GPU to 72 status...

[Madpoo]

Originally Posted by chalsall

A follow up.

Just look at this query to get an idea what we're dealing with.

No disingenuous assumption intended, but a stupid AI has a bit of a problem dealing with that with a scarce resource on a five minute timescale.

Aurashift has some nice resources and it seems like he blazes through the P-1 work fast and efficiently, so that should be fine. Mr P-1 however, looks like a lot of old assignments in there, over a month old and estimated completions sometime early next year. I have no knowledge of that user so I don't have an opinion one way or another on that.

[chalsall]

Originally Posted by Madpoo

Mr P-1 however, looks like a lot of old assignments in there, over a month old and estimated completions sometime early next year. I have no knowledge of that user so I don't have an opinion one way or another on that.

Mr. P-1 was actually the genesis of GPU72. He was coordinating GPU TF'ing (via PMs and email) before GPU72 came online.

To put on the record, I don't think anyone is being malicious here. But, when being only a few hours ahead in the TF'ing for P-1'ing, sometimes things get handed out sub-optimally TF'ed.

Sub-optimal, but not the end of the world.

[chalsall]

Originally Posted by LaurV

Today, after this discussion, I enabled the proxy again. Remark that the computer is set to get First Time Tests. Guess what kind off assignments I have? Of course, you guessed right: I just got 22 pfactor lines in the last hour.

Do you remember what I suggested you do?

In case you do not... Change the settings for each client, and have the client communicate with Primenet through the proxy. Then change the settings back to what you want (in this case, LL) and communicate again. That should fix things.

It's probably a SPE on my part, but this should solve the problem.

[LaurV]

Originally Posted by chalsall

But, when being only a few hours ahead in the TF'ing for P-1'ing, sometimes things get handed out sub-optimally TF'ed.

Let me finish Manfred's DC (275M mark passed with a match, ETA ~11 days) and then I promise I bring 1800GHDays/Day LLTF power till the end of the year

Regarding the second part, LL versus P-1 assignments, I have to end the working day and go home to try that (now lunch break, but still at job), but I am skeptical, because I did that in the past (and I just "did it", well, partially, when switching back to the proxy, after all these months when I was getting right assignments from PrimeNet). I remember some other (Petrw?) having the same problem and you "tickled" some small wire inside of that server, which solved the problem. But wait first till I get home, and I will be back to you. The P-1 assignments may have been finished by now, and if your suggestion won't work, it is just the right time to get some new of them, requesting LL work

[retina]

Originally Posted by chalsall

... a SPE on my part ...

Submassive Pulmonary Embolism?

Swedish Penis Enlarger?

[LaurV]

@Chris: don't reply to retina, he knows the meaning, he only likes to hear you saying it...

edit: actually, you can reply, we all like to hear you saying it...

[Dubslow]

Originally Posted by LaurV

@Chris: don't reply to retina, he knows the meaning, he only likes to hear you saying it...

edit: actually, you can reply, we all like to hear you saying it...

It's a term I've incorporated in my mental reviews of my own code

[LaurV]

Nope, it won't work. Going back and forth how many times you want, it cycles like that:

Getting assignment from server
PrimeNet success code with additional info:
GPU72 assigned P-1 factoring work.
Got assignment c10c95e323f2f65...: P-1 M73578259
Sending expected completion date for M73578259: Oct 10 2015
Getting assignment from server
PrimeNet success code with additional info:
GPU72 assigned P-1 factoring work.
Got assignment 3b2e431a68a2ff63...: P-1 M73578199
Sending expected completion date for M73578199: Oct 11 2015
Updating computer information on the server
Getting assignment from server
PrimeNet success code with additional info:
Server assigned Lucas Lehmer primality test work.
Got assignment 84839C0DBCB...1C61B: LL M76914287
Sending expected completion date for M76914287: Nov 03 2015
Getting assignment from server
PrimeNet success code with additional info:
Server assigned Lucas Lehmer primality test work.
Got assignment A6ACC9637F5...E96: LL M76915441
Sending expected completion date for M76915441: Nov 03 2015

Your proxy hasn't even a little pain in the button about what type of assignment I want. You need to tickle that little wire. Now, going through this procedure, and because it happened again, I remember the odd things that happened at that time (documented on the forum), which made me to delete the proxy: one core was getting P-1 assignments and another core was getting normal, LL/DC assignments when I was using the proxy. Which was pissing me off.

It doesn't matter anyhow, I won't use the proxy (and I will do the P-1, I won't unreserve them, in spite of the fact they were unwanted, I need some P-1 activity on that account ) and it is not affecting me. I only resurrected this discussion because people wonder "why so many P-1 assignments", "we are concentrating too much power on P-1", etc. This may be a reason: they get P-1 assignments when they ask for another type of work.

[petrw1]

Originally Posted by LaurV

Nope, it won't work. Going back and forth how many times you want, it cycles like that:.

I found the sequence of events is very specific....give me a little time to dig out the exact post somewhere in the previous couple hundred pages. :P

[kladner]

For me, this has worked:

Stop P95
Change work type to something you don't want to end up with
Have P95 communicate and send completion dates
Change work types to what you want
Have P95 communicate with server
Log into PrimeNet and see what it says your work types are. Correct, if necessary.
Have P95 communicate with server
See if the settings in P95 are correct
Unreserve unwanted assignments, which will have come in during these contortions.
Make sure P95 settings and PrimeNet settings agree

YMMV. Not all of these steps may be necessary. This sequence is aimed at beating things into submission.

[chalsall]

Originally Posted by kladner

YMMV. Not all of these steps may be necessary. This sequence is aimed at beating things into submission.

LaurV, could you please give this a try?

For that particular machine, the GPU72 database doesn't see a change for each Computer_CPU record since 2013-11-26.

Again, it's probably a Stupid Programmer Error (SPE) () on my part, but it would be interesting to analyse the logs generated by the attempt.