|
[QUOTE=diamonddave;282192]Hmmm the logic is flawed... PrimeNet reserve at least 2-4 time the number of completed test per day. So if 200 test gets completed on a given day, anywhere between 400 and 800 will have been assigned. It's easy to see why when we know that only 25-50% of any assignment ever get completed.[/QUOTE]
Damn! You're right. [QUOTE=diamonddave;282192]Since we know that Prime95 will never further factor exponent factored to 71 in those range, even if it would have been optimal GPU-wise. If the spider could somehow monitor the number of exponent buffered by PrimeNet so that it didn't have to advance the wavefront and always release the lowest exponent factored to 71. Then I think the system would be optimal! :cool:[/QUOTE] Well, right now the system passes back to PrimeNet any candidates above M53 factored to 71 and above, and any above 55M regardless of state. (The two currently "out" in the M55 block were assigned before this change in the heuristics.) So based on your above analysis (which I agree with), if any GPU Workers want to adapt this strategy all they need to do is adjust the limits on the LLTF Assignment page. |
[QUOTE=diamonddave;282194]This doesn't reserve the exponent (BTW it's what I use). If I pick a month worth of work in a range, it might not have been assigned when I grabbed it, but that status could have changed in the following month and we are back at square one.
[/QUOTE] Admittedly there is a small window. There is a chance that from the time I generate the list until they are added to my worktodo.txt someone may have assigned them. If I notice that some of my assignments show up as N/A I would remove them from my worktodo.txt. However, I tend to work in ranges that don't have a lot of activity. The window will be smaller if you added the new work to worktodo.txt instead of .add. |
[QUOTE=petrw1;282203]Admittedly there is a small window. There is a chance that from the time I generate the list until they are added to my worktodo.txt someone may have assigned them. If I notice that some of my assignments show up as N/A I would remove them from my worktodo.txt. However, I tend to work in ranges that don't have a lot of activity. The window will be smaller if you added the new work to worktodo.txt instead of .add.[/QUOTE]
With P-1 the window is small since Prime95 will reserve the exponent. With further TF there is no such thing and no way to mark the exponent as reserved. Or is there an unknown way to reserve lets say range 11.1M to 11.2M for TF 64->65 on PrimeNet that I don't know of? |
[QUOTE=cheesehead;282148]That depends rather heavily on the B1 and B2 bounds, doesn't it?[/QUOTE]
cheesehead, I'm simply trying to get a ballpark estimate on how long it should take to complete a P-1. The ones that get assigned to my not-so-modern CPUs prior to starting an LL test take just a couple of days. Therefore I wonder why it would take nine months for someone to complete a P-1, as in the case under discussion. Rodrigo |
[QUOTE=diamonddave;282192]PrimeNet assign for LL testing at least 2-4 time the number of completed test per day. [/QUOTE]
Makes me wonder if with todays cheap RAM and fast networks there would be merit in ocassionally sending the LL workfile to the server so that if it is abandoned the new assignee does not have to start over. Ocassionally could be once a day or two ... and maybe only after the test reaches 10 or 25% ... and maybe only if it assigned to a non-trusted tester |
[QUOTE=petrw1;282206]Makes me wonder if with todays cheap RAM and fast networks there would be merit in ocassionally sending the LL workfile to the server so that if it is abandoned the new assignee does not have to start over.
Ocassionally could be once a day or two ... and maybe only after the test reaches 10 or 25% ... and maybe only if it assigned to a non-trusted tester[/QUOTE] In this day and age of P2P network like bittorrent... We could distribute the seeding of backup file (encrypted so the bad poacher wouldn't steal your work), lets say in 1 Ghz-Day increment. Then if an exponent is reclaimed by PrimeNet the latest file would be marked as a DC candidate. When the new worker get to the same point as the previous worker, the residue would be compared and if we have a match the backup file would be marked as DC accepted. Now when the exponent get assigned as DC or reassigned as LL the known good backup file would be used with the decryption key provided by PrimeNet... |
cheesehead, in what way are TF and P-1 assignments overlapping? You said that multiple times, yet I can't see why. They search for very different factors, as markr and ckdo well know. What markr did was not assigned to anybody.
@dd: I think 1 GHz is a bit often. Maybe every 5 or 10 GHz? Either way, a massive increase in throughput I think will be the result. But that does not belong in this thread anyways. |
[QUOTE=Rodrigo;282205]I'm simply trying to get a ballpark estimate on how long it should take to complete a P-1.[/QUOTE]At the time it was an untested exponent, given a decent machine to work on, M7,018,901 would've spent about [url=http://mersenne-aries.sili.net/prob.php?guess_saved_tests=2&exponent=7018901]0.06Ghz-days[/url] on P-1. As [i]cheesehead[/i] said, the selected bounds can have a tremendous impact on the time spent. As it happens, [i]ckdo[/i] reports spending roughly 0.366GHz-days on the effort due to [url=http://mersenne-aries.sili.net/prob.php?exponent=7018901&b1=445000&b2=13572500&factorbits=63]higher bounds[/url] (which were [url=http://mersenne-aries.sili.net/7018901]still insufficient[/url] to find the factor that was found by TF).
[QUOTE=petrw1;282206]Makes me wonder if with todays cheap RAM and fast networks there would be merit in ocassionally sending the LL workfile to the server so that if it is abandoned the new assignee does not have to start over.[/QUOTE]There are roughly 150,000 L-L assignments out at any given time. Assuming a nominal savefile size of 5MB, that's not just 750GB of storage (disk space is cheap, that's unlikely to be much of a problem) but a hefty amount of data transfer -- 10+TB/month if all workers report every two days. That's what makes the scheme unworkable. However, narrowing down the criteria to only (for example) [once per month + assignment >25% complete + estimated completion >30days] that would eliminate a large portion of the potential reports. It would still be a lot of storage and data transfer for what might be limited benefit. Is there any data on [i]how far[/i] the 75% assigned-then-dropped assignments ever get? I suspect a bulk of them are registered, then the user sees it'll take a month to complete and never runs Prime95 again. Or were "accidentally" registered when trying to run overclocking stress tests. Or any number of similar never-intend-to-complete assignments on which very little useful work has been done. |
[QUOTE=James Heinrich;282213]At the time it was an untested exponent, given a decent machine to work on, M7,018,901 would've spent about [url=http://mersenne-aries.sili.net/prob.php?guess_saved_tests=2&exponent=7018901]0.06Ghz-days[/url] on P-1. As [i]cheesehead[/i] said, the selected bounds can have a tremendous impact on the time spent. As it happens, [i]ckdo[/i] reports spending roughly 0.366GHz-days on the effort due to [url=http://mersenne-aries.sili.net/prob.php?exponent=7018901&b1=445000&b2=13572500&factorbits=63]higher bounds[/url] (which were [url=http://mersenne-aries.sili.net/7018901]still insufficient[/url] to find the factor that was found by TF).
