Newer milestone thread - Page 94

retina · 2014-01-17, 16:57

Quote:

Originally Posted by chalsall

As an example, I plan to "poach" 45456781. Assigned to one of "curtisc's" machines for 1033 days; clearly stuck on P-1 Stage Two. I can complete it on one-half of one of my machines in about four days.

Does anyone consider this immoral? (And, if it turns out to be Prime, I'd be happy to let him add it to his collection.)

I won't comment on the morality of it. But why not just inform curtisc of the problem and see what he wants to do about it.

chalsall · 2014-01-17, 17:00

Quote:

Originally Posted by retina

I won't comment on the morality of it. But why not just inform curtisc of the problem and see what he wants to do about it.

Excellent suggestion. I will do so.

Although I wonder why he wasn't already aware of this himself....

c10ck3r · 2014-01-17, 17:47

Quote:

Originally Posted by chalsall

As an example, I plan to "poach" 45456781. Assigned to one of "curtisc's" machines for 1033 days; clearly stuck on P-1 Stage Two. I can complete it on one-half of one of my machines in about four days.

Does anyone consider this immoral? (And, if it turns out to be Prime, I'd be happy to let him add it to his collection.)

Considering the fact that it hasn't been actually LL'd (only Stage 1 of P-1 completed), I can't fathom calling this poaching in practice. In theory, maybe, but in practice, you haven't cost him any resources, except once it is complete if the machine continues to hang up on P-1 and isn't found.

petrw1 · 2014-01-17, 17:55

Quote:

Originally Posted by chalsall

However, George / Scott never actually implemented the removal process. The client will drop an assignment (even if has had work done on it already) if it receives the appropriate message from the server, and then get a new assignment which is actually useful.

I think it is at least partially implemented.

I have on several ocassions lost assignments on day 61 of no progress. I have a few borged PCs that are hard to get to or that work sporadically....great guns for a couple months then nothing for a few only to return. :/

chalsall · 2014-01-17, 18:04

Quote:

Originally Posted by petrw1

I have on several ocassions lost assignments on day 61 of no progress. I have a few borged PCs that are hard to get to or that work sporadically....great guns for a couple months then nothing for a few only to return. :/

That's a slightly different thing, but thanks for bringing it up...

If a client doesn't report for 60 days, then Primenet will "recycle" the candidate and assign it to someone else. The next time your client checks in it will be told (in the meta) "You don't have a valid AID" and the client will drop the assignment and then request another.

The issue being discussed here is that Primenet should report back to the client the same message if the assignment is over a year old (or, arguably, if the claimed "Estimated Completion Date" - "Assignment Date" is greater than one year).

But it doesn't...

Yet.

ATH · 2014-01-17, 18:31

Quote:

Originally Posted by chalsall

As an example, I plan to "poach" 45456781. Assigned to one of "curtisc's" machines for 1033 days; clearly stuck on P-1 Stage Two. I can complete it on one-half of one of my machines in about four days.

I already started that one a few days ago. I can't do it in 4 days, as it is an older machine running 2 LL's at once, but in like 14 days or so.

chalsall · 2014-01-17, 18:36

Quote:

Originally Posted by ATH

I already started that one a few days ago. I can't do it in 4 days, as it is an older machine running 2 LL's at once, but in like 14 days or so.

LOL...

OK, I hadn't yet placed it on a machine, so I'll leave it to you to "poach".

TheMawn · 2014-01-17, 19:37

Just to pitch in here I also agree that "poaching" near the bottom of the pile shouldn't be a huge deal.

I believe that one year is a very, VERY generous amount of time to do any one assignment. If your device takes one year to give a result, it is contributing virtually nothing to the project and is very likely going to have encountered at least one error in the quadrillions of cycles it spent.

I think even six months (I might push three) is too damn long for an LL test. I feel bad about doing TF on a CPU and my laptop, even at 24/7, would take weeks to do a single test in the DC range and I find even that to be too long, so I stick with P-1 as the only other useful work it can do.

Realistically, if your computer has too little RAM to do P-1 and the CPU is too slow to do an LL in less than three months, it should probably be in a heap somewhere.

If three or six months is too pushy, then at least consider removing assignment extensions on any assignment in demand. I don't care if a 70M first-time LL is going to take four months because no relevant milestone involving that assignment could possibly be met by then. On the other hand, some two-year-and-still-going assignment that at least 20 of us would gladly take and do in two days should definitely be recycled.

