mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   PrimeNet (https://www.mersenneforum.org/forumdisplay.php?f=11)
-   -   Assignments? (https://www.mersenneforum.org/showthread.php?t=2690)

ET_ 2004-07-02 17:03

[QUOTE=S00113]
I think we should take another approach to handing out work. [/QUOTE]

You assume that everyone joining the project has read the user manual and is able to select number of hours and kind of reservation.

I don't think so. Many people use to start an application just clicking on it, and to forget about it. They usually put default values in the boxes just to see "if it works".

To get rid of long-standing exponents, I think there should be an automated algorithm to calculate the effective CPU daily work, and request exponents based upon that algorithm.

Luigi

I_like_tomatoes 2004-07-02 18:45

[QUOTE=S00113]And let the server deny requests for exponents which will take more than 120 days or so to complete.[/QUOTE]
What about people with slower computers that want to look for primes that are larger then the ones yet discovered?
Sure, it'll take forever for somebody with a 386 to do an LL test on a number bigger then M41, but if it isn't slowing somebody else's progress down, we shouldn't stop them.
If you're going to deny the longer requests, at the very least let anybody who wants tackle the longer exponents (although perhaps give them a warning that it'll take a while if it's going to take more then six months or so). They won't be slowing down the countdowns - in fact, it would speed up the countdowns by letting someone with a faster computer have a go.

patrik 2004-07-02 19:03

Yes, assigning first time tests to the slowest computers may be a good idea since it would prevent them from holding up any milestones. (The "tested once" milestones are really not very important since we have so many errors.) And assigning first time tests to the REALLY slowest computers may prevent them from holding up any milestones for the first four years, or so.

Matthias C. Noc 2004-07-02 19:05

[QUOTE=S00113]It is very discouraging when you find that the exponent you have been working on the last four months was poached...[/QUOTE]

Four month would be no problem, I was talking about those who need years or how nfortino wrote "only one has a chance of finishing in a semi-finite amount of time".

[QUOTE=S00113] think we should take another approach to handing out work. Let us decide how large exponents we want by the number of days it will take to complete it, and hand out exponents in lots of different ranges at the same time. This will let everyone make steady progress and see exponents complete regularly at the speed he or she decide. And let the server deny requests for exponents which will take more than 120 days or so to complete.[/QUOTE]

Yes, something like that would be good. I think a stricter expiration rule might do the trick too, e.g. if procress per week on a small exponent is not bigger than 5 % (except holidays), then a warning should be send, after three warnings the exponent should expire and returned to the pool.

[QUOTE=I_like_tomatoes]What about people with slower computers that want to look for primes that are larger then the ones yet discovered?
Sure, it'll take forever for somebody with a 386 to do an LL test on a number bigger then M41, but if it isn't slowing somebody else's progress down, we shouldn't stop them.
If you're going to deny the longer requests, at the very least let anybody who wants tackle the longer exponents (although perhaps give them a warning that it'll take a while if it's going to take more then six months or so). They won't be slowing down the countdowns - in fact, it would speed up the countdowns by letting someone with a faster computer have a go....[/QUOTE]

If one isn't holding up the progress I see no problem at all. In the higher ranges one year or more to complete won't matter in the foreseeable future. So there is no problem.

Best,

Matthias C. Noch

ET_ 2004-07-02 20:14

[QUOTE=Matthias C. Noc]
If one isn't holding up the progress I see no problem at all. In the higher ranges one year or more to complete won't matter in the foreseeable future. So there is no problem. [/QUOTE]

Apart from discovering M42 with one year of delay...

Luigi

Matthias C. Noc 2004-07-02 21:11

[QUOTE=ET_]Apart from discovering M42 with one year of delay...
[/QUOTE]

Yes, maybe, but it will take a lot of time to prove that M42 is really M42 and not M44. IMHO the real worth of this project is that we are checking all exponents, and are not just adding some new, but isolated M-primes.

Best,

Matthias C. Noch

Primehack 2004-07-02 22:22

I think you are forgetting something very important.

This is a proyect we are all contributing as a TEAM, and I think we need everybody contribution. The sum of all contributions is what gives this proyect its impressive computing power...

If you think your supercomputers (which btw I really don't think you have) can do this work alone then please don't bother everybody else: Become LMHs and enjoy your superb speed.

Best Regards,
EMC

cheesehead 2004-07-06 10:47

[QUOTE=Matthias C. Noc]Hi,

you asked what harm slow-completer are causing. I have three points:[/quote]
(... each of which explains how impatient people can affect the project, but none of which demonstrates any effect on the project by the slow-completers themselves, which is what I asked for! :-)
[quote]1.) The project is not only about <..snip..> If it takes years to finish one small exponent, which under today’s normal performance would take only 4 of 5 days to calculate, it is holding up the whole process without need.[/quote]
No, it isn't!!

Let's say the fast machine does the small exponent in 4 days, while the slow machine does it in 400 days. Then their relative speeds are 100 and 1. With the fast machine alone, total speed of GIMPS progress is 100 speed units. With both machines, total speed of GIMPS progress is 101 speed units, which is 1% faster. The slow machine is speeding up GIMPS, not slowing it down.

If the slow machine were not taking care of exponent A it is working on, the fast machine [u]would have to be diverted from working on a different exponent B in order to take care of exponent A[/u]. [u]That[/u] (i.e., elimination of the slow machine's contribution) would hold up the project for the time it would take the fast machine to work on exponent A instead of exponent B! It is the elimination of the slow machine, not its participation, that would hold up the project!