There are roughly 150,000 L-L assignments out at any given time. Assuming a nominal savefile size of 5MB, that's not just 750GB of storage (disk space is cheap, that's unlikely to be much of a problem) but a hefty amount of data transfer -- 10+TB/month if all workers report every two days. That's what makes the scheme unworkable. However, narrowing down the criteria to only (for example) [once per month + assignment >25% complete + estimated completion >30days] that would eliminate a large portion of the potential reports. It would still be a lot of storage and data transfer for what might be limited benefit. Is there any data on [i]how far[/i] the 75% assigned-then-dropped assignments ever get? I suspect a bulk of them are registered, then the user sees it'll take a month to complete and never runs Prime95 again. Or were "accidentally" registered when trying to run overclocking stress tests. Or any number of similar never-intend-to-complete assignments on which very little useful work has been done.[/QUOTE] In a distributed project you would think that keeping a backup of the file would also be distributed? Think P2P. So 5000 active user in last 30 days and lets have 10 copies distributed across the network. Also of those 150000 How many actually are being worked on and not just sitting in a queue? So we get 750 / 5000 * 10 or about 1.5 Gb per user. Numbers could be tweaked, but its definitely doable. With such a system, no more P-1 work would be wasted and work could be continued in the future. |
But coding a P2P system is way beyond GIMPS's coding abilities, not to mention it negates the primary pull of P95, which is just being a background program that uses CPU cycles and minimal bandwidth.
|
[QUOTE=James Heinrich;282213]At the time it was an untested exponent, given a decent machine to work on, M7,018,901 would've spent about [URL="http://mersenne-aries.sili.net/prob.php?guess_saved_tests=2&exponent=7018901"]0.06Ghz-days[/URL] on P-1. As [I]cheesehead[/I] said, the selected bounds can have a tremendous impact on the time spent. As it happens, [I]ckdo[/I] reports spending roughly 0.366GHz-days on the effort due to [URL="http://mersenne-aries.sili.net/prob.php?exponent=7018901&b1=445000&b2=13572500&factorbits=63"]higher bounds[/URL] (which were [URL="http://mersenne-aries.sili.net/7018901"]still insufficient[/URL] to find the factor that was found by TF).
[/QUOTE] Thanks, James, for the info and the links. If I read the information correctly, [I]ckdo[/I] has no shortage of [URL="http://www.mersenne.org/report_top_500/"]firepower[/URL] available. Thus, it leads one to wonder why such a tiny assignment[URL="http://www.mersenneforum.org/showpost.php?p=282119&postcount=946"], reportedly[/URL] allocated in March, would have still not been completed well into December (as of the time that this discussion started). There must be a good explanation. Rodrigo |
[QUOTE=Rodrigo;282223]Thus, it leads one to wonder why such a tiny assignment[URL="http://www.mersenneforum.org/showpost.php?p=282119&postcount=946"], reportedly[/URL] allocated in March, would have still not been completed well into December (as of the time that this discussion started). There must be a good explanation.
Rodrigo[/QUOTE] It will be a case of making sure the machines are well stocked with future work. [URL="http://www.mersenneforum.org/showpost.php?p=222205&postcount=454"]This old posting[/URL] offers some insight. |
[QUOTE=Brian-E;282250]It will be a case of making sure the machines are well stocked with future work. [URL="http://www.mersenneforum.org/showpost.php?p=222205&postcount=454"]This old posting[/URL] offers some insight.[/QUOTE]
Brian-E, Not only does that old posting offer some insight, it offers a lot of insight -- thanks! I take it that ckdo has very infrequent access to a powerful machine(s), and that's the reason he wants to stock up on assignments. It's all much clearer now. Rodrigo |
[QUOTE=KyleAskine;282172]Do you? Not only this incident, but there was the other case where you tried to poach from a completely different project!! And you admitted you knew about it! [URL]http://mersenneforum.org/showpost.php?p=279785&postcount=121[/URL]
If you had found a factor the amount of wasted processor time would have been absolutely mammoth from NFS@home. There are tons of numbers out there that have not been assigned to people to do whatever worktype you want to do. Don't be selfish. Think about the people who are working on the exponent. I would be pissed if I turned in a P-1 and was not given credit because someone else already found that factor with TF. Or if I was working on it and someone told me they had already done it.[/QUOTE] Yes, I do! As I said, my definition of poaching is different, and I don't give a dime on yours. Sue me. I admitted and I will always admit of occasionally poaching, I did that a couple of times. If I would turn it a factor for that, first: how do you know I would report it before nfs@home do? Well, I would, most probably, but how can you accuse me before I do something? Does it fit into your head that I could possible already have a factor for that number and keep silent? Assume I have one, should I report it and SAVE nfs@home of another 6 months of coming WASTING TIME, or continue to keep silent? (I don't have a factor for it, this just for the sake of the flame war that we are already fighting). Assume you LL test an 100M or 1B exponent, a task that takes one (or more) year(s) with the hardware you have, and you are already half through it. Assume some guy, who has NO IDEA about gimps, or math, or factoring programs, but it has a fistful of luck, turns out a factor for your exponent, from the thin air. Should he report it and save you a half year of following futile work, so you can move on to some more useful things, as another exponent? Or should he keep silent and let you waste another half year of resources? Which is more selfish? Now substitute "a guy with no idea" with "LaurV". Or any guy from here around. How does that change the equation?? Then substitute "LL" with "P-1" or "ECM". How does that change the equation?? It does not change. But we humans are very subjective. One guy wanting to go to a movie for which the ticket is 5 dollars, arrive there and finds out that the ticket is doubled in price, it is now 10, he will still buy the ticket and go to the movie, most of the guys. One guy having a ticket already, bought in advance, for 5 dollars, and arriving at the theater finds out that he forget his ticket home, he will not pay 5 dollars to get a new ticket, but it will return home to get it. In both cases the loss would be 10 dollars, for the same movie, but people are very subjective about it. Running TF for exponents assigned to other people for longer jobs is NOT poaching, as long as you do not make this regularly and with a mean purpose. I did this few times in my 8-years GIMPS life, maybe 10 or 20 times, when I was particularly interested in some exponents, and I DO NOT consider myself a poacher. |
[QUOTE=LaurV;282277]Assume you LL test an 100M or 1B exponent, a task that takes one (or more) year(s) with the hardware you have, and you are already half through it. Assume some guy, who has NO IDEA about gimps, or math, or factoring programs, but it has a fistful of luck, turns out a factor for your exponent, from the thin air. Should he report it and save you a half year of following futile work, so you can move on to some more useful things, as another exponent? Or should he keep silent and let you waste another half year of resources? Which is more selfish? [/QUOTE]
A problem that I see with this, there is no way for Joe user to tell Jane user that he found a factor, and that she should stop. |
Which I already mentioned, but without jane and joe, in my first post on this thread. I said (not the same words, but the same idea): "you should write to him, so he can stop any work he eventually started, and move on to the next assignment".
If I would get a mail saying "hey, stop that LL (or P-1), I have a factor for your exponent!", then I would kiss the sender every time (most probably he won't like me kissing him, but that is a different story). He SAVED my time. But people do not want to understand that, and still crumple on that "credit". I like the credit, for making fun here around. Do someone need the credit I got by P-1-ing M2061? Well, tell me how to transfer it to you. Even George asked people to TF his exponents. What is the big fuss? Get it done, move on! No personal attachment. I (this time the assignee) do LL or a long P-1. One wants to waste his time to TF my exponent? Be my guest! You can turn up a factor, so you will SAVE my time, and I can move on to next exponent. You can end up empty handed, then you wasted YOUR time. You should not do that, if you are sane, unless you have particular interest in MY exponent. There is nothing for me (the assignee) to lose, and nothing for you (the poacher) to gain. What is not clear? I (the poacher this time, when I was so) fully understood I would waste my time. I can give you a list with the expos I poached, because I STILL HAVE the same interest in them. |
On the trivia front, I'm about to hit 100 P-1's and 2000 TF runs :smile:
6 P-1 factors and 21 TF factors. EDIT: 22 TF factors. Another one just came in. |
[QUOTE=LaurV;282279]If I would get a mail saying "hey, stop that LL (or P-1), I have a factor for your exponent!", then I would kiss the sender every time (most probably he won't like me kissing him, but that is a different story).[/QUOTE]Assume for a moment that I am LLing a 100M exponent and you TF'ed it to 80 bits and found a factor. [U]You would be unable to contact me.[/U](BTW my PrimeNet ID is not the same as for the forum). [B][I]That[/I][/B] is my point. Most PrimeNet users are unreachable. I don't think that George wants to play match maker. Nor is PrimeNet set to send such notifications. Nor is Prime95 set to receive such a notification.
