results not needed - Page 4

kladner · 2015-10-16, 15:44

I just reported a P-1 result for M77170517.

Quote:

Sending result to server: UID: kladner/pod64, M77170517 completed P-1, B1=655000, B2=13100000, E=12, We4: CC5D4ADB, AID:

However, someone got there first:

Quote:

PrimeNet error 40: No assignment P-1 result for M77170517 was not needed

Quote:

2015-10-16 josibo NF-PM1 B1=705000, B2=13571250

snme2pm1 · 2015-10-16, 22:38

It is of low significance, however there are unfriendly occasions when results are submitted which prematurely expire outstanding assignments.
http://www.mersenne.org/report_expon...exp_hi=&full=1
That exponent had time remaining for lopper to complete an ECM assignment.
I note that ramgeis lodged a TF factor found result before such expiry.
I am aware of this because I held the same TF factor found result for several weeks, but very deliberately abstained from submitting such result until the official primenet expiry realisation.
I was doubtful that lopper would return a result before expiry, but that's not the point.

LaurV · 2015-10-17, 04:10

Quote:

Originally Posted by kladner

I just reported a P-1 result for M77170517.
However, someone got there first:

He just has higher limits than you have. If you did not delete your checkpoint files, raise B2 and continue. For less than a hour work you can top the reported limits, and don't lose the credit (and may be lucky and find a factor, too)

LaurV · 2015-10-17, 04:20

Quote:

Originally Posted by snme2pm1

I am aware of this because I held the same TF factor found result for several weeks, but very deliberately abstained from submitting such result until the official primenet expiry realisation.

You have a factor? Send the factor!
The worst can happen is that the assignee does not check, and he continues to waste resources until the assigned work is done, but he would do that anyhow, if you don't send the factor.
The best can happen is either the assignee or his program (P95 checks that) will see that the work is not needed anymore, and switch to other work, therefore speeding the tings up. Some credit would be anyhow lost, either his, if he already started any work on that specific assignment, or yours, if you fond a factor and did not report anything. Our elders have a say: better than my mother crying, is your mother crying...

In case you have a factor and "hold it", a lot of worse things can happen, like you lose the factor, or miss the right time to report it and the work is reassigned for LL or even DC, etc, lots of other people waste their resources. My advice: send the factor.

Madpoo · 2015-10-17, 06:45

Quote:

Originally Posted by LaurV

You have a factor? Send the factor!
The worst can happen is that the assignee does not check, and he continues to waste resources until the assigned work is done, but he would do that anyhow, if you don't send the factor.
The best can happen is either the assignee or his program (P95 checks that) will see that the work is not needed anymore, and switch to other work, therefore speeding the tings up. Some credit would be anyhow lost, either his, if he already started any work on that specific assignment, or yours, if you fond a factor and did not report anything. Our elders have a say: better than my mother crying, is your mother crying...

In case you have a factor and "hold it", a lot of worse things can happen, like you lose the factor, or miss the right time to report it and the work is reassigned for LL or even DC, etc, lots of other people waste their resources. My advice: send the factor.

Ditto that... I'd go ahead and check in the factor.

Like you said, worst case they report it in and maybe it gets marked as suspect in which case there's a good chance it would be reassigned almost right away, leaving a very narrow window of time to check in the factor and prevent someone getting the DC and starting on it.

LaurV · 2015-10-17, 06:45

It doesn't really fit in this category, but I didn't want to make a new thread for it, I think something odd is going on with the server, which may be connected or not with the "results not needed" story.

I use one of my computer to reserve work, P95 is not crunching, but only stays open and reserves P-1 work. If you "pause" the workers, it will still connect daily to the server and report results/get assignments, etc. I did this in the past, and it was working well.

Another computer (which has GPU) does the work. Now, it just reported two P-1 results, which of course, were assigned by PrimeNet (yes, they are not in the bunch wrongly assigned by GPU72, but were directly requested as P-1, and assigned by PrimeNet, together with other, total 38 P-1 jobs).

One reported result went well, but for the other one, things got weird. I just found now that instead of being marked as completed, the job was expired, and it was considered as "poached". For which reason, it was not deleted from the worktodo (technically, I still can "finish" the job, and report the result, which will become "not needed"). I deleted it manually from the worktodo (the other one was deleted automatically)

Note that the job is done on the GPU, on another computer, which reported "manually" the result (by a script), but the reservation is done by this computer which I use now, and I only use it to reserve work. The things are simple: it tries to connect to the server, and reserves work, because the worktodo is getting low, the job is done (by the other computer's GPU, which reads from worktodo file), then, when this computer connects daily to the server, it finds that the key is not valid, and the assignment is deleted from the worktodo. Because of that, worktodo shrinks, and when it is too less work to do, other work is brought in. Which in turn, is read by, and done, in the other computer (is GPU work).

This worked well in the past. I just re-started last week, and now it got the keys scrambled somehow. I still have to get access to the other computer (in few hours) and see if it used the right key when it reported the results, before telling for sure that is a server problem, but most probably yes, it is a server problem, and the right keys were used, otherwise the other exponent would not be "completed", but "poached by myself", too.

(edit: now I got totally crazy, not only replying to my own posts, 3 time in a row, but also poaching my own assignments

)

(edit 2: luckily, Madpoo's crosspost saved me this time, haha)

snme2pm1 · 2015-10-17, 07:58

Quote:

Originally Posted by Madpoo

Ditto that... I'd go ahead and check in the factor.

