mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet > MISFIT

Reply
 
Thread Tools
Old 2015-02-24, 15:45   #12
kladner
 
kladner's Avatar
 
"Kieren, ktony"
Jul 2011

2·3·37·43 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
cat *.txt > merged.txt

Even works in DOS.
Windows 7 command line does not like "cat". However,
Code:
copy results_exported_130*.txt merged.txt
works in my circumstances. The "130*" might be different on other systems. Those happen to be the first 3 digits following "results_exported_". For me, at least, files which have letters instead of numbers in that position are "per instance" results, while numbers correspond to combined results. The numbers are needed to exclude files beginning with "backup_", which again, are for individual GPUs.
kladner is online now   Reply With Quote
Old 2015-02-24, 17:31   #13
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013
Ͳօɾօղէօ

53348 Posts
Default

Quote:
Originally Posted by kladner View Post
Windows 7 command line does not like "cat".
I was wrong. It's `type` in DOS.
Mark Rose is offline   Reply With Quote
Old 2015-02-24, 21:35   #14
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

2·401 Posts
Default

Personally I would use the files in directory GIOM_SENT as it contains only combined results (not per cpu)

D:\GIMPS\TF\GIOM_SENT>copy r*.txt master_report.txt
D:\GIMPS\TF\GIOM_SENT>notepad master_report.txt


Of course this assumes MISFIT is performing results uploading.
swl551 is offline   Reply With Quote
Old 2015-02-24, 22:59   #15
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11·157 Posts
Default

Alright. Thanks folks.
TheMawn is offline   Reply With Quote
Old 2015-06-21, 01:32   #16
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11·157 Posts
Default

I have added my GTX 980 Ti to the TF effort. It's actually a very impressive card, and I look forward to un-crippling it by putting a proper cooler on it...

For now I have a GTX 980 Ti and a GTX 660 Ti on the same system and there is a very gross throughput difference between them so I think I need to adjust something in MISFIT as a result.

I came back to my computer to see that GTX 980 Ti had stalled at some point and the first thing I think of is that worktodo.txt ran out. I used the MISFIT calculator and just added both of my GPUs' throughputs (250 + 750 = 1000 GHz Days per day) and applied all the values. I think what happened is MISFIT has no way to know that one of the instances is cranking through so much more work than the other, because my GTX 660 Ti has 23 assignments queued (it usually has a half dozen or so).

I set MISFIT to balance worktodo files when the row difference is 5 or more. Does this make sense as an adjustment?
TheMawn is offline   Reply With Quote
Old 2015-06-21, 16:12   #17
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

2×401 Posts
Default

Hi, Work Balance is designed to handle the differences in card throughput just as you described.

The default setting of 5 is fine in most cases. I've never needed to change it, but it was implemented back when the project was doing factoring to 72, not 75. That right there will change the physical count of assignments typically stored on your PC. To compensate adjust the re-balance value or increase the number of assignments stored on your PC to ensure there is time to get the work moved around before a run-out occurs.


Another issue is the defaults for how MFAKTC implemented workToDo.Add

(from my release notes of MISFIT VERSION 2.10.0)
Quote:
Added configuration to push work assignments into WorkToDo.add instead of WorkToDo.txt
Note: The existence of a WorkToDo.add is a blocking condition and MISFIT will not add more work to an existing .ADD file
This block also applies to Wrapping-up processing and balancing work.
MISFIT will display errors if it collides with an .ADD file
Note: As of this writing if mfaktc exhausts all work from WorkToDo.txt it will gracefully end.
The problem is that upon restart it does not inbound work that might be waiting in a .ADD file and exits with protest that no work is available.

****
The default interval for mfaktc to fully process an .ADD file is 20 minutes which increases the likelihood of this condition occurring.
If it happens then you will have to manually move rows into WorkToDo.txt
*****
The shortest possible .ADD process cycle is 1 minute. See mfaktc.ini for more information on WorkFileAddDelay

So MISFIT cannot re-balance work in a timely manner if you allow MFAKTC to sit on the .ADD file for such a long time.

Let me know how it goes.

Thx

Scott

Last fiddled with by swl551 on 2015-06-21 at 16:15
swl551 is offline   Reply With Quote
Old 2015-06-21, 20:21   #18
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11·157 Posts
Default

Okay, thanks. 5 seems to work just fine. I've increased my fetch threshold from the calculated 1250 to 1875 GHz-Days which is what the value would be if both of my cards were GTX 980 Ti's. That is still far, far less than a week's worth of buffer so there are no issues there.

I'm actually still on 2.9.5. I'll have to dig back through the mfaktc and misfit notes to understand why people wanted misfit to use worktodo.add, since MISFIT kind of does the same thing already.
TheMawn is offline   Reply With Quote
Old 2015-06-21, 22:00   #19
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

8,861 Posts
Default

Quote:
Originally Posted by TheMawn View Post
I'm actually still on 2.9.5. I'll have to dig back through the mfaktc and misfit notes to understand why people wanted misfit to use worktodo.add, since MISFIT kind of does the same thing already.
Race conditions.
chalsall is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
questions about misfit jschwar313 MISFIT 1 2016-02-02 00:32
Misfit feature request. Aramis Wyler MISFIT 15 2015-03-04 21:13
MISFIT 2.9 swl551 MISFIT 44 2014-12-15 12:46
MISFIT for MFAKTx swl551 MISFIT 186 2014-02-05 23:27
(archive)MISFIT swl551 MISFIT 584 2013-03-19 20:13

All times are UTC. The time now is 20:58.

Thu Apr 9 20:58:45 UTC 2020 up 15 days, 18:31, 1 user, load averages: 1.07, 1.39, 1.52

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.