Prime95 version 25.5

[henryzz]

in prime95 25.5 i dont get an option for recieving ecm or p-1 work i only get the options that were on the old server
do i have to put something in prime.txt to access the new server

[garo]

You need to send George a PM to get the 'secret' line that allows you to beta-test with the new server.

[henryzz]

thanks

[James Heinrich]

Strange problem: Thread #3 seems to be getting "stuck". It gets through its ECM-stage1 work and then just freezes.
(it's now 17:29, 8+ hours since last update message)

[Dec 21 09:07] Work thread starting
[Dec 21 09:07] Using FFT length 12K
[Dec 21 09:07] ECM on M250969: curve #1 with s=7722991907350575, B1=50000, B2=5000000
[Dec 21 09:07] M250969 curve 1 stage 1 at prime 21937 [43.87%].
[Dec 21 09:07] M250969 curve 1 stage 1 at prime 29753 [59.50%]. Time: 19.866 sec.
[Dec 21 09:07] M250969 curve 1 stage 1 at prime 37501 [75.00%]. Time: 20.252 sec.
[Dec 21 09:08] M250969 curve 1 stage 1 at prime 45293 [90.58%]. Time: 20.127 sec.
[Dec 21 09:08] Stage 1 complete. 721475 transforms, 1 modular inverses. Time: 12.121 sec.
[Dec 21 09:08] Round off: 0.21875

Attempting to stop all worker threads doesn't cause it terminate, I actually have to kill (with TaskManager, file|exit isn't sufficient) all of Prime95 to be able to restart the threads (since you can't restart workers until they've all finished a previous stop request).

edit: Restarting Prime95 after killing it resumes M250696 ECM stage1 at 43.87%, presumably the point of the last savefile.

edit2: Before trying to stop workers Prime95 uses ~100% CPU across 4 threads on 4 cores. After attempting to stop workers, it's down to 0% CPU usage but worker #3 never successfully terminates.

[henryzz]

with this version there isnt a way to set how much work u want queued or to just finish the work u r on

my hundredth post

[Andi47]

Originally Posted by henryzz

my hundredth post

...and no single questionmark, full stop or other punctuation used yet

[Prime95]

Originally Posted by James Heinrich

Playing with my new quad-core, I have two (of 4) threads fighting for RAM. Each of them start their Stage2, see another thread is trying to use RAM, then restart with new settings and they just keep looping without either of them actually claiming RAM and doing any real work

James, I've looked at the code and can't see a problem. Unfortunately, I'm not sure if the current code is the code you are running. My bad. I only recently resumed using source control.

Please download 25.6 and see if the problem still happens. If it does, email me the prime.txt, local.txt, worktodo.txt files. Send the save files too.

[Prime95]

Originally Posted by Andi47

Reliability, Confidence  	1.0, 0.0

Is it possible that my memory settings cause Prime95 to repeatedly restart stage 2 of an ecm curve with new memory setting every time when daytime and nighttime begins, and then never finishes the curve - and so the server assumes that the cpu is not reliable and refuses to assign anything?

Reliability (scale of 0.0 to 1.0) and confidence (scale of 0 to 10) is only used in assigning LL tests. It is only updated when LL results are reported. Right now the server thinks you have a very reliable machine - but no confidence in that assumption.

I can easily adjust the reliability / confidence thresholds for assigning the smallest first-time and double-check LL tests.

[Prime95]

Originally Posted by Andi47

Has the server run out of "ECM on small Fermat-Numbers" work? I have one worker thread set to this type of work, but it has got several "TF on Mersenne numbers" assignments.

I looked at your log file. The server thinks you want trial factoring work. The server is out of that work type.

[Prime95]

While making a change so that the server assigned ECM work when it is out of TF work I accidentally obliterated a PHP file. So at 11:30 EST the v5 server started acting strangely. At first refusing connections, then handing out ECM work instead of TF-LMH work. Should be all fixed now. Sorry.

[harlee]

Originally Posted by Prime95

At first refusing connections, then handing out ECM work instead of TF-LMH work. Should be all fixed now. Sorry.

Thanks for the info, just saw all of the ECM work. Since I got the same ECM exponent, I went through the server and unreserved the exponents.

Also, can you check the rolling average on v25.5? Mine doesn't seem to be changing, even if I run two prime95 programs for 24hrs a day for several days on my P4.