Like you said, worst case they report it in and maybe it gets marked as suspect in which case there's a good chance it would be reassigned almost right away, leaving a very narrow window of time to check in the factor and prevent someone getting the DC and starting on it.

Factor results simply would not be reasonably marked as suspect; sever-side they are very quickly self evidently proven or not.
Here I'm not talking about numbers anywhere near the LL double check region anyway.
There is no issue of reassignment, since mersenne.org does not conveniently offer assignment of such low numbers at all.

lycorn · 2015-10-17, 08:39

True. I think the only precaution you should take is to stay clear of ramgeis working ranges, so you don´t step on each other toes. I noticed you are working on "dangerously" close exponents.

snme2pm1 · 2015-10-17, 10:01

Quote:

Originally Posted by lycorn

True. I think the only precaution you should take is to stay clear of ramgeis working ranges, so you don´t step on each other toes. I noticed you are working on "dangerously" close exponents.

Well, I didn't notice him active at times when I reserved 2200 bricks of work in that region at a time.
The problem with the mid 5M region was that it had lots of ECM assignments, which is why of I left some of it alone.
Evidently ramgeis has been trying to clear the remainder, however many unfinished ECM jobs linger.
I have continued into the 6M region, again reserving 2200 like slabs at a time.
Not many other folk reserve like that. At least none that I know.

But seriously off topic; I've just be out to see the local festival fireworks.
Found a scruffy bloke that felt inclined to walk into the ladies toilet, which wife waiting at the door wasn't pleased about.
Now I'm not a big fellow, but I am a rare non-black hair person in this very asian suburb; I thought I might have to take him down, he seemed small enough, but after a little dancing with him and pointing out the security cameras he stopped facing me off.
By the way, he didn't have black hair!

lycorn · 2015-10-17, 11:15

Quote:

Originally Posted by snme2pm1

By the way, he didn't have black hair!

Are you sure you weren´t dancing with yourself?

LOL.

snme2pm1 · 2015-10-17, 11:24

Quote:

Originally Posted by lycorn

Are you sure you weren´t dancing with yourself?

LOL.

The event was reported to security officers, in retrospect it isn't funny at all.

2015-10-16, 22:38	#35
snme2pm1 "Graham uses ISO 8601" Mar 2014 AU, Sydney 363₈ Posts	M5618611 It is of low significance, however there are unfriendly occasions when results are submitted which prematurely expire outstanding assignments. http://www.mersenne.org/report_expon...exp_hi=&full=1 That exponent had time remaining for lopper to complete an ECM assignment. I note that ramgeis lodged a TF factor found result before such expiry. I am aware of this because I held the same TF factor found result for several weeks, but very deliberately abstained from submitting such result until the official primenet expiry realisation. I was doubtful that lopper would return a result before expiry, but that's not the point.

2015-10-17, 06:45	#39
LaurV Romulan Interpreter Jun 2011 Thailand 2·5·31² Posts	P-1 expired It doesn't really fit in this category, but I didn't want to make a new thread for it, I think something odd is going on with the server, which may be connected or not with the "results not needed" story. I use one of my computer to reserve work, P95 is not crunching, but only stays open and reserves P-1 work. If you "pause" the workers, it will still connect daily to the server and report results/get assignments, etc. I did this in the past, and it was working well. Another computer (which has GPU) does the work. Now, it just reported two P-1 results, which of course, were assigned by PrimeNet (yes, they are not in the bunch wrongly assigned by GPU72, but were directly requested as P-1, and assigned by PrimeNet, together with other, total 38 P-1 jobs). One reported result went well, but for the other one, things got weird. I just found now that instead of being marked as completed, the job was expired, and it was considered as "poached". For which reason, it was not deleted from the worktodo (technically, I still can "finish" the job, and report the result, which will become "not needed"). I deleted it manually from the worktodo (the other one was deleted automatically) Note that the job is done on the GPU, on another computer, which reported "manually" the result (by a script), but the reservation is done by this computer which I use now, and I only use it to reserve work. The things are simple: it tries to connect to the server, and reserves work, because the worktodo is getting low, the job is done (by the other computer's GPU, which reads from worktodo file), then, when this computer connects daily to the server, it finds that the key is not valid, and the assignment is deleted from the worktodo. Because of that, worktodo shrinks, and when it is too less work to do, other work is brought in. Which in turn, is read by, and done, in the other computer (is GPU work). This worked well in the past. I just re-started last week, and now it got the keys scrambled somehow. I still have to get access to the other computer (in few hours) and see if it used the right key when it reported the results, before telling for sure that is a server problem, but most probably yes, it is a server problem, and the right keys were used, otherwise the other exponent would not be "completed", but "poached by myself", too. (edit: now I got totally crazy, not only replying to my own posts, 3 time in a row, but also poaching my own assignments ) (edit 2: luckily, Madpoo's crosspost saved me this time, haha) Last fiddled with by LaurV on 2015-10-17 at 06:54

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
tf results not needed	MatWur-S530113	GPU to 72	10	2017-12-09 15:36
Results not Needed	RMAC9.5	PrimeNet	3	2013-06-26 13:16
not needed	zeit	PrimeNet	3	2008-04-25 08:03
Help needed	AntonVrba	Math	3	2007-03-06 10:55
V24.12 QA help needed	Prime95	Software	5	2005-06-17 15:54

2015-10-17, 08:39	#41
lycorn Sep 2002 Oeiras, Portugal 2700₈ Posts	True. I think the only precaution you should take is to stay clear of ramgeis working ranges, so you don´t step on each other toes. I noticed you are working on "dangerously" close exponents.