|
Not that it's very much yet in the grand scheme of things, but just a few minutes ago, I found my second factor ever on a 57M range exponent.
|
[QUOTE=Uncwilly;282281]Assume for a moment that I am LLing a 100M exponent and you TF'ed it to 80 bits and found a factor. [U]You would be unable to contact me.[/U](BTW my PrimeNet ID is not the same as for the forum). [B][I]That[/I][/B] is my point. Most PrimeNet users are unreachable. I don't think that George wants to play match maker. Nor is PrimeNet set to send such notifications. Nor is Prime95 set to receive such a notification.[/QUOTE]
I also addressed this, when I said "what to do if the other user (assignee, thanks cheesehead) is an anonymous, or not a member". (@cheesehead: by "not a member", I was understanding the assignee is not a member on this forum, and you can not contact him. I saw your post with the question, but had no time to reply at that moment). All in all, I will say it again: 1. I agree that people should not poach. 2. Poaching is BAD! 3. I do not poach. 4. I do not consider TF as poaching, unless you do it quite regularly, and with a mean interest, OR unless the other guy (the assignee) is also doing a TF job (that is stealing a TF job, to TF on it, which is even worse then poaching, is stealing!) End of story. It could happen occasionally, if you get a particular interest of one of the exponents. It happens to me (to "poach" other people) about 1-2 times per year. |
[QUOTE=LaurV;282277]
Running TF for exponents assigned to other people for longer jobs is NOT poaching, as long as you do not make this regularly and with a mean purpose. I did this few times in my 8-years GIMPS life, maybe 10 or 20 times, when I was particularly interested in some exponents, and I DO NOT consider myself a poacher.[/QUOTE] In this case you are setting a long LL as the baseline. But no one is accusing you of running a TF on someone else's LL. You ran a TF on someone's P-1, and you ran a massive ECM on an entire project's NFS. So you specifically attempted to factor a number that someone else is attempting to factor. I don't understand how you don't see the problem here. Beyond that, it isn't even okay for you do to it if someone were to be running LL. Think about it. There is some x probability (I've heard 1/70 thrown around) that you do find a factor when you TF someone's LL. Does this person save some time on their LL [B]if[/B] you can somehow notify them that their test is not needed? Absolutely. Do you get whatever it was that you wanted when you set out to TF this already assigned exponent. I suppose so. But GIMPS as a whole is who loses in this case. You have wasted whatever time that person already spent on that LL. There are plenty of other numbers out there to TF. You can find one that has not yet been LL'ed. This is the best allocation of resources for the project. Saving LLx2 is much better than saving LLx1.5 (or whatever), plus there are no hurt feelings when you save LLx2. |
[QUOTE=KyleAskine;282294]In this case you are setting a long LL as the baseline. But no one is accusing you of running a TF on someone else's LL. You ran a TF on someone's P-1, and you ran a massive ECM on an entire project's NFS. So you specifically attempted to factor a number that someone else is attempting to factor. I don't understand how you don't see the problem here.
Beyond that, it isn't even okay for you do to it if someone were to be running LL. Think about it. [/QUOTE] Well it wasn't LaurV that did the TF on someone else P-1 [QUOTE=markr;282119]7.0M to 7.019M is now checked to 2^59, and I'll take it to 2^60. [URL="http://mersenne-aries.sili.net/exponent.php?exponentdetails=7018901"]M7018901[/URL] has a 58 bit factor. It was assigned to Carsten Kossendey for P-1 (since March): I don't usually poach but since P-1 would be very unlikely to find a factor with k = 43 x 258029413 I hope he won't mind too much.[/QUOTE] |
[QUOTE=KyleAskine;282294]
But GIMPS as a whole is who loses in this case. You have wasted whatever time that person already spent on that LL. There are plenty of other numbers out there to TF. You can find one that has not yet been LL'ed. This is the best allocation of resources for the project. Saving LLx2 is much better than saving LLx1.5 (or whatever), plus there are no hurt feelings when you save LLx2.[/QUOTE] Half his post was about that this paragraph is entirely true, and he does such only when he has a particular interest in the exponent involved, rather than just 'vanilla' work for GIMPS. Just pointing that out. |
[QUOTE=James Heinrich;279291]To list those exponents (that I've found so far) in one place for easy reference:[/quote]
2 more: [url=http://mersenne-aries.sili.net/6802123]M6,802,123[/url], [url=http://mersenne-aries.sili.net/6853937]M6,853,937[/url], [url=http://mersenne-aries.sili.net/6853967]M6,853,967[/url], [url=http://mersenne-aries.sili.net/6854297]M6,854,297[/url], [url=http://mersenne-aries.sili.net/6888719]M6,888,719[/url], [url=http://mersenne-aries.sili.net/6935129]M6,935,129[/url], [url=http://mersenne-aries.sili.net/6937501]M6,937,501[/url], [url=http://mersenne-aries.sili.net/6961751]M6,961,751[/url], [url=http://mersenne-aries.sili.net/6984797]M6,984,797[/url], [i][url=http://mersenne-aries.sili.net/7012963]M7,012,963[/url][/i], [b][url=http://mersenne-aries.sili.net/8263357]M8,263,357[/url][/b], [b][url=http://mersenne-aries.sili.net/8272073]M8,272,073[/url][/b], [url=http://mersenne-aries.sili.net/8289409]M8,289,409[/url], [url=http://mersenne-aries.sili.net/8855257]M8,855,257[/url] |
[QUOTE=diamonddave;282296]Well it wasn't LaurV that did the TF on someone else P-1[/QUOTE]
My reading comprehension skills are not up to snuff. So sorry about half of my accusations. The rest of my points still stand though. |
[QUOTE=LaurV;282153]That's life, my dear. I don't want to go through a flame war with you, but if it is a must, I will.[/QUOTE]To whom are you addressing this?
Are you saying that you don't yet understand the ethics of the GIMPS/PrimeNet system? |
[QUOTE=Rodrigo;282205]cheesehead,
I'm simply trying to get a ballpark estimate on how long it should take to complete a P-1. The ones that get assigned to my not-so-modern CPUs prior to starting an LL test take just a couple of days. Therefore I wonder why it would take nine months for someone to complete a P-1, as in the case under discussion.[/quote]Rodrigo, You seem to be assuming that your nine-months case involves a single P-1 assignment, since you compare it with a single-assignment case of your own. Have you ever considered the possibility that someone might get more than one assignment at a time (which is perfectly legitimate)? Only the assignment that is listed first in the worktodo file may be the one actively in progress, but the user may wish to have multiple assignments queued up. M7018901 might well be queued farther down in the user's worktodo. There's nothing about that situation that violates any GIMPS or PrimeNet rules. A reminder: PrimeNet has algorithms to limit how long an assignment may pass without progress. If the user's progress is within those rules -- as indicated by the fact that the assignment still exists -- no one else has any business poaching it. Anyone who thinks there's some exceptional circumstances can ask George Woltman to intervene. |
[QUOTE=Rodrigo;282223]Thanks, James, for the info and the links.