Who would advance the work on exponent B while the fast machine worked on exponent A? Do you ever take [u]that[/u] into account?

It seems to me that the objectors to the slow-completers systematically overlook that if the slow-completers were not participating, faster machines [u]would have to be diverted from other work[/u] in order to perform the work that the slow machines would have performed, thus [u]slowing down[/u] progress on the exponents that the faster machines would otherwise have been doing.
[quote]That doesn’t help to get a better understanding of the distribution of those Mersenne primes.[/quote]
Tell us how slowing GIMPS progress from a speed of 101 to a speed of 100 helps to get a better understanding of the distribution of Mersenne primes than continuing at speed 101 would get.
[quote]2.) It makes a bad impression about the organisation of the project when we can’t finish a data range in an acceptable amount of time.[/quote]
What is that acceptable amount of time? Who defines the acceptable amount? Will you admit that your argument is only about the emotions of impatient people, not about any real harm done by the slow-completers?
[quote]We are crunching very large numbers, but fail to finish much smaller exponents. That doesn’t look good.[/quote]
But we haven't failed to finish any much smaller exponent!! If you disagree, tell us one exponent which this project has "failed to finish". [i]Note that that can't include any exponent on which work is currently in progress, or else you're declaring that all of the thousands of assignments which are not yet completed are ones that we are "fail[ing] to finish". If you're saying that all the thousands of currently uncompleted assignments are failures to finish exponents, then your argument doesn't seem to be worth anything.[/i]
[quote]3.) We are doing a very abstract stuff here. To keep people motivated one needs to reach small and big milestones from time to time. As we don’t find a new Mersenne every month other milestones have to keep people interested.[/quote]
Okay. For the purposes of this discussion, I'll accept that.

But ever since this project's beginning, we [u]have[/u] been reaching small and big milestones all along, despite the participation of those darned slow-completers. In fact, the slow-completers have been helping GIMPS reach those milestones 1% or 2% faster than we would have without their help.
[quote]Countdown to testing all exponents below M(13466917) once: 4
Countdown to testing all exponents below M(20996011) once: 9,549
Countdown to testing all exponents below M(24036583) once: 42,739
Countdown to proving M(13466917) is the 39th Mersenne Prime: 39,154
Countdown to proving M(20996011) is the 40th Mersenne Prime: 204,395
Countdown to proving M(24036583) is the 41st Mersenne Prime: 271,351[/quote]
... and GIMPS will reach each of those milestones, just as it has been reaching earlier milestones all along.
[quote]If we loose one normal cruncher with an up-to-date PC due to frustration about no progress[/quote]
[i]No[/i] progress? What do you mean [i]no[/i] progress? GIMPS is making progress every day, so why do you write about "no progress"?

... and I note again that your argument is about the feelings of impatient people, not about any real harm done by slow completers. I'm not saying feelings are unimportant -- I'm saying that although I asked for arguments about real harm by slow completers, so far all I've gotten are arguments about feelings of impatient people. I just want folks to call a spade a spade.
[quote]we are loosing much more power then when we loose 10 angry slow-crunchers because their exponents were reassigned in time to keep the project rolling.[/quote]
But the project [u]is[/u] rolling. I haven't yet seen you offer any evidence that it isn't!!
[quote]It will close to be impossible to keep the project running in the much higher areas with only slow-completers.[/quote]
Why do you exaggerate? No one except you has said anything about running the project with only slow-completers.

Can't you tolerate the participation of a small percentage of slow-completers? Neither I nor anyone else has proposed that slow-completers would constitute a large proportion of participants -- it just ain't gonna happen. So what's your worry?
[quote]We need as much high-power as we can get now that we really working big numbers.[/quote]
I agree.

So why don't the folks with high power simply just continue to participate as they have been?

Do you want to lose 1-2% (or whatever it is) of GIMPS speed by eliminating the slow-completers?

Primehack 2004-07-06 21:02

Cheesehead, you have stated in a far more clever way my point of being a team.

I completely agree with you.
EMC

dsouza123 2004-07-07 19:06

Back to the main issue, what is the projected time for the last 4 to finish ?

Years or months ? Or an aproximate date.

JuanTutors 2004-07-07 19:06

I think it's fair to say that, had it not been for the slow computers, we would not know that M24036583 is prime. Personally, I don't see any way for us to have gotten to test M20996011 either if it weren't for those "slow computers".

Plus, if you have 1000 slow computers running at the same rate as 100 new, fast computers, then those 1000 slow produce the same quantity of results in a given time as the fast 100. Since the slow 1000 are also not working synchronously, they are producing results just as if they were the fast 100, and really the only difference between the slow 1000 and fast 100 is the quantity of numbers that are being worked on. Were it not for the fact that this program's creators have allowed us to have access to information about specific numbers, it would be completely impossible to tell the difference between numbers done by slow and fast computers. (In other words, other computer speeds are irrelevant if you don't know their speeds, making their speed irrelevant even if their speed is known.)


Plus, if you need one supreme, nonmathematical reason, let it be that you joined a project that its creators wanted everyone to be able to participate and help in, independent of how much they could do in any given amount of time.


All times are UTC. The time now is 12:21.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.