OFFICIAL "SERVER PROBLEMS" THREAD

[kriesel]

Manual submission of an LL result advises to upgrade to v30.3 (the current version for prime95 or mprime). Even when the result line submitted clearly indicated it was from CUDALucas, for which the current version is v2.06, supporting neither Jacobi check for LL, nor PRP nor GEC nor PRP proof.
(And in this case since CUDALucas does not include user/cpu/gpu info in the result line, the server has no way to know whether the gpu involved is compatible with gpuowl.)

 M( 161000009 )C, 0xcb6eec7da5fc040a, offset = 9507, n = 9216K, CUDALucas v2.06beta

Some branching based on application indicated in the result seems in order, to avoid recommending upgrading GPU applications to a cpu app version number.
All prime95 instances feasible have been upgraded to prime95 v30.3 already here. But the various gpu applications are not prime95.

[James Heinrich]

Originally Posted by kriesel

Manual submission of an LL result advises to upgrade to v30.3 .. even when the result line submitted clearly indicated it was from CUDALucas
Some branching based on application indicated in the result seems in order, to avoid recommending upgrading GPU applications to a cpu app version number.

Thanks, the warning check should be more selective now. Please let me know if you get any further inappropriate warnings.

[kriesel]

The form that produces https://www.mersenne.org/assignments...xtf=1&exdchk=1
needs an exclude CERTS check box. Certs are a type of brief PRP DC and throw off the count of first tests active before reaching a milestone exponent level. But Certs are included in the list and count even when exclude double checks is in effect.

[Uncwilly]

Yet again we have someone taking 100's of exponents for TF down in the DC Cat 0-1 range.
The user "Hacker" has over 700 assignments in the 53-53M range and the lowest ETA is over 4.5 years. It might be time for honest contributors to ignore any TF assignments in DC Cat 0-2. Or give up on working the bottom and work elsewhere.

[Prime95]

Originally Posted by Uncwilly

Yet again we have someone taking 100's of exponents for TF down in the DC Cat 0-1 range.
The user "Hacker" has over 700 assignments in the 53-53M range and the lowest ETA is over 4.5 years. It might be time for honest contributors to ignore any TF assignments in DC Cat 0-2. Or give up on working the bottom and work elsewhere.

A weird case. User account created in 2011, so it seems unlikely that this is a mischief maker. Requests were from prime95 because expected completion dates were sent. At first I thought it might be a "runaway" execution (perhaps unable to write to worktodo.txt thus requesting more and more). However, the requests stopped at 700 some-odd.

I freed the assignments. If it happens again, it isn't too hard to suspend the account.

[Aramis Wyler]

I used to like putting DC work on my slower processors, but now I just group them up and work on PRPs because in this new era of Prime95 30+ the DCs are actually a limited and vanishing resource. This is volunteer work and people can do as they like - I am certainly not suggesting rule or anything - but I think that the DCs should probably be saved for prime95 < v30, as that should be the only work they do.


There are infinite larger numbers to work on.

[kriesel]

Originally Posted by Aramis Wyler

I used to like putting DC work on my slower processors, but now I just group them up and work on PRPs because in this new era of Prime95 30+ the DCs are actually a limited and vanishing resource. This is volunteer work and people can do as they like - I am certainly not suggesting rule or anything - but I think that the DCs should probably be saved for prime95 < v30, as that should be the only work they do.

There are infinite larger numbers to work on.

S..l...o...w systems may be too slow for DC soon too. P-1 or Cert work would be ok a while. Eventually, it's better to shut them down and invest the electricity saved into newer faster hardware.
(But yeah, I have some ancient gear slogging along now.)

There is value to getting the remaining LLDC done, up to 58M, or to 83M. It clarifies whether a prime below Mp48* to Mp51* got missed in the first pass due to an error.

[kriesel]

PRP reservations are listed as LL in exponent lookup.
https://www.mersenne.org/report_expo...exp_hi=&full=1 is an example of a PRP first test assignment incorrectly displaying as LL first test work type. These are primenet-connected prime95 where it hasn't started running that assignment yet.

There are cases where work type says LL first test but stage says PRP; https://www.mersenne.org/report_expo...exp_hi=&full=1 for example. These are primenet-connected prime95 in progress.

Manual assignments appear to be handled correctly: https://www.mersenne.org/report_expo...exp_hi=&full=1

[ATH]

Originally Posted by Prime95

A weird case. User account created in 2011, so it seems unlikely that this is a mischief maker. Requests were from prime95 because expected completion dates were sent. At first I thought it might be a "runaway" execution (perhaps unable to write to worktodo.txt thus requesting more and more). However, the requests stopped at 700 some-odd.

I freed the assignments. If it happens again, it isn't too hard to suspend the account.

Ehh now Curtisc has approximately the same assignments?

https://www.mersenne.org/assignments...first=1&exp1=1

[Uncwilly]

Originally Posted by ATH

Ehh now Curtisc has approximately the same assignments?

https://www.mersenne.org/assignments...first=1&exp1=1

I bet that the screen name is the same, but not the e-mail associated with the account.
And the estimated completion date is nearly 8 years away.

[kriesel]

scroll down. 10939d/365 ~30 years.

This shall not stand.