If I read the information correctly, [I]ckdo[/I] has no shortage of [URL="http://www.mersenne.org/report_top_500/"]firepower[/URL] available. Thus, it leads one to wonder why such a tiny assignment[URL="http://www.mersenneforum.org/showpost.php?p=282119&postcount=946"], reportedly[/URL] allocated in March, would have still not been completed well into December (as of the time that this discussion started). There must be a good explanation. Rodrigo[/QUOTE] There's a rather simple explanation: I have a lot of similar assignments - 8,000 or some such. While this may seem like an enormous number, I have already completed 64,000+ of those. I'm rather positive none of the rest will expire, even if the final result is a long way down the road. :smile: |
[QUOTE=cheesehead;282318]Have you ever considered the possibility that someone might get more than one assignment at a time (which is perfectly legitimate)?[/QUOTE]
cheesehead, Until I read [URL="http://www.mersenneforum.org/showpost.php?p=282250&postcount=982"]this post [/URL]by Brian-E which provided an explanation of ckdo's situation after I posed my question, no -- the possibility of stockpiling thousands of assignments for completion months or years down the road had never crossed my mind. Now I have learned that this does happen. I have seven computers doing various kinds of work for GIMPS. Following the KISS principle, they all receive their assignments automatically from the PrimeNet server as current exponents approach completion. The machines doing TF have a 5-day queue. I [U]am[/U] aware that it's possible to increase the size of the queue, but I never had a need to do that or indeed (till yesterday) saw any point in doing so. (The machines I use are at home, so I have at-will access to them.) The only hands-on experience I have with stockpiling assignments in any way is with the GPU I use to LL with CUDALucas. The first time I tried the manual assignment thing, I thought I'd done it wrong and ended up getting two exponents. Now I'm more careful and only do one at a time, on the consideration that should something happen to me, then these assignments will have to wait so long before somebody else can take them up. [QUOTE=cheesehead;282318]Only the assignment that is listed first in the worktodo file may be the one actively in progress, but the user may wish to have multiple assignments queued up. M7018901 might well be queued farther down in the user's worktodo. There's nothing about that situation that violates any GIMPS or PrimeNet rules.[/QUOTE] So let me see if I get this -- you could have 10,000 P-1's (or whatever) lined up, and as long as PrimeNet keeps getting reports of progress made on the batch [B]as a whole[/B] with enough regularity, it doesn't matter that the last ones of those 10,000 won't actually be done for two years, five years, or whatever. Is that correct? Rodrigo |
[QUOTE=Rodrigo;282350]cheesehead,
Until I read [URL="http://www.mersenneforum.org/showpost.php?p=282250&postcount=982"]this post [/URL]by Brian-E which provided an explanation of ckdo's situation after I posed my question,[/QUOTE]Unfortunately, I hadn't yet read that post from Brian-E when I composed my preceding one, or else I wouldn't have posted my last one at all. [quote]no -- the possibility of stockpiling thousands of assignments for completion months or years[/quote]not years AFAIK -- see below [quote]So let me see if I get this -- you could have 10,000 P-1's (or whatever) lined up, and as long as PrimeNet keeps getting reports of progress made on the batch [B]as a whole[/B] with enough regularity, it doesn't matter that the last ones of those 10,000 won't actually be done for two years, five years, or whatever. Is that correct?[/quote]Last I heard, the PrimeNet limit was one year or less. So, not multiple years AFAIK, but I could be mistaken. You wouldn't have 10,000 P-1s assigned unless you had lots of computers, so they wouldn't be in progress just one at a time. |
[QUOTE=cheesehead;282351]
Last I heard, the PrimeNet limit was one year or less. So, not multiple years AFAIK, but I could be mistaken. [/QUOTE] I watch this very close on a small project I'm working on, and I think I have it figured out ... let me know if I'm missing something. There two options, the first is getting an assignment from PrimeNet automatically. This assignment has a life of 60 days from the last update from your computer. If your computer updates everyday, it could conceivably never expire. If it updates every 7 days, the life will reduce by 1 per day for the week, down to 53 days, then get upped again to 60 when you report, until it is finished. This is true for all of your assignments of this type, not just the one that is active (one per core actually). If you don't report back to the server within the 60 days, regardless of your status with the exponent, it gets returned and made available. The other option is the manual reservation. These are good for 180 days. There is no updating that I know of for these, so if you don't finish them within the 180 days, they get returned. Of course, there is always the option to manually extend them as well. Doug |
As for LL-100M, it said "Current assignment won't be finished within a year -- may be unreserved then'. But clearly this doesn't happen to ckdo, but it's also true his work would not be automatically assigned anyways, so potentially that 1 year rule wouldn't apply. Otherwise drh is correct.
|
[QUOTE=ckdo;282324]There's a rather simple explanation: I have a lot of similar assignments - 8,000 or some such. While this may seem like an enormous number, I have already completed 64,000+ of those. I'm rather positive none of the rest will expire, even if the final result is a long way down the road. :smile:[/QUOTE]
ckdo, This is an opportunity for me, cheesehead, and others following this discussion to learn more about the process of obtaining a large number of assignments. In the last several posts, we've been wondering what one needs to do in order to receive 10,000 (or, as in your case, 8,000) P-1's. Is it some sort of manual process? How many months' worth of work can you (anyone) reserve? Can you use the automated Prime95 settings to request that much work? Thanks! Rodrigo |
I could get about 300 exponents by cheating the server to believe I have a very fast computer (30GHz) in the last 5 minutes (now let's see how I get rid of them :smile:).
I put in prime.txt something like" DaysOfWork=60 MaxExponents=10000 UnreserveDays=20000 and in local.txt: OldCpuSpeed=30000 NewCpuSpeedCount=0 NewCpuSpeed=0 CpuSpeed=30000 This would do it.... [edit offtopic] About the "poaching" of M1061, just to clarify the things in my favour :P, my job started long ago, on a very slow computer, which I somehow forgot about, as I already [URL="http://mersenneforum.org/showpost.php?p=279785&postcount=121"]explained there[/URL], and I was a bit surprised too, to see the job done and the credit given. I believe I started it [B]before[/B] nfs@home begun sieving on it, or even poly selection, this just to clarify the things. This takes time, it took over two months, you don't get 122Ghz-days credit for a P-1 overnight job... [end of edit, and end of the discussion about poaching, from my side] |
[QUOTE=Rodrigo;282391]what one needs to do in order to receive 10,000 (or, as in your case, 8,000) P-1's.[/QUOTE]My process is nothing fancy:[list=1][*][url=http://mersenne-aries.sili.net/p1small.php]Get assignments[/url] (from wherever you like)[*][url=http://mersenne-aries.sili.net/balance.php]Balance the assignments[/url] across 4 workers[*]Dump them into Prime95 and let it chat with the server for an hour or so reserving all the assignments[*]After reservation is complete, check for and remove and N/A assignment (probably assigned to someone else)[/list]I currently have about 2000 assignments reserved that way.
|
[QUOTE=cheesehead;282135]As is common with poachery justifications, that leaves out something important.