ATH · 2014-01-17, 21:04

Quote:

Originally Posted by TheMawn

I believe that one year is a very, VERY generous amount of time to do any one assignment. If your device takes one year to give a result, it is contributing virtually nothing to the project and is very likely going to have encountered at least one error in the quadrillions of cycles it spent.

I always quote Prime95 own idea for the assignment rules: http://www.mersenneforum.org/showpos...&postcount=443

I make sure the assignments I "poach" are well beyond those limits, and I would never poach an exponent with progress above say 70% even if it has been active for a long time, and has a long time to go.

chalsall · 2014-01-17, 21:14

Quote:

Originally Posted by ATH

I always quote Prime95 own idea for the assignment rules: http://www.mersenneforum.org/showpos...&postcount=443

Interestingly, this was written more than three years ago....

petrw1 · 2014-01-17, 22:58

This proposal excludes manual assignments but even they need some limit don't they?

Case in point: sorry it formatted poorly.

Exponent
Work
Type

Stage, %
age
days

days
to go

Estimated
Completion Next
Update Updated Assigned Userid CPU Name
28525907 D 1510 2014-02-26 2013-12-21 2009-11-29 CaptainEntropy Manual testing
28526873 D 1510 2014-02-26 2013-12-21 2009-11-29 CaptainEntropy Manual testing
28527307 D 1510 2014-02-26 2013-12-21 2009-11-29 CaptainEntropy Manual testing
28530247 D 1453 2014-02-23 2013-12-21 2010-01-25 CaptainEntropy Manual testing
28530767 D 1142 2014-03-06 2013-12-21 2010-12-02 CaptainEntropy Manual testing
28530937 D 1142 2014-03-06 2013-12-21 2010-12-02 CaptainEntropy Manual testing

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Newer X64 build needed	Googulator	Msieve	73	2020-08-30 07:47
Performance of cuda-ecm on newer hardware?	fivemack	GMP-ECM	14	2015-02-12 20:10
Cause this don't belong in the milestone thread	bcp19	Data	30	2012-09-08 15:09
Newer msieves are slow on Core i7	mklasson	Msieve	9	2009-02-18 12:58
Use of large memory pages possible with newer linux kernels	Dresdenboy	Software	3	2003-12-08 14:47

2014-01-17, 19:37	#1031
TheMawn May 2013 East. Always East. 11×157 Posts	Just to pitch in here I also agree that "poaching" near the bottom of the pile shouldn't be a huge deal. I believe that one year is a very, VERY generous amount of time to do any one assignment. If your device takes one year to give a result, it is contributing virtually nothing to the project and is very likely going to have encountered at least one error in the quadrillions of cycles it spent. I think even six months (I might push three) is too damn long for an LL test. I feel bad about doing TF on a CPU and my laptop, even at 24/7, would take weeks to do a single test in the DC range and I find even that to be too long, so I stick with P-1 as the only other useful work it can do. Realistically, if your computer has too little RAM to do P-1 and the CPU is too slow to do an LL in less than three months, it should probably be in a heap somewhere. If three or six months is too pushy, then at least consider removing assignment extensions on any assignment in demand. I don't care if a 70M first-time LL is going to take four months because no relevant milestone involving that assignment could possibly be met by then. On the other hand, some two-year-and-still-going assignment that at least 20 of us would gladly take and do in two days should definitely be recycled.

2014-01-17, 22:58	#1034
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 4695₁₀ Posts	This proposal excludes manual assignments but even they need some limit don't they? Case in point: sorry it formatted poorly. Exponent Work Type Stage, % age days days to go Estimated Completion Next Update Updated Assigned Userid CPU Name 28525907 D 1510 2014-02-26 2013-12-21 2009-11-29 CaptainEntropy Manual testing 28526873 D 1510 2014-02-26 2013-12-21 2009-11-29 CaptainEntropy Manual testing 28527307 D 1510 2014-02-26 2013-12-21 2009-11-29 CaptainEntropy Manual testing 28530247 D 1453 2014-02-23 2013-12-21 2010-01-25 CaptainEntropy Manual testing 28530767 D 1142 2014-03-06 2013-12-21 2010-12-02 CaptainEntropy Manual testing 28530937 D 1142 2014-03-06 2013-12-21 2010-12-02 CaptainEntropy Manual testing