mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   PrimeNet (https://www.mersenneforum.org/forumdisplay.php?f=11)
-   -   Useless DC assignment (https://www.mersenneforum.org/showthread.php?t=12397)

lycorn 2009-09-06 19:58

Useless DC assignment
 
I started testing (DC) yesterday the exponent 21759931.
Today I checked its status on the server and I found that it had already been successfully DCed. It appears as assigned to me, though.
Now the funny bit:
There is one result submitted by one user, with no date.
And a matching result submitted by a second user on the 5th of September, 2009.
The exponent was assigned to me on the 17th of June 2009, included in a bunch that I have been testing.
The result with no date must have been the 1st time LL test, done before v5 went online (that´s why there is no date). This means that either the exponent was assigned to me while it was being tested by the second user, or it was assigned to him while it was being tested by me. It was a goood thing that I checked the status, otherwise the work would be lost. Now this is not the kind of thing I (and I reckon the majority of the users) do on a regular basis. Normally I am given an exponent to test and don´t care to check its status on the server.
There is, I think, a third possibility: the second user let the exponent expire, and submitted the result at a later date.
I think this should be looked into. If someone lets an exponent expire, and said exponent is reassigned to someone else, then the original assignee should no longer be allowed to submit the result, because it´s frustrating for the new assignee to complete the test and receive the "exponent not needed" message, and no credit for the work.
Any thoughts?

S45653 2009-09-07 04:30

Seems I've been doing a triple-check too: see status of exponent 21563923. It has been LL-tested by three different users, first entry without a date, all residues match. I did get credit, though. Some 17 GHz-days, that probably would have been better spent working on another exponent.

lycorn 2009-09-07 07:15

Absolutely.
The main point of my post wasn´t about credit, but on the waste of resources. Additionaly, the associated feeling of uselessness wrecks your motivation.

garo 2009-09-07 10:18

lycorn, the third possibility is what most likely happened in your case. Unfortunately your suggestion of what to do in this case conflicts with the GIMPS philosophy of never turning down a completed result. Your result should have been accepted and you should have been given credit since you were a legitimate assignee. However, I don't see how the waste of resources can be avoided unless GIMPS can somehow tell your prime95 client to stop working on the exponent as soon as someone else returns a DC on your assignment. Even then, some people might not be pleased to have say a 95% complete assignment be kicked out and them being deprived of the credit. No easy solution here. Also, occasional triple-checks are useful for GIMPS as they provide error rate data that can be useful.

Mini-Geek 2009-09-07 11:24

[quote=garo;188890]lycorn, the third possibility is what most likely happened in your case. Unfortunately your suggestion of what to do in this case conflicts with the GIMPS philosophy of never turning down a completed result. Your result should have been accepted and you should have been given credit since you were a legitimate assignee. However, I don't see how the waste of resources can be avoided unless GIMPS can somehow tell your prime95 client to stop working on the exponent as soon as someone else returns a DC on your assignment. Even then, some people might not be pleased to have say a 95% complete assignment be kicked out and them being deprived of the credit. No easy solution here. Also, occasional triple-checks are useful for GIMPS as they provide error rate data that can be useful.[/quote]
I've got an idea: when a matching DC is returned on a candidate that is already assigned, set a flag on that assignment so that the next time Prime95 communicates the status on that assignment, PrimeNet tells it to stop and report its percent completion, then give partial credit based on that percentage. (e.g. if you're at 90%, and the credit would be 14, give 14*.9=12.6) (or, if it's reporting that it's complete, check that the reported residue matches the other two residues and give full credit, or whatever it normally does if you report a result that ends up not matching)

garo 2009-09-07 11:42

Good idea.

Brian-E 2009-09-07 12:01

I question if the work is useless.

All smaller exponents (up to 1M at least I think) for which no factors have been found, have been LL-tested 3 times.

Presumably the intention is for that to be done further? In the long run the exponents which are now being double-checked will be going through a phase of triple-checks and your exponent will have that work already done.

Mini-Geek 2009-09-07 12:41

[quote=Brian-E;188894]I question if the work is useless.

All smaller exponents (up to 1M at least I think) for which no factors have been found, have been LL-tested 3 times.

Presumably the intention is for that to be done further? In the long run the exponents which are now being double-checked will be going through a phase of triple-checks and your exponent will have that work already done.[/quote]
I'm 99% sure that triple checks are very close to being totally useless. (unless for some other purpose...definitely not useful in the same way that double checks are useful) I think the odds of Prime95 having a bug that would produce matching residues (which is very small) is better than the chance that two 64-bit residues just happened to match when they're both wrong, so a third check wouldn't be useful.
The extra LL results (triple-, quadruple-, etc. checks) in small exponents is likely due to people playing around. (one time, I went through a list of small-ish primes and re-discovered the first couple dozen or so Mersenne primes, just for fun :smile: if I had submitted my results to PrimeNet, they'd appear as more matching LL results on those small numbers)

Brian-E 2009-09-07 13:07

[quote=Mini-Geek;188898]I'm 99% sure that triple checks are very close to being totally useless. (unless for some other purpose...definitely not useful in the same way that double checks are useful) I think the odds of Prime95 having a bug that would produce matching residues (which is very small) is better than the chance that two 64-bit residues just happened to match when they're both wrong, so a third check wouldn't be useful.
The extra LL results (triple-, quadruple-, etc. checks) in small exponents is likely due to people playing around. (one time, I went through a list of small-ish primes and re-discovered the first couple dozen or so Mersenne primes, just for fun :smile: if I had submitted my results to PrimeNet, they'd appear as more matching LL results on those small numbers)[/quote]
Yes I agree with you really. An example of some other purpose was hinted by garo:

[quote=garo;188890](...)
Also, occasional triple-checks are useful for GIMPS as they provide error rate data that can be useful.[/quote]
but for purposes of purely being certain that no other Mersenne numbers in the given range are prime, the only serious use of triple checking would seem to be with completely different client software to allow for the possibility of a bug as you mention, and that is largely handled by the random offset at the start of a LL test, we would hope.

Can anyone tell us why so much of the lower end of the exponents have been triple checked? Is it just people having fun, is it for the error rate data, or are there other concrete reasons too?

lycorn 2009-09-07 13:09

I think Mini-Geek´s suggestion is definitely worth to be considered.
Another alternative could be, when a exponent expires, to warn the assignee that his/er assignment is no longer valid, and as such will not be accepted by the server in the future, and suggesting that the work should be interrupted and another work unit be requested from the server.
I am of the opinion that these cases shall be dealt with in a manner that "penalizes" the original assignee that let the exponent expire, as he/she has already have the chance to do the task, and not the new assignee that has done his lot and is delivering a finished test to the server in due time.
There is no perfect solution, though. I acknowledge that fact.

markr 2009-09-07 13:35

[QUOTE=Brian-E;188902]Can anyone tell us why so much of the lower end of the exponents have been triple checked? Is it just people having fun, is it for the error rate data, or are there other concrete reasons too?[/QUOTE]
Most of them have only two [I]64-bit[/I] residues. Early tests produced shorter residues...

Exponent,User name,Computer name,Residue,Date found
50087,Unknown,,6421,
50087,George Woltman,,7F0C06BD9A0B6421,
50087,Brian J. Beesley,cnsj,7F0C06BD9A0B6421,


All times are UTC. The time now is 21:08.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.