"Didn't know that extending TF was considered poaching..." -- It wasn't [I]only[/I] extending TF. It was extending TF [I]while someone else had a legitimate assignment[/I] (as you yourself acknowledged in your preceding sentence, but not this one). [I]That overlap with a proper assignment[/I] is what distinguishes poaching from legitimate non-overlapping work. Why did you leave out that overlapping-an-assignment factor and try to pretend that the bare TF extension was what was accused to be poachery?[/QUOTE] These are 7M exponents, someone is just going through either 1) seeing if a bad P-1 had been done in the past or 2) trying to factor as many things as possible. I don't see the problem myself, I used James' website to grab a bunch of P-1 in the 7.1-7.2M range but they are running on an old laptop not connected to the internet, so no assignment, but it spits out a result every 42 minutes, so every couple of days I copy the results to thumbdrive and submit. If I ran those on this machine, it'd probably spit one out every 3-4 minutes lol. If someone "poaches" one of mine due to it not being "assigned" oh well. [QUOTE=ckdo;282324]There's a rather simple explanation: I have a lot of similar assignments - 8,000 or some such. While this may seem like an enormous number, I have already completed 64,000+ of those. I'm rather positive none of the rest will expire, even if the final result is a long way down the road. :smile:[/QUOTE] None will expire as long as your computer is on the net talking to the server every x days and your unreservedays=x is not exceeded. |
[QUOTE=James Heinrich;282407]My process is nothing fancy:[LIST=1][*][URL="http://mersenne-aries.sili.net/p1small.php"]Get assignments[/URL] (from wherever you like)[*][URL="http://mersenne-aries.sili.net/balance.php"]Balance the assignments[/URL] across 4 workers[*]Dump them into Prime95 and let it chat with the server for an hour or so reserving all the assignments[*]After reservation is complete, check for and remove and N/A assignment (probably assigned to someone else)[/LIST]I currently have about 2000 assignments reserved that way.[/QUOTE]
Ahh, so THAT's how it's done! Thank you, James. Very cool. Rodrigo |
Found a nice 131 bit 40 digit factor
M50,232,683 has the factor
3,312,819,927,398,148,201,283,212,927,854,847,645,967 I think that's 3.312 Duodecillion. I put the commas in for davieddy k = 3[SIZE=2][SUP][SIZE=2]2[/SIZE][/SUP] × 487 × 503 × 691 × 45497 × 248161 × 404507 × 4739381[/SIZE] Here it is on mersenne-aires: [URL]http://mersenne-aries.sili.net/exponent.php?exponentdetails=50232683[/URL] |
[QUOTE=Chuck;282854]M50,232,683 has the factor
3,312,819,927,398,148,201,283,212,927,854,847,645,967 I think that's 3.312 Duodecillion. I put the commas in for davieddy k = 3[SIZE=2][SUP][SIZE=2]2[/SIZE][/SUP] × 487 × 503 × 691 × 45497 × 248161 × 404507 × 4739381[/SIZE] Here it is on mersenne-aires: [URL]http://mersenne-aries.sili.net/exponent.php?exponentdetails=50232683[/URL][/QUOTE] Nice one! Longer than the longest one listed on that site! |
[QUOTE=KyleAskine;282859]Nice one! Longer than the longest one listed on that site![/QUOTE]
:office: |
I can get almost [URL="http://www.wolframalpha.com/input/?i=3.1445*24%2F13&t=crmtb01"]6 GHz-Days per day[/URL] on one core of P-1 on an i7-2600k... that's 23 GHz-Days per day across the CPU... o_O (and it was literally 13 hours to within 60 seconds O_o)
|
Self-assignment help
[QUOTE=James Heinrich;282407]My process is nothing fancy:[LIST=1][*][URL="http://mersenne-aries.sili.net/p1small.php"]Get assignments[/URL] (from wherever you like)[*][URL="http://mersenne-aries.sili.net/balance.php"]Balance the assignments[/URL] across 4 workers[*]Dump them into Prime95 and let it chat with the server for an hour or so reserving all the assignments[*]After reservation is complete, check for and remove and N/A assignment (probably assigned to someone else)[/LIST]I currently have about 2000 assignments reserved that way.[/QUOTE]
So I followed your instructions and those here: [URL]http://mersennewiki.org/index.php/Self-assign[/URL] But, it doesn't seem to update the lines at all. If they start out N/A (from GPU to 72) they don't update. I also tried one from the site in your step 1 and that didn't change either. A few questions: Do you add the assignments to worktodo.txt or worktodo.add What would a typical line look like before the communication For your step 3, do you do anything special to get Prime95 started communicating or do you just wait for it to do it itself? Should this work with P-1 assignments from GPU to 72? Thanks for the help. |
[QUOTE=flashjh;283217]So I followed your instructions and those here: [URL]http://mersennewiki.org/index.php/Self-assign[/URL]
But, it doesn't seem to update the lines at all. If they start out N/A (from GPU to 72) they don't update. I also tried one from the site in your step 1 and that didn't change either. A few questions: Do you add the assignments to worktodo.txt or worktodo.add What would a typical line look like before the communication For your step 3, do you do anything special to get Prime95 started communicating or do you just wait for it to do it itself? Should this work with P-1 assignments from GPU to 72? Thanks for the help.[/QUOTE] N/A means Not Assigned (to PrimeNet). These will never update. TF and P-1 GPU to 72 assignments are all N/A. Though I thought assignments from James' site were NOT N/A and would get updated? |
[QUOTE=petrw1;283219]N/A means Not Assigned (to PrimeNet).
These will never update. TF and P-1 GPU to 72 assignments are all N/A. Though I thought assignments from James' site were NOT N/A and would get updated?[/QUOTE] Thanks for the info. You're right, his assignment's are not N/A. I forced a manual communication and it talks to the server but never changes that line. Maybe I'm not waiting long enough or selecting the option? |
[QUOTE=flashjh;283222]Thanks for the info. You're right, his assignment's are not N/A. I forced a manual communication and it talks to the server but never changes that line. Maybe I'm not waiting long enough or selecting the option?[/QUOTE]
Can I ask what line you are expecting to see a change on? Worktodo.txt line should look like: PFactor=<long alphanumeric string>,1,2,<exponent>,-1,72,2 (Unless you selected PMinus in which case it looks a little different but close) |
[QUOTE=petrw1;283226]Can I ask what line you are expecting to see a change on?
Worktodo.txt line should look like: PFactor=<long alphanumeric string>,1,2,<exponent>,-1,72,2 (Unless you selected PMinus in which case it looks a little different but close)[/QUOTE] This is the line I copied. I took a line from James' page so I could practice with the self-assignment. I don't have the <long alphanumeric string> because I thought that is what I was expecting when PrimeNet assigned the number to me (or make it N/A) which would mean it's probably assigned to someone else. Pfactor=1,2,11201131,-1,64,2 |
[QUOTE=flashjh;283227]This is the line I copied. I took a line from James' page so I could practice with the self-assignment. I don't have the <long alphanumeric string> because I thought that is what I was expecting when PrimeNet assigned the number to me (or make it N/A) which would mean it's probably assigned to someone else.
Pfactor=1,2,11201131,-1,64,2[/QUOTE] Assuming that is the actual exponent: per the following it is not assigned to anyone: [url]http://www.mersenne.org/report_exponent/?exp_lo=11201131&exp_hi=10000&B1=Get+status[/url] The line looks correct. When you use the client menu to contact primenet (and I am assuming Windows because I have not used the Unix/Linux version) it will either give you an assignment key and tell you the computed completion date OR it will give you an error that says something like: "Invalid assignment; already assigned". The only idea I have is that when you contact the server are you also checking the line that says: "Send new expected completion dates to the server"? When you click Test... Status does the assignment show up with a completion date? If not (and I am grasping at straws here) the line may not be in the correct file (worktodo.txt) in the Prime95 directory. |
Or try to exit (not just close and it minimize to the traybar) Prime95 before editing worktodo.txt
|
An update.
Earlier I had tried putting it in a .add file, stopped and started and then forced comminication with updated status but it didn't give me an ID. I just checked and after about 2 hours it updated with an assignment ID. So, it worked I just had to wait. |
[QUOTE=flashjh;283245]Earlier I had tried putting it in a .add file[/QUOTE]It takes some time for Prime95 to pull the contents of worktodo.add into worktodo.txt (but with the advantage of not needing to stop/exit/restart Prime95). If you want to make sure your assignment-registration is processed right away, you can stop and exit Prime95, edit worktodo.txt, and then restart Prime95. It will most likely immediately contact the server to register all the unregistered assignments it now finds in worktodo.txt
|
P-1 seems to be stuck at 100%
I recall reading about this a couple years ago in other threads....
