -0400, Nathan Russell wrote: However, that is still drawing a distinction between new and experienced users. Well, so what if it does? It is an extremely minor difference. New users would still get first time LL tests - it just means that they wouldn't get exponents that someone else abandoned. It is an exponent thing, not a user thing. Many organizations have distinctions for new members. Non-voting members, provisional members, Fraternity pledges, etc. To a new user, what does it matter if he gets a brand new exponent instead of one that someone else abandoned in the same range? Why would a new user prefer an exponent someone else abandoned over one that no one else has been assigned? To me, it seems that (if anything) he would prefer a new one because there is a possibility of the person who abandoned it getting back to work on it and finishing it in the mean time. For that matter, if a milestone is delayed by a month or two, it doesn't significantly hurt everyone's overall odds of finding a prime. I know, but it would be nice to know whether what seems to be M38 is or isn't M38. I can empathize with you here. However, I was a new user only a little over a year ago, and if someone had said on the mailing list at that time that new users should be given assignments chosen so that they couldn't harm milestones, I would have been upset. They would still be contributing towards milestones. If there are exponents below a milestone that never have been assigned, they would get them. Yes, it would be nice to say that we know for sure which Mersenne prime is the thirty-eighth, but doing so does not speed us towards discovering the thirty-ninth. But there could be one smaller than what now seems to be #38, because there are exponents in that range that haven't had even 1 LL. For that matter, I am sure that there are users who have run a single exponent and then left, Well, that's OK. Their work helps. Does anyone have an idea of the % of people who start and then quit w/o finishing an assignment? ++ | Jud McCranie | || | former temporary part-time adjunct | | instructor of a minor university | ++ _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Re: 26 exponents
At 09:44 PM 5/14/2001 -0400, Nathan Russell wrote: However, that is still drawing a distinction between new and experienced users. Well, how about this - new users can get an exponent that has been abandoned several times, but they must check in at least once a month to report the percentage done and expected completion date to show that they are making reasonable progress. It could even be automatic. Or maybe check in at 1 month, 2 months after that, and then every three months? Or is that too elitist too? ++ | Jud McCranie | || | former temporary part-time adjunct | | instructor of a minor university | ++ _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne Digest V1 #850
Mersenne Digest Monday, May 14 2001 Volume 01 : Number 850 -- Date: Sun, 13 May 2001 11:03:50 +0200 From: mohk [EMAIL PROTECTED] Subject: Re: Mersenne: Pentium 4 owners - pre-beta prime95 release available At 03:02 13.05.2001, you wrote: Hi, The P4-optimized version of prime95 is ready for adventurous users to try out. Ordinarily I would not release a trial version at this point, but I'm sure P4 owners are itching for a 3x performance boost. If you do not own a P4, then DO NOT DOWNLOAD THIS VERSION. It contains no new features and your results will be marked invalid if we later discover a bug in this code. Don't you say, this code enhances the Athlon as well? regards, Mohk _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers -- Date: Sun, 13 May 2001 12:13:28 +0200 From: Steinar H. Gunderson [EMAIL PROTECTED] Subject: Mersenne: Re: Pentium 4 owners - pre-beta prime95 release available On Sun, May 13, 2001 at 11:03:50AM +0200, mohk wrote: Don't you say, this code enhances the Athlon as well? No, it doesn't. It's that much faster mainly because it utilizes the SSE2 instructions that are new on the P4, which the Athlon doesn't have (yet). /* Steinar */ - -- Homepage: http://members.xoom.com/sneeze/ _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers -- Date: Sun, 13 May 2001 11:48:48 - From: Brian J. Beesley [EMAIL PROTECTED] Subject: Re: Mersenne: Pentium 4 owners - pre-beta prime95 release available On 12 May 2001, at 21:02, George Woltman wrote: If you try out this version, please let me know of any problems. It would be helpful if you ran double-check assignments for a few weeks. If anyone wants to be a bit more adventurous than taking ordinary PrimeNet DC assignments, I have a limited number of exponents which will require double-checking at some stage for which interim residues at 1 million iteration checkpoints are available. The point here is that if you also set the program for million iteration checkpoints (InterimFiles=100 in prime.ini) you will be able to catch any discrepancies early without having to continue the run to the end. (Of course there is a chance that my checkpoint residues may not be correct - but the systems they are running on seem to be pretty reliable!) The exponents for which I have interim residues are as follows: 7056503, 7092307, 7094839, 7477177, 7731131, 7731259, 7871953, 7927697, 7979819, 8164349, 8167919, 8242327, 8255881, 8369941, 8380621, 8434523, 8549371, 8555881, 8563109, 8573107, 8615273, 8650927, 8655991, 8715337, 8719367, 8767471, 8900161, 8907359, 9057403, 9117209, 9189533, 9443311, 9475097, 9607691, 9617929, 9743639, 9743803, 10068077, 10126621, 10319891, 10386067, 10809893, 11173153, 11561909, 11570227, 11636479, 11772809, 12141223 12450397. If anyone is interested in testing any of these with the P4 code, let me know around what size exponent you're willing to take on; I'll reply by email allocating a unique exponent with a list of the appropriate million iteration interim residues. Regards Brian Beesley _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers -- Date: Sun, 13 May 2001 11:57:41 -0500 (CDT) From: [EMAIL PROTECTED] Subject: Mersenne: Re: story of scientists travelling by train Brian J. Beesley wrote: Mathematician (after long pause): No, in Scotland there exists at least one sheep, at least one side of which appears to be black. What do you mean, long pause ??? :-) Richard Woods _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers -- Date: Sun, 13 May 2001 12:10:38 -0500 (CDT) From: [EMAIL PROTECTED] Subject: Mersenne: Re: story of scientists travelling by train Brian J. Beesley wrote: As the train crosses the border into Scotland, they observe a black sheep in an otherwise empty field. (* snip *) Mathematician #2 (after reading story): Does that mean that the black sheep is both the additive and the multiplicative identity? :-) Richard Woods _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers