Re: Mersenne: prime95 - v21 progress
On Sun, 11 Mar 2001 07:55:04 -0600, Steve wrote: [...] As others have mentioned, the problem is NOT slow machines but rather abandoned exponents, which has nothing to do with machine speeds. exactly... and that is the point... maybe someone with access to the server logs can give us a rough estimation of the expired/checked in exponents ratio, so we better know of what we are talking about... If it is too high, then I think it would be definitevly worth to find a way to reduce it, because indirectly it affects the average test time and recently there have been some 'complaints' about the time to complete a exponent and this beeing a reason not to join/continue the project... recently there were also several remarks regarding the fact, that one has to wait relatively long compared to other distibuted projects just to get mentioned for the first time in the top producers list. maybe we could do something for those who are in for 'competition' on a more individual/machine basis. It takes a long time to climb the 'highscore table' with just one computer. What I was thinking of was a sort of 'fastest machine of the day/week/month highscore table' which mainly shows a ranking of the best machines which checked in/updated the last day/week/month, what I think is easy to compute and can be done after every update/check in. These informations and maybe also some other statistical data can be used in combination with the screen saver idea and displayed on the user screen saying something like: "This computer is working on the GIMPS... right now it's the xxx fastest machine in the project running at , overall ranking is ; peak performance was achived on 00-00-. Estimated time to complete active work: " and so on... ...just my 0.02 euro ;-) greetings, Siegmar PS: talking about mailing list server improvements in another thread... I just noticed, that there is no "Reply-To: [EMAIL PROTECTED]" line in the header. I would find it usefull if it would be like in other mailing lists where a reply goes by default to the list and not to the sender. _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
On 12 Mar 2001, at 0:20, Alexander Kruppa wrote: (1) Removing these assignments from PrimeNet and managing them seperately. Anyone who is prepared to make special arrangements to acquire these assignments is unlikely to default by reason of lack of commitment. Someone is (or was?) doing something similar - David Campeau (sp?), aka diamonddave, who set up scripts to grab small exponents right after the Primenet servers recycle run and completed them in ascending order of size. That was unofficial. I meant official, and I meant removing the lagging exponents from PrimeNet so that they simply can't be grabbed using this sort of technique. In any case, the "lagging" LL assignments are now in the DC range. The "diamonddave" technique simply wouldn't be effective. This doesnt help with the exponents that are currently reserved with a expected run time of several years - but if the user abandons them, the server will recycle them 60 days after the last update from the user which doesnt seem like an overly long delay for the project. If the users chooses to complete the exponent on a slow machine or on a machine that is not in use very often, then I dont see why he shouldnt be allowed to do so, delaying a milestone or not. My suggestion wouldn't cope with that, either. Except by "officially" pirating the assignment. I would argue that a competent individual administering a relatively small amount of assignments manually should be able to make the required judgements. BTW I agree with Alex that milestones are not the be-all and end-all of the project. I'm simply trying to find something that will satisfy the "pushers" without upsetting those who are perfectly happy for a run to take a long time. Providing, of course, completion does eventually occur. 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: prime95 - v21 progress
-Original Message- From: Jeramy Ross [EMAIL PROTECTED] To: [EMAIL PROTECTED] [EMAIL PROTECTED] Date: Sunday, March 11, 2001 5:10 AM Subject: Re: Mersenne: prime95 - v21 progress Martijn wrote: *SNIP* Another solution that will work: Have as default a 7 day check in period at most and only a grace period of 7 days (not 60). Let the user set the check in period to a higher value only via the expert menu and after results have been checked in. That way abandoned exponents get released in 14 instead of 80 days. Kind regards, Martijn That idea sounds the least painful of all that have been discussed so far (to me at least), and since the discussion of what people want in a new version of Prime95 has also been floating arround... this sounds like a good suggestion to be submitted. It would not only take care of a problem, but would also not be so harsh to those who own slower machines. A win-win situation from those points of view. Great idea, Martijn!! Best wishes, Jeramy I agree this is a good idea, although the 7-day grace period may be a little drastic. But even reducing that just from 60 to 30 days (along with a 7-day default check in) would release abandoned exponents in 37 days instead of 88. This would recycle them more than twice as fast, greatly enhancing the odds of someone eventually getting the assignment who will actually finish it. And I don't believe it would be necessary to put the default changes in the expert menu; most people seem to not bother changing the defaults anyway, and it is already impossible to change the defaults until after the first exponent is assigned (unless you start/set up the program while offline). As others have mentioned, the problem is NOT slow machines but rather abandoned exponents, which has nothing to do with machine speeds. This change would have no effect on those of us who have been regularly completing our work and reporting it in, regardless of whether or not we have slow machines. Regards, Steve Harris _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: R: Mersenne: prime95 - v21 progress
At 08:13 AM 3/10/01 +0100, you wrote: this without any polemic spirit, please belive me, enjoy yourself and try to explain yourself why it is tolerable to wait one year or more for testing M(3325) and is not tolerable to wait eight years for testing M(325): is it the first figure really more important than the first just because of its size? Actually, the latter is more important (to me) because of its proximity to the lower boundary of untested numbers. Not knowing the former is of little importance today in comparison, because there are substantially more untested exponents below it. I'm only saying that I'd rather see the profject working a bit more linearly, as opposed to "first come first served" now, so that all machines can contribute while at the same time, we make linear progress from zero forward, instead of expending the bulk of our efforts on exponents double or higher above those untested ones. _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
RE: Mersenne: prime95 - v21 progress
I agree this is a good idea, although the 7-day grace period may be a little drastic. But even reducing that just from 60 to 30 days (along with a 7-day default check in) would release abandoned exponents in 37 days instead of 88. This would recycle them more than twice as fast, greatly enhancing the odds of someone eventually getting the assignment who will actually finish it. The downside is that there would probably be a great increase in the number of assignments which are still running but don't complete before the expiry date. Clearly there is a balance to be struck somewhere, but 7 days seems to me to be _ludicrously_ short. In fact, as assignments take progressively longer to run, the "grace period" should be extending, rather than contracting. We should also bear in mind the very valuable contributions made by those people who do not have permanent (or near-permanent) network connections, and those people who are using clients without the PrimeNet communication protocol. Requirement to check in frequently is off-putting to these people. (Some would put it a lot stronger than that!) I don't think we want to risk driving these people out of the project. I have said nothing about not being able to set a longer period, just set the default one low! As others have mentioned, the problem is NOT slow machines but rather abandoned exponents, which has nothing to do with machine speeds. I fail to see how reducing the check-in interval would have any impact on the "problem". Those people who are checking in every 28 days aren't running into the 60-day expiry deadline. The reduction of the default check-in interval would recycle the exponents quicker of people grabbing but not testing exponents. The 60 day expiry value is a server parameter, not a client parameter. In any case, as I explained above, I think that a drastic reduction in the value would be dangerous. Certainly not for first time assignments, the results are re-used as double checks right away. People with non frequent contact could put their time between check ins up. (could we let the grace period be the same as the time set between check ins on the client with a minimum of 7 days?) Might I suggest a couple of alternative approaches. Both of these would require the identification of exponents which are "seriously lagging" - perhaps the 100 smallest outstanding LL and the 100 smallest outstanding DC assignments. (1) Removing these assignments from PrimeNet and managing them seperately. Anyone who is prepared to make special arrangements to acquire these assignments is unlikely to default by reason of lack of commitment. Seems reasonable, see proposal on bottom (which saves the 100 smalles checks needed) (2) Alternatively, awarding double PrimeNet CPU time credit for the completion of these assignments. The downside to this is that, as well as requiring changes to the server software, recycled "small" exponents would have to be released at random times of the day, to prevent them being systematically "grabbed" by a few users. Will not work, someone working on a exponent is always good (also if he/she takes a year) Someone who will forget about their exponent and not check in will not care if he/she would have gotten double points for these assignments. Now the next proposal: Let Primenet reassign automatically only exponents with a exp date -5. Make it possible that a user can explicitly request an exponent with an exp date between -0.1 and -5. (Primenet could have 2 reactions ok (and the list gets updated at the next hour) or not ok (exponent not expired (yet/anymore)/exponent has been checked in)) Primenet is already able to assign unassigned specific exponenents to someone. (Just try to send an update on a unreserved exponent, you will be listed) When trying to check in a status on an reserved exponent you will get an error message. I just do not know how it exactly works with recycling expired exponents. Most of the time they come aroun 6:00 status time, but sometimes the come really bulky at "random" times. This makes it possible for the people that want to "hunt" the holes to hunt those holes. Maybe we could also update the client to not accept exponents it will work on for longer than lets say 2 years, and then let the client get another type of assignment automatically. Kind regards, Martijn _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
"Brian J. Beesley" wrote: (1) Removing these assignments from PrimeNet and managing them seperately. Anyone who is prepared to make special arrangements to acquire these assignments is unlikely to default by reason of lack of commitment. Someone is (or was?) doing something similar - David Campeau (sp?), aka diamonddave, who set up scripts to grab small exponents right after the Primenet servers recycle run and completed them in ascending order of size. And boy, did he get flamed for it. Right, flamed - because some only looked at the 20 or so small exponents he had reserved and tought he was holding them back. Maybe those who complain about slow "linear" progress could take on a similar approach? (except for the getting flamed part, I hope.) Write a script that (or get up early/stay up late to) get exponents right after the servers recycle run and immediately release the biggest ones so you have enough work for, say, 30 days. Maybe David can help with the scripts or general tips for maintaining the exponents list, *especially* safe guards so that reserved exponents dont pile up in an uncontrolled manner. If everyone sticks to the rules - dont poach, dont hog exponents - this should solve most of the problem with the least possible hassle. Those who want to see milestones soon can help to get there, the rest can just go on with their regularly assigned work. This doesnt help with the exponents that are currently reserved with a expected run time of several years - but if the user abandons them, the server will recycle them 60 days after the last update from the user which doesnt seem like an overly long delay for the project. If the users chooses to complete the exponent on a slow machine or on a machine that is not in use very often, then I dont see why he shouldnt be allowed to do so, delaying a milestone or not. Ciao, Alex. _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
I said the idea to set the DEFAULTS lower was a good one. Anyone who is serious about the project could very easily change those defaults. I have several machines with dialup access which don't connect with the server for weeks at a time, but a 7-day default would not affect me in the least. I change the settings on each machine depending on its circumstances; I have some set for 39 days. (Even a 1-day default wouldn't bother those machines, but I did say I thought 7 days might be a little drastic.) The people who abandon exponents are the ones who would not bother changing the defaults, hence returning them sooner. Steve Harris -Original Message- From: Brian J. Beesley [EMAIL PROTECTED] To: Steve [EMAIL PROTECTED] Cc: [EMAIL PROTECTED] [EMAIL PROTECTED] Date: Sunday, March 11, 2001 2:44 PM Subject: Re: Mersenne: prime95 - v21 progress On 11 Mar 2001, at 7:55, Steve wrote: Another solution that will work: Have as default a 7 day check in period at most and only a grace period of 7 days (not 60). Let the user set the check in period to a higher value only via the expert menu and after results have been checked in. That way abandoned exponents get released in 14 instead of 80 days. That idea sounds the least painful of all that have been discussed so far (to me at least), and since the discussion of what people want in a new version of Prime95 has also been floating arround... this sounds like a good suggestion to be submitted. It would not only take care of a problem, but would also not be so harsh to those who own slower machines. A win-win situation from those points of view. Great idea, Martijn!! I agree this is a good idea, although the 7-day grace period may be a little drastic. But even reducing that just from 60 to 30 days (along with a 7-day default check in) would release abandoned exponents in 37 days instead of 88. This would recycle them more than twice as fast, greatly enhancing the odds of someone eventually getting the assignment who will actually finish it. The downside is that there would probably be a great increase in the number of assignments which are still running but don't complete before the expiry date. Clearly there is a balance to be struck somewhere, but 7 days seems to me to be _ludicrously_ short. In fact, as assignments take progressively longer to run, the "grace period" should be extending, rather than contracting. We should also bear in mind the very valuable contributions made by those people who do not have permanent (or near-permanent) network connections, and those people who are using clients without the PrimeNet communication protocol. Requirement to check in frequently is off-putting to these people. (Some would put it a lot stronger than that!) I don't think we want to risk driving these people out of the project. As others have mentioned, the problem is NOT slow machines but rather abandoned exponents, which has nothing to do with machine speeds. I fail to see how reducing the check-in interval would have any impact on the "problem". Those people who are checking in every 28 days aren't running into the 60-day expiry deadline. The 60 day expiry value is a server parameter, not a client parameter. In any case, as I explained above, I think that a drastic reduction in the value would be dangerous. Might I suggest a couple of alternative approaches. Both of these would require the identification of exponents which are "seriously lagging" - perhaps the 100 smallest outstanding LL and the 100 smallest outstanding DC assignments. (1) Removing these assignments from PrimeNet and managing them seperately. Anyone who is prepared to make special arrangements to acquire these assignments is unlikely to default by reason of lack of commitment. (2) Alternatively, awarding double PrimeNet CPU time credit for the completion of these assignments. The downside to this is that, as well as requiring changes to the server software, recycled "small" exponents would have to be released at random times of the day, to prevent them being systematically "grabbed" by a few users. Regards Brian Beesley _ 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
Re: Mersenne: prime95 - v21 progress
Brian J. Beesley wrote: I fail to see how reducing the check-in interval would have any impact on the "problem". Those people who are checking in every 28 days aren't running into the 60-day expiry deadline. For one, the reduced check in time would allow the closer watch of "suspect" users who download a exp., run the program for a few days and then ditch out on the project. While it is true that people may run the program for upwards of a month or so and then ditch. The idea of having them check in about every week AS A DEFAULT too keep tabs on them while they have their FIRST exp. would aid in finding someone who ditches out. The process in which we use this aid is up in the air. The 60 day expiry value is a server parameter, not a client parameter. In any case, as I explained above, I think that a drastic reduction in the value would be dangerous. I think it is realized that the 60 day value is a server parameter, and the change in value would be only effective for FIRST TIME users. Besides, I don't know of many users who have a problem checking in every week to keep things moving smoothly. Best wishes, Jeramy _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
On 9 Mar 2001, at 17:27, Jeff Woods wrote: Actually, the next obvious milestone is checking all below M(6972593) for the first time. There are 67 exponents unchecked at all below that exponent, and that number has been VERY VERY slow to reduce, mainly due to number campers or 386's trying to test that number. Or fast systems slowing because an animated screensaver is being run? This can easily happen if you ask someone else to run Prime95 on their system as a favour. I also massaged today's assignments report, and found that there are over 200 exponents assigned over a year ago (and some as far back as 1998), NOT including those expected to take that long (i.e. 33 million+). Some exponents have been run for over a year, and have "days to run" estimates of 2900 days or more -- yes, nearly EIGHT YEARS. I remember getting "wound up" about this shortly after I joined the project. George replied that these problems have a way of sorting themselves out; experience proves him right. The point is that we could crank through these laggards if the Primenet server would have simply ensured they were assigned to a "top 1000" producer, or to a machine of sufficient calibre and reliability (historically, per prior test results). To which [EMAIL PROTECTED] (Mikus Grinbergs) replied: Is that what we want - an elitist organization which SEGREGATES those participants to whom we do not attribute "sufficient calibre" Well _I_ don't! Back to Jeff Woods: You said you had a good reason not to do that, but didn't want to post it here (you were going to mail it privately to Henk). Why not discuss it here? Now I believe Henk is a responsible individual, and I've no reason to suspect Jeff is any less so. The system we have at the moment is reasonably robust and will stand a certain amount of abuse. However, abuse on a large scale will break it. I don't want to be responsible for that. What I mailed to Henk privately amounts to a minor form of abuse of the system. One paragraph of that private message reads: (start quote) Obviously you should be careful when doing this, else you are likely to be accused of "pirating" assignments. Also there would be chaos if several people were doing this, which is why this reply is being sent to you only and not to the list. (end quote) Jeff, if you (or anyone else, for that matter) want to take advantage of the idea I mailed to Henk, I suggest you mail Henk privately and discuss amongst yourselves how you're going to coordinate your combined effort. My contribution to your "sturmgruppe" ends here because I don't believe that anyone's CPU cycles are inherently more valuable than anyone else's. Oh, and if you happen to find a new Mersenne prime whilst you're working in this mode, I would hope that you'd be prepared to share the credit with any other person who happens to "own" the PrimeNet assignment at the time. 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: prime95 - v21 progress
Hello, because of the discussion about the speed on the slowest exponents I did some calculations: Result: The limits like 750 MHz or even 350 MHz for the smalles exponents are completely ridiculous, certainly if we do not take into account the number of it is on, a 1000 MHz machine on for 8x5 hours a week will take longer than PII 233 on 24x7. If we want to get it done as soon as possible we must calculate what would be fastest: Observation the relatively slow machines are often used as servers / private firewalls and on 24x7 On the other hand relatively slow machines may be hardly on. Some calculations (benchmark page) on the fictive exponent 400 yields 24x7 on a: 486@33: 1Y+182 days (Ok that seems long) PI@60: 43 days (Seems we can wait on that, as long as the machine is reliable) PII@233: 10 days (Seems as of now it really does not matter anymore) Cel@300: 8 days PIII@450: 6 days PIII@1000: 3 days Athlon@1200: 2 days Some calculations on the ficive exponent 800 (486 left out) PI@60 : 177 days (Doubtfull) PII@233 : 40 days (Seems we can wait on that, as long as the machine is reliable) Cel@300 : 31 days PIII@450 : 23 days PIII@1000 : 12 days Athlon@1200 : 9 days Some calculations on the ficive exponent 1200 (486 left out) PI@60 : 1Y+29 days (Seems to long) PII@233 : 87 days Cel@300 : 68 days PIII@450 : 50 days PIII@1000 : 27 days Athlon@1200 : 17 days When we take into account that the timeout offset is 60 days (so someone starting an assignment and not doing anything at all costs 60 days, nobody works on the exp. so that is a delay of the entire project) We should probably reassign exponents to machines that have already finished at least 2 exponents, and based on that info will return this assignment within 60 days. That would be like doublechecks reassigns for PI / PII / Celeron, Primality reassigns for PIII / Athlon. Exponents in the lower 10% of a range should be reassigned if no progress has been reported for 60 days. A good better solution optimizing for progress would be to re-assign expired exponents to machines that have finished exponents already and will finish them in for instance approximately 20 days (Assigning the smaller exponents to slower machines and larger exponents to faster machines) The actual number of days can be calculated from the "ballpark". So that optimal progress is made. We must try to keep slow machines in as long as possible as they really contribute to the progress. So optimal progress will be made by giving smaller machines small exponents and larger machines large. Kind Regards, Martijn _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
On 9 Mar 2001, at 0:45, Henk Stokhorst wrote: Maybe it would be a good idea to have a special version of prime95 that has an option to request exponents that have expired after having been reserved for a long time without any progress being on the work for that exponent. No need for this. Henk, I'll mail you privately explaining why. The server should issue those exponents only to people who have that option. Two problems here: (a) needs a server fix; with due respect I think there are other things which need fixing more urgently e.g. keeping proper tabs on P-1 so that work is not replicated wastefully. (b) reeks of elitism. Whilst I understand your motives, I think there are a lot of people who would object for that reason. That would help prevent exponents expiring multiple times. Is that _really_ a problem? If anything it only points out a lack of commitment to the project amongst some of the contributors. Regards Brian Beesley _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: prime95 - v21 progress
Robert van der Peijl [EMAIL PROTECTED] wrote on Wed, 7 Mar 2001 03:11:22 +0100: We would have to uglify (is that good English?) the user interface however, to make the whole thing believable. In answer to your question, Alice didn't think so: `I only took the regular course,' said the Mock Turtle with a sigh. `What was that?' inquired Alice. `Reeling and Writhing, of course, to begin with,' the Mock Turtle replied; `and then the different branches of Arithmetic-- Ambition, Distraction, Uglification, and Derision.' `I never heard of "Uglification," Alice ventured to say. `What is it?' The Gryphon lifted up both its paws in surprise. `What! Never heard of uglifying!' it exclaimed. `You know what to beautify is, I suppose?' `Yes,' said Alice doubtfully: `it means--to--make--anything-- prettier.' `Well, then,' the Gryphon went on, `if you don't know what to uglify is, you ARE a simpleton.' from Alice's Adventures in Wonderland, by Lewis Carroll Lewis Carroll was a Professor of Mathematics (a rather dull one, if the stories are true) but this small book, written to entertain a little girl, is what has eternalized his name (not to put a damper on anyone's day). By the way... "Eternalized"... Is that good English? Best regards, Dennis Pope _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
At 11:09 AM 3/9/01 +, you wrote: That would help prevent exponents expiring multiple times. Is that _really_ a problem? If anything it only points out a lack of commitment to the project amongst some of the contributors. *THAT* is precisely the reason that such people should not be assigned the very exponents that are most necessary to reaching the next "milestones". For those of us that measure progress both in Teraflops and in milestones, we've not seen much lately, mainly due to the things that Henk was complaining about. _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
At 10:00 PM 3/9/01 +, you wrote: Isn't the "problem" just that the milestones are rather a long way apart at the moment? Does it _really_ matter that there's an outstanding exponent around 3.25 million which needs double-checking when there are so many more to go before we complete double-checking up to 6973593 - which is the next obvious milestone, after completing coverage of first tests up to that value, and always assuming that we don't find another prime smaller than Hajratwala's Number in the meantime? Actually, the next obvious milestone is checking all below M(6972593) for the first time. There are 67 exponents unchecked at all below that exponent, and that number has been VERY VERY slow to reduce, mainly due to number campers or 386's trying to test that number. I also massaged today's assignments report, and found that there are over 200 exponents assigned over a year ago (and some as far back as 1998), NOT including those expected to take that long (i.e. 33 million+). Some exponents have been run for over a year, and have "days to run" estimates of 2900 days or more -- yes, nearly EIGHT YEARS. The point is that we could crank through these laggards if the Primenet server would have simply ensured they were assigned to a "top 1000" producer, or to a machine of sufficient calibre and reliability (historically, per prior test results). You said you had a good reason not to do that, but didn't want to post it here (you were going to mail it privately to Henk). Why not discuss it here? Somehow the laggards always do get swept up in the end; there seems to be an adequate supply of self-appointed "gap fillers" (or pirates, according to your point of view). The result is that many of the assignments which are the "longest overdue" eventually end up with three, four or even more completed LL tests. Said "overchecking" could be eliminated by ensuring that the oldest exponents are assigned to the most reliable machines possible. _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
On Fri, 09 Mar 2001 10:12:43 -0500 Jeff Woods [EMAIL PROTECTED] wrote: At 11:09 AM 3/9/01 +, you wrote: That would help prevent exponents expiring multiple times. Is that _really_ a problem? If anything it only points out a lack of commitment to the project amongst some of the contributors. *THAT* is precisely the reason that such people should not be assigned the very exponents that are most necessary to reaching the next "milestones". For those of us that measure progress both in Teraflops and in milestones, we've not seen much lately, mainly due to the things that Henk was complaining about. I consider the attitude "GOTTA GET IT DONE" to be rat-race oriented rather than thank-you-for-participating oriented. It implies to those with less-than-state-of-the-art equipment: "__You__ are an obstacle in the way of GOTTA GET IT DONE -- there are no seats for you on this bus". Would the world come to an end tomorrow if the GIMPS participants were more tolerant of those who do __not__ measure progress in Teraflops ? mikus _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
On Fri, 09 Mar 2001 17:27:42 -0500 Jeff Woods [EMAIL PROTECTED] wrote: The point is that we could crank through these laggards if the Primenet server would have simply ensured they were assigned to a "top 1000" producer, or to a machine of sufficient calibre and reliability (historically, per prior test results). Is that what we want - an elitist organization which SEGREGATES those participants to whom we do not attribute "sufficient calibre" ? Why not discuss it here? I thought this project was an association of VOLUNTEERS. I believe we should welcome ALL who offer to participate. mikus _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
Joshua Zelinsky wrote on Thursday, March 08, 2001 3:38 AM I'm betting this is a rather significant penalty! You have to use the wall clock to time your iterations / minute. Prime95 will report the same time, but it is only measuring the time to do an iteration, not the additional time to write to the display. Still, the CPU time required to write one line to the display is considerably less than that required to execute one iteration of an LL test on a large exponent. I just did a test as George suggested. Thanx, I never get around to actually testing things. I had been wondering myself about the performance penalty. The average ratio for the with results at every 1 iteration came out to be .540 per for .487 every 100 iterations. I'm running a PIII 600 megahertz and I didn't realize that I had my webbrowser open when I ran the test which seriously invalidates the results. That much difference -- 10 PERCENT! That's too rich for my blood. I'll have to shtick with the standard setting. I might run a test again later when I get a chance... Let us know what you come up with; I own a P3-600 myself. By the way, I think Joshua's suggestion from Feb 12 2001 for Prime95 to have an option for disk writes to occur based on the number of iterations, rather than the time passed, is a sound idea. I hope it will make it into Prime95 v21. Brian Beesley wrote on Wed, 7 Mar 2001 20:56:12 - Don't fix what ain't broke. Right on. If you must change the name [Prime95]... Yes, if someone successfully challenges our rights to that name. Has anyone of us bothered to register the name Prime95 as a trademark? I sure haven't. ..., go for PriMe. That looks like a very modern name to Me. By the way, a release of v21 in more than 6 months would suit me just fine. And, George, hand-holding can be rewarding too, it just takes up so much time. I think most people wouldn't mind the slight file size increase with the optional screensaver part. George's efforts to understand the strengths and weaknesses of the P4 architecture could present an opportunity for the engineers at Intel to pitch in. It may be in the chip manufacturer's interest! Also, on Feb 3 2001 Joshua Zelinski wrote: Recently people have been discussing that the long times it takes to finish a batch of work makes GIMPS less appealing than other distributed computing systems. What if we gave out factoring in small bunches... Yeah, I like that idea! Giving peoply bite-sized chuncks would make them come back, craving for more. It would make GIMPS less appalling, that is, more appealing to the novice. Even a factoring assignment from 2^n to 2^(n+1) could theoretically be cut up into up to 16 little snack-sized appetizers. (All this talk is starting to wet my appetite) I'm off to lunch now, cheers. Robert. _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
L.S., Maybe it would be a good idea to have a special version of prime95 that has an option to request exponents that have expired after having been reserved for a long time without any progress being on the work for that exponent. The server should issue those exponents only to people who have that option. That version should only be available to people who have fast (700 MHz or more) machines running most of the day. That would help prevent exponents expiring multiple times. YotN, Henk Stokhorst _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
Henk Stokhorst wrote: L.S., Maybe it would be a good idea to have a special version of prime95 that has an option to request exponents that have expired after having been reserved for a long time without any progress being on the work for that exponent. The server should issue those exponents only to people who have that option. That version should only be available to people who have fast (700 MHz or more) machines running most of the day. That would help prevent exponents expiring multiple times. YotN, Henk Stokhorst Nope bad idea smaller exponent = smaller runtime = lower clock frequency so it would then be better to have them assigned to machines that are slow real pentiums for instance (It does really not matter if such an exponent is finished in 4 or 30 days. The multple expiering problem is an entirely fake one. It does not hinder progress, it only makes (relatively small) exponents unavailable for 90 days, in that time we work on other exponents and life goes on. Martijn Kruithof _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