I have a P-1 that is sitting at 100% for several minutes. When I check that Status... it reports back the current minutes as the expected completions time....and has been doing so for about 10 minutes now. I am running 2 workers on a Q9550 with 2 cores each. Never mind.....it finished just as I sent this.... :embarassed: Still seemed to be stuck on 100% too long though. |
[QUOTE=James Heinrich;279291]To list those exponents (that I've found so far) in one place for easy reference:[/QUOTE]I've finished the batch of this range that I've been working on, so this should be the final list (10 newest ones in bold) of exponents with P-1 factors that should've been found with the original bounds used, but weren't for some reason:
[url=http://mersenne-aries.sili.net/6802123]M6,802,123[/url], [url=http://mersenne-aries.sili.net/6853937]M6,853,937[/url], [url=http://mersenne-aries.sili.net/6853967]M6,853,967[/url], [url=http://mersenne-aries.sili.net/6854297]M6,854,297[/url], [url=http://mersenne-aries.sili.net/6888719]M6,888,719[/url], [url=http://mersenne-aries.sili.net/6935129]M6,935,129[/url], [url=http://mersenne-aries.sili.net/6937501]M6,937,501[/url], [url=http://mersenne-aries.sili.net/6961751]M6,961,751[/url], [url=http://mersenne-aries.sili.net/6984797]M6,984,797[/url], [i][url=http://mersenne-aries.sili.net/7012963]M7,012,963[/url][/i], [b][url=http://mersenne-aries.sili.net/7113173]M7,113,173[/url][/b], [b][url=http://mersenne-aries.sili.net/7116209]M7,116,209[/url][/b], [b][url=http://mersenne-aries.sili.net/7120891]M7,120,891[/url][/b], [b][url=http://mersenne-aries.sili.net/7124597]M7,124,597[/url][/b], [b][url=http://mersenne-aries.sili.net/7132063]M7,132,063[/url][/b], [b][url=http://mersenne-aries.sili.net/7144853]M7,144,853[/url][/b], [b][url=http://mersenne-aries.sili.net/7145129]M7,145,129[/url][/b], [b][url=http://mersenne-aries.sili.net/7175459]M7,175,459[/url][/b], [b][url=http://mersenne-aries.sili.net/7178657]M7,178,657[/url][/b], [b][url=http://mersenne-aries.sili.net/7183223]M7,183,223[/url][/b], [url=http://mersenne-aries.sili.net/8263357]M8,263,357[/url], [url=http://mersenne-aries.sili.net/8272073]M8,272,073[/url], [url=http://mersenne-aries.sili.net/8289409]M8,289,409[/url], [url=http://mersenne-aries.sili.net/8855257]M8,855,257[/url] |
[QUOTE=petrw1;283523]
I have a P-1 that is sitting at 100% for several minutes. When I check that Status... it reports back the current minutes as the expected completions time....and has been doing so for about 10 minutes now. I am running 2 workers on a Q9550 with 2 cores each. Never mind.....it finished just as I sent this.... :embarassed: Still seemed to be stuck on 100% too long though.[/QUOTE]The GCDs at ends of stage 1 and 2 take several minutes -- might one have been in progress? IIRC the titles and icon statuses never mention the GCD. - - - Low-priority request for prime95/mprime: when the GCD starts, change the "100%" on those lines to "100%, but GCD still in progress" or something like that. |
Mine says "GCD in progress... this may take a few minutes". It doesn't report percentage during that time.
|
[QUOTE=cheesehead;283567]The GCDs at ends of stage 1 and 2 take several minutes -- might one have been in progress?[/QUOTE]
Nope; it spent at least 10 minutes on 100% of Stage 2 ; then the GCD kicked in. |
This bug has never been fixed. It's not reproducible and I've examined the code several times and cannot figure out how it happens.
|
Factor
M52038281 has a factor:
23,156,409,264,727,587,094,789,739,703,316,217 [URL]http://mersenne-aries.sili.net/exponent.php?exponentdetails=52038281[/URL] P-1 found a factor in stage #2, B1=505000, B2=10352500 |
Looks like something is up with that K value on James' website.
|
[QUOTE=KyleAskine;283725]Looks like something is up with that K value on James' website.[/QUOTE]There is, I noticed that just now when [i]flashjh[/i] posted his factor. I need to look at that tomorrow. B1/B2 are accurate though.
|
[QUOTE=petrw1;283604]Nope; it spent at least 10 minutes on 100% of Stage 2 ; then the GCD kicked in.[/QUOTE]
Have you tried telling Prime95 to report the completion percentage to four or even six decimal places? What might be happening is that anything above 99.9xxx% is being reported as 100.00%. Or, if the assignment is within x minutes of the ETA, Prime95 simply rounds off the completion percentage to 100. |
[QUOTE=NBtarheel_33;283914]Have you tried telling Prime95 to report the completion percentage to four or even six decimal places? What might be happening is that anything above 99.9xxx% is being reported as 100.00%. Or, if the assignment is within x minutes of the ETA, Prime95 simply rounds off the completion percentage to 100.[/QUOTE]
IIRC, the percentage is always rounded down. |
[QUOTE=flashjh;283718]M52038281 has a factor:
23,156,409,264,727,587,094,789,739,703,316,217 [URL]http://mersenne-aries.sili.net/exponent.php?exponentdetails=52038281[/URL] P-1 found a factor in stage #2, B1=505000, B2=10352500[/QUOTE] Looking through my stuff on the site, and I found this: [url]http://mersenne-aries.sili.net/exponent.php?exponentdetails=52389509[/url] [url]http://mersenne-aries.sili.net/exponent.php?exponentdetails=48104261[/url] These are my only two results. Huh. For the second, it displays correctly if you go the factor page, rather than the exponent page: [url]http://mersenne-aries.sili.net/exponent.php?factordetails=1574475018803497078755023[/url] And now suddenly the first one works fine from either page. And so does the second. Did I catch this right as you fixed it? |
[QUOTE=KyleAskine;283725]Looks like something is up with that K value on James' website.[/QUOTE]And it's fixed now.
edit: I probably broke whatever [i]Dubslow[/i] was looking at while he was looking at it, and created confusion. :smile: |
[QUOTE=Dubslow;284147]Looking through my stuff on the site, and I found this:
[URL]http://mersenne-aries.sili.net/exponent.php?exponentdetails=52389509[/URL] [URL]http://mersenne-aries.sili.net/exponent.php?exponentdetails=48104261[/URL] These are my only two results. Huh. For the second, it displays correctly if you go the factor page, rather than the exponent page: [URL]http://mersenne-aries.sili.net/exponent.php?factordetails=1574475018803497078755023[/URL] And now suddenly the first one works fine from either page. And so does the second. Did I catch this right as you fixed it?[/QUOTE] How does one read these "Show stats for user" pages? I looked up [URL="http://mersenne-aries.sili.net/stats.php?showuserstats=Luigi+Morelli"]Luigi's[/URL] info, and it says that he's done 8 TFs, 3 P-1's, and no LLs, and then only since December 25 -- whereas in fact we know he's done a lot more than that, for a much longer period of time. Whereas if you look up [URL="http://mersenne-aries.sili.net/stats.php?showuserstats=Dubslow"]Dubslow's[/URL] page, it seems to provide much more realistic-looking (but still incomplete, going back to July) figures and timeframes for exponents processed. |
Each user has to choose to upload the data there. I have never deleted my Prime95 results.txt, and I've uploaded that twice, most recently earlier today. I also upload my mfaktc results to his site -- but have only been doing that for about a month. (Yes, that means I upload it to PrimeNet and then his site before deletion.) There's also another computer I had running P95 for a couple of months over the summer and previous semester, but that has been in a dumpster for a month or so now, so no access to its results.txt. (Edit: Also occurs to me there's a few CUDALucas runs I did 6 months ago that aren't in my Prime95 results. End result is all my LL's should be there, 90%+ of the DC's, and around 60-80% of the P-1. Also maybe 50% of TF.)
TL;DR: Each user chooses to upload data. For me at least, I have some data that I can't upload any more. Luigi must have just started, or only does it sometimes. |
In the last month or so I've started retaining more of the data I spider from PrimeNet. Previously I would catch most of the new primenet data, and use it to update the status of exponents, but not associate any specific work with a specific user, but I have now started to (partially) do that. Any results.txt data that you (as the user in question) upload to my site should override any previously-spidered data (since it's more detailed and accurate). If anyone notices problems (duplicate entries, for example) then please let me know.
|
[QUOTE=James Heinrich;284239]In the last month or so I've started retaining more of the data I spider from PrimeNet. Previously I would catch most of the new primenet data, and use it to update the status of exponents, but not associate any specific work with a specific user, but I have now started to (partially) do that. Any results.txt data that you (as the user in question) upload to my site should override any previously-spidered data (since it's more detailed and accurate). If anyone notices problems (duplicate entries, for example) then please let me know.[/QUOTE]
Thanks for your work on this. |
@Dubslow
@James Heinrich Thanks for filling me in on this. These pages could become a treasure-trove of data. Wonder if it's possible to feed this information to James automatically as PrimeNet receives it, saving users the effort of submitting it by hand and James the effort of pulling it off PrimeNet. I'd speculate that that would require either adding code to Prime95 so that it reports to James also, or reprogramming the PrimeNet server to copy the data to James. I second what flashjh said! Keep up the great work. Rodrigo |
k?
Alright, I've read all over the place and searched online - can't find this info anywhere.
When testing P-1 from the menu of Prime95, if it finds a factor what is the best way to factor the factor to get k like [URL="http://mersenne-aries.sili.net/exponent.php?factordetails=1919565392651028059642113943994258791"]here[/URL]? Thanks, Jerry |
Why not just use that site to get k? The other thing I do is type factor-1 into Wolfram Alpha, which factors it for you. (I'm missing the point, aren't I? What are you trying to say?)
|
[QUOTE=Dubslow;284441]Why not just use that site to get k? The other thing I do is type factor-1 into Wolfram Alpha, which factors it for you. (I'm missing the point, aren't I? What are you trying to say?)[/QUOTE]
You're not missing the point. I was putting in numbers, for example, 900[SUP]200[/SUP]-1 and his site doesn't accept factors for numbers like that (that I could find). Wolfram Alpha did it though... Thanks, Jerry |
[QUOTE=flashjh;284440]When testing P-1 from the menu of Prime95, if it finds a factor what is the best way to factor the factor to get k like [URL="http://mersenne-aries.sili.net/exponent.php?factordetails=1919565392651028059642113943994258791"]here[/URL]?[/QUOTE]"k" as displayed on my site is the prime factorization of (<factor> - 1) / 2 / exponent.
So in your example:[quote]k = (1919565392651028059642113943994258791 - 1) / 2 / 53470619 factor(1919565392651028059642113943994258790) = [color=red]2[/color] × 5 × 29 × 13463 × 17477 × 34327 × 449437 × 34101721 × [color=red]53470619[/color] k = 5 × 29 × 13463 × 17477 × 34327 × 449437 × 34101721[/quote]Subtracting 1 from the found factor is trivial, so you can plug that into anywhere that will give you a prime factorization, for example on [url=http://mersenne-aries.sili.net/factor.php?n=1919565392651028059642113943994258790]my site[/url] (very basic) or [url=http://www.alpertron.com.ar/ECM.HTM]Dario Alpern's ECM factoring program[/url] (far more capable). |
[QUOTE=flashjh;284443]I was putting in numbers, for example, 900[SUP]200[/SUP]-1 and his site doesn't accept factors for numbers like that[code]UID: flashjh/JerryLaptop, 960^119-1 has a factor: 1712794039984598036816946462161826939231808917401926495950795748706050556554081[/code][/QUOTE]No, it doesn't. My site strictly handles Mersenne data, like PrimeNet, with the exception that my site also handles exponents up to M(2^32).
|
[QUOTE=James Heinrich;284480]No, it doesn't. My site strictly handles Mersenne data, like PrimeNet, with the exception that my site also handles exponents up to M(2^32).[/QUOTE]
That's the one I tried to load manually. |
Large Exponent P-1
[QUOTE=cheesehead;257998]Module gwnum.h in the Windows source has the definition
#define MAX_PRIME_SSE2 596000000L /* SSE2 bit limit */ 595999993 is probably the largest prime < 596000000 LL, ECM and P-1 all use FFTs, and 596000000 is the maximum exponent the FFTs can now handle AFAIK. Perhaps the higher exponents were processed with a Mac version of the software rather than a Windows version.[/QUOTE] Is there any way to P-1 very large exponents yet? I would like to P-1 a M999XXXXXX exponent. Jerry |
[QUOTE=flashjh;284847]Is there any way to P-1 very large exponents yet? I would like to P-1 a M999XXXXXX exponent.
Jerry[/QUOTE] The largest implemented FFT size in P95 is 32MB whose limit is as mentioned. A 56M FFT should be able to test your exponent in question. It'd be relatively trivial for George to create it, but not sure if he's ina ny rush to do this :smile: Without a suitable FFT implementation, there's no way to P-1 these exponent(s). How far have you TF-ed them? |
[QUOTE=axn;284850]The largest implemented FFT size in P95 is 32MB whose limit is as mentioned. A 56M FFT should be able to test your exponent in question. It'd be relatively trivial for George to create it, but not sure if he's ina ny rush to do this :smile:
Without a suitable FFT implementation, there's no way to P-1 these exponent(s). How far have you TF-ed them?[/QUOTE] Thanks for the info. The exponent was TFd to 79, not by me. I was looking at some reports and saw a few M999XXXXXX exponents that needed P-1. So I grabbed one to test on my P-1 system only to find that it doesn't work. That's when I went reading and found the current limit for P-1. I searched for an update but didn't find it (until now). Hopefully when v27.1 x64 is final it will allow the bigger FFTs? :wink: |
[QUOTE=flashjh;284854]Thanks for the info. The exponent was TFd to 79, not by me.[/QUOTE]
I think the optimal TF depth is around 81-82 bits by CPU or 84-85 bits by GPU. If you have access to a GPU, I suggest to TF it all the way to 84 bits _at least_, before even contemplating about P-1. |
Seconding axn. In fact, raising: There are only 13 expos TF-ed to 79, and all are 999'96x'xxx or higher. Thinking how long a LL would take for one of that, and thinking we would TF at least 1% of the time worth of it, then there would make no sense to think about P-1, and to ever dream about LL unless they are TF-ed to at least 82-83 bits on CPU, or 86-87 bits on GPU.
|
[QUOTE=LaurV;284863]Seconding axn. In fact, raising: There are only 13 expos TF-ed to 79, and all are 999'96x'xxx or higher. Thinking how long a LL would take for one of that, and thinking we would TF at least 1% of the time worth of it, then there would make no sense to think about P-1, and to ever dream about LL unless they are TF-ed to at least 82-83 bits on CPU, or 86-87 bits on GPU.[/QUOTE]
TF to 87 bits estimated completion via one core on my CPU is 2036-09-12 00:36, not going to happen :smile:. Still waiting on an ETA from mfaktc... |
[QUOTE=flashjh;284867]Still waiting on an ETA from mfaktc...[/QUOTE]
Didn't want to say before doing it, so I just did a 77 to 78 of 999'998'687 which took a bit less then 5 hours using about 45% of one of my GPU's (and got 30 GHz-days for it!). That would be about 95 days to go for 86 to 87 bits, or about 190 days to go all the way from 0 to 87. Well, maybe we could start P-1 a bit earlier:smile:. But anyhow, that is dreaming, and LL is sleeping dead, for these exponents. |
Jaws exponents
[url]http://mersenne-aries.sili.net/exponent.php?exponentdetails=999999131[/url]
245 GHz-days to TF to 80 730 GHz-days for default P-1 for M999999131 27583 days to LL once. "You're gonna need a [faster CPU]" --Police Chief Martin Brody |
[QUOTE=LaurV;284931]Didn't want to say before doing it, so I just did a 77 to 78 of 999'998'687 which took a bit less then 5 hours using about 45% of one of my GPU's (and got 30 GHz-days for it!). That would be about 95 days to go for 86 to 87 bits, or about 190 days to go all the way from 0 to 87. Well, maybe we could start P-1 a bit earlier:smile:. But anyhow, that is dreaming, and LL is sleeping dead, for these exponents.[/QUOTE]
2[sup]86[/sup]-2[sup]87[/sup] would take [url=http://mersenne-aries.sili.net/credit.php?worktype=TF&exponent=999998687&frombits=86&tobits=87]15,671 GHz-days[/url] 2[sup]78[/sup]-2[sup]87[/sup] would take [url=http://mersenne-aries.sili.net/credit.php?worktype=TF&exponent=999998687&frombits=86&tobits=87]31,282 GHz-days[/url] (starting from where it is now) 2[sup]1[/sup]-2[sup]87[/sup] would take [url=http://mersenne-aries.sili.net/credit.php?worktype=TF&exponent=999998687&frombits=1&tobits=87]31,342 GHz-days[/url] (starting from scratch) You can get an approximate idea of how many GHz-days/day your GPU can process [url=http://mersenne-aries.sili.net/mfaktc.php]here[/url]. |
[QUOTE=LaurV;284931]Didn't want to say before doing it, so I just did a 77 to 78 of 999'998'687 which took a bit less then 5 hours using about 45% of one of my GPU's (and got 30 GHz-days for it!). That would be about 95 days to go for 86 to 87 bits, or about 190 days to go all the way from 0 to 87. Well, maybe we could start P-1 a bit earlier:smile:. But anyhow, that is dreaming, and LL is sleeping dead, for these exponents.[/QUOTE]
Has anyone used Moore's law to project when CPUs / GPUs will feasibly be able to LL these exponents? |
[QUOTE=flashjh;284942]Has anyone used Moore's law to project when CPUs / GPUs will feasibly be able to LL these exponents?[/QUOTE]Depends how you define "feasible", but I'd say roughly 15 years to be able to LL M999M on common hardware in approximately 1 month.
|
[QUOTE=James Heinrich;284943]Depends how you define "feasible", but I'd say roughly 15 years to be able to LL M999M on common hardware in approximately 1 month.[/QUOTE]
My god!!! I hadn't thought about it, but you're probably correct! :smile: By that time our systems will be so powerful that Perl (and PHP et al languages) will have modules which can be included to implement the "DWIM" directive.... :cool: (DWIM: Do What I Mean (not what I told you to do).) |
[QUOTE=James Heinrich;284940]
You can get an approximate idea of how many GHz-days/day your GPU can process [url=http://mersenne-aries.sili.net/mfaktc.php]here[/url].[/QUOTE] Not us AMD folk!! |
[QUOTE=James Heinrich;284943]Depends how you define "feasible", but I'd say roughly 15 years to be able to LL M999M on common hardware in approximately 1 month.[/QUOTE]
Is that based on one core or more? Do we have any clue how many cores cpus will have then? Will Prime95 be parallel enough by then to usefully run a single test on 128 cores(maybe possible by then?)? |
[QUOTE=KyleAskine;284953][QUOTE=James Heinrich;284940]You can get an approximate idea of how many GHz-days/day your GPU can process [url=http://mersenne-aries.sili.net/mfaktc.php]here[/url].[/QUOTE]Not us AMD folk!![/QUOTE]All mfakto users are encouraged to email me benchmarks including 4 critical pieces of data [b]running a single instance of mfakto[/b]:
1) GPU model + clockspeed 2) assignment (exponent, startingbits, endingbits) 3) wall time it took to process the assignment 4) average GPU usage Average SievePrimes value, and CPU model/clockspeed are useful as well, but not required. Once I get enough data then AMD users can also benefit from that page. :smile: |
[QUOTE=henryzz;284957]Is that based on one core or more? Do we have any clue how many cores cpus will have then? Will Prime95 be parallel enough by then to usefully run a single test on 128 cores(maybe possible by then?)?[/QUOTE]
I'm not sure it can get any more parallel than it is now, which isn't very much. (Unless that guy who was writing a IBDWFFT for GPU comes back from being afk...) |
Large p-1 factor found
Hi,
I found a large factor for M77999869: p61 = 1815615329758341197381057535338581863482459211614518833203801 123-bit number = p24 195493151840292702688361 * p37 9287360261302658846450143394743421041 factoring of p24-1 = 2^3 * 5 * 43 * 136093 * 10707139 * 77999869 factoring of p37-1 = 2^4 * 3^5 * 5 * 7 * 17 * 283 * 12239 * 14033 * 45377 * 23336543 * 77999869 This is the biggest one I've found so far. |
[QUOTE=timbit;285224]123-bit number[/QUOTE]Nice find.
To clarify, your 61-digit composite factor is 200 bits, but the larger of the two prime factors is still 123-bit. [url]http://mersenne-aries.sili.net/M77999869[/url] |
Does Prime95 report the composite factor, or does it realize that it's composite before submission?
|
[QUOTE=Dubslow;285262]Does Prime95 report the composite factor, or does it realize that it's composite before submission?[/QUOTE]Prime95 finds and reports the composite factor; PrimeNet determines that it's composite upon submission (as does my site).
|
Brent-Suyama extension
I wonder how to calculate the largest possible factor of a Mersenne number that could be found by the P-1 Brent-Suyama extension as it is used in prime95. That is if:
Exponent = M B1 = B1 B2 = B2 Which is the largest possible factor that could be found by the Brent-Suyama extension? Maybe a calculator could be included in [URL]http://mersenne-aries.sili.net/[/URL] |
[QUOTE=aketilander;287115]Which is the largest possible factor that could be found by the Brent-Suyama extension?[/quote]What is the largest factor that could be found without it? Very large factors could be found, if they existed and were improbably-smooth. It's very easy to examine a factor and determine either the needed bounds to find it, or whether given bounds would find the factor, but I don't think there's any particular limit to the size of factors that could be found with any particular bounds (? someone correct me if I'm wrong, please).
[QUOTE=aketilander;287115]Maybe a calculator could be included in [URL]http://mersenne-aries.sili.net/[/URL][/QUOTE]I'd be happy to, if someone can explain the math in very simple words that I could understand. :cmd: ([i]Mr. P-1[/i] had given me [url=http://www.mersenneforum.org/showpost.php?p=280258&postcount=915]a nice explanation of the Brent-Suyama extension[/url], but I don't know what to do with that... :help::cry:) |
[QUOTE=Dubslow;284970]I'm not sure it can get any more parallel than it is now, which isn't very much.[/QUOTE]
:confused: When I run 4 25M DC tests on my i5-750 (which is now 3 year old technology) OC'd to 3200 my per iteration time for each core in just over .020 seconds. When I use 3 cores in parallel for 1 DC the time is just under .007 seconds. The remaining core DC time drops slightly to about .0197. I am getting about 96% effeciency on the 3 cores in parallel and about 102% efficiency in the 4th cores. Seems like pretty decent parallelization to me. It still quite good with all 4 cores on 1 DC; the per iteration time drops to about .0056 which is close to 90% efficiency. |
| All times are UTC. The time now is 10:25. |
Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.