Hi, At 03:11 AM 3/7/2001 +0100, Robert van der Peijl wrote: Could it be that's because everyone is holding their breath for prime95 v21? But that might still take quite a while, since AFAIK no public announcement has been made about its scheduled completion date. I have no idea when a new release will be available. I've done no work on prime95 from v20 release until last December. Since then all available time has been devoted to a P4 version of prime95. Much of this has been writing code samples to understand the strengths and weaknesses of the architecture. This is not particularly easy - a lot like reverse engineering chip design. After a P4 recoding, I'll look at what new features to add to a v21 release. I liked the screensaver ideas recently suggested, but would like to have one executable that runs like it does now, as a screensaver only, or both. I've received many other minor suggestions which I dutifully write down and implement the easiest and/or most useful. In any event, I cannot imagine a release of v21 in less than 6 months. Hey, how about calling it primeXP? Several versions ago, I thought about renaming prime95 to prime98. However, the upgrade issues are enormous. I can just imagine hundreds of users running both prime95 and primeXP on their machine - then a flood of email questions and hand-holding. I was thinking of switching my iterations between screen outputs to 1 for Prime95, just to make it "feel faster." However, I wasn't sure how much resources it would take up. I checked and it seems negligible. Is that true? I'm betting this is a rather significant penalty! You have to use the wall clock to time your iterations / minute. Prime95 will report the same time, but it is only measuring the time to do an iteration, not the additional time to write to the display. Have fun, George _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: prime95 - v21 progress
I'm betting this is a rather significant penalty! You have to use the wall clock to time your iterations / minute. Prime95 will report the same time, but it is only measuring the time to do an iteration, not the additional time to write to the display. Still, the CPU time required to write one line to the display is considerably less than that required to execute one iteration of an LL test on a large exponent. I just did a test as George suggested. The average ratio for the with results at every 1 iteration came out to be .540 per for .487 every 100 iterations. I'm running a PIII 600 megahertz and I didn't realize that I had my webbrowser open when I ran the test which seriously invalidates the results. I might run a test again later when I get a chance but right now it looks like George is right. Sincerely, 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
Mersenne: prime95 - v21 progress
Joshua Zelinsky wrote on Tuesday, March 06, 2001 at 17:08:37 -0500 : The list has been pretty quiet lately... Could it be that's because everyone is holding their breath for prime95 v21? But that might still take quite a while, since AFAIK no public announcement has been made about its scheduled completion date. Of course, that date would depend on the intended new features. Still, some indication would useful -- say, between 3 and 9 months from now?? Personally, it would help me in deciding whether to continue translating the v20 readme file into my language (Dutch), or to wait for the new stuff. Hey, how about calling it primeXP? It would allow the new user to run only 50 iterations, after which time registration is mandatory. We would have to uglify (is that good English?) the user interface however, to make the whole thing believable. And, existing software should no longer be able to execute with our new version. Those two criteria may be quite hard for us to fulfill. What software company could we ask to implement those changes for us? The Good Thing (TM) would be, that no new features would be necessary. But then we'd have to charge a lot of money for the product. The revenues could go to Bill Gates' family -- to help them extend their house. This is a little embarrassing but ... That's quite allright, this is a private conversation, right? :-) I was thinking of switching my iterations between screen outputs to 1 for Prime95, just to make it "feel faster." Actually, I have several alternative suggestions to make it, how did you put it, 'feel faster'? Do something fun. Don't look at prime95's output. Go read a good book. Go on a long vacation. You name it. However, I wasn't sure how much resources it would take up. I checked and it seems negligible. Is that true? If you say so. You say you checked, right? But, seriously: yes, a first time LL test on a new exponent would complete about as fast with "iterations between screen outputs" = 1. Are there any other reasons to set the number that low? You tell me -- are there? Hope this answers your question. It doesn't answer mine. Cheers, Robert. P.S. Sorry for this lame response, but I was getting bored looking at prime95's screen output. ;-) _ Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers