mersenneforum.org  

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

Closed Thread
 
Thread Tools
Old 2012-11-21, 20:35   #45
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

14428 Posts
Default

Quote:
Originally Posted by flashjh View Post
I'll give it a shot, but I use mfaktc and do the whole run at the same time. If it's getting the info from the work file, it should show 71-73, right? Or is it actually getting the info from the instance window?
So you have your mfaktc.ini file set to
Stages=0

If so then there is no way to infer the bit range based on the factor line. I do not read it from the console output.

Oh well.... I can't support every possible deviation from the defaults.
swl551 is offline  
Old 2012-11-21, 20:44   #46
flashjh
 
flashjh's Avatar
 
"Jerry"
Nov 2011
Vancouver, WA

1,123 Posts
Default

Hey, no biggie. I run stages=0 because it is a little more efficient. AFAIK mfakto supports stages=0 also, but I haven't used AMD cards in a while.

I'm fine with just seeing the current exponent.
flashjh is offline  
Old 2012-11-21, 20:56   #47
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

2×401 Posts
Default

Quote:
Originally Posted by flashjh View Post
Hey, no biggie. I run stages=0 because it is a little more efficient. AFAIK mfakto supports stages=0 also, but I haven't used AMD cards in a while.

I'm fine with just seeing the current exponent.
Flash. Help me out here. Are you saying the exponent changes? I thought the bit level changed but the exponent was fixed throughout the duration of the factoring process.
swl551 is offline  
Old 2012-11-21, 21:10   #48
flashjh
 
flashjh's Avatar
 
"Jerry"
Nov 2011
Vancouver, WA

1,123 Posts
Default

Quote:
Originally Posted by swl551 View Post
Flash. Help me out here. Are you saying the exponent changes? I thought the bit level changed but the exponent was fixed throughout the duration of the factoring process.
You're right, the exponent is the same for the entire worktodo line. I just meant that I can ignore the bit level and just see the exponent that is currently being worked.
flashjh is offline  
Old 2012-11-21, 21:14   #49
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

32216 Posts
Cool

Quote:
Originally Posted by flashjh View Post
You're right, the exponent is the same for the entire worktodo line. I just meant that I can ignore the bit level and just see the exponent that is currently being worked.
Phew. otherwise I'd have to state I really don't have a clue!!
swl551 is offline  
Old 2012-11-22, 19:03   #50
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

2·401 Posts
Default 2.0.8.1

VERSION 2.0.8.1
1. Added detection of unexpected modifications (writes) to workToDo.txt files during Balance. This would be possible if mfaktO/C modified the file after the balance process started. An error dialog is shown with the file that had the collision. If you ever see this error let me know. (see screenshot)

2. The entire HTML response received from GIMPS during GIOM events is saved in the GIOM_LOGS directory. This will allow you to verify what happened during an IO event.. Useful for research.

3. GridStats "Working" column no longer assumes the factor is done in stages so it just shows the exact value found in the top row of workToDo file.

4. Ability to view all results in a consolidated file. Useful for research

Get from http://www.mersenneforum.org/misfit/
Attached Thumbnails
Click image for larger version

Name:	CollisionError.PNG
Views:	147
Size:	14.0 KB
ID:	8894  
swl551 is offline  
Old 2012-11-24, 20:32   #51
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

32216 Posts
Default 2.0.10 is out there. A handful of small changes

Work this weekend has been toward fine tuning MISFIT and improvement in detecting error/alert conditions that are possible within your MFAKTO/C instances.

VERSION 2.0.10
1. Modified the alerting regarding GHZdayWorkToDo. The determination to send an email in response to a low work condition is now the secondary method of alerting. The primary method is now a dialog message. Sending email is optional. Since the previous implementation was ambiguous you will have to go into the configuration and re-select your preferred options.

2. Added Exclamation Sound and blink red every 30 seconds to Error/Alert Dialog.

3. fixed bug where stalled instance cell being red was lost at next updateGridStats cycle.

4. Added duplicate exponent detection to the UpgradeGridStats function. Will show any duplicates.
swl551 is offline  
Old 2012-11-25, 10:42   #52
NormanRKN
 
NormanRKN's Avatar
 
Jul 2012
Saarland / Germany

1048 Posts
Thumbs up

Thank you Scott.
NormanRKN is offline  
Old 2012-11-27, 08:35   #53
Bdot
 
Bdot's Avatar
 
Nov 2010
Germany

3·199 Posts
Default

Hi Scott,

great thing! And as usual, I'm pretty late in noticing the news ...

I have suggestions/comments on some of the posts I've read in this thread.

Did you know that mfakto uses lock files when manipulating worktodo or results files? If mfakto.ini has the line
WorkFile=w1.txt
then it will try to create a file w1.txt.lck (atomically using open(..., O_EXCL|O_CREAT,...) ). If that fails because it already exists, then mfakto will wait for ever increasing delays (between 1 ms and 1s) and retry endlessly. Similar for the results file.
So if you create the .lck file before touching worktodo or results files, mfakto should not interfere, avoiding the mentioned synchronization issues. I once talked about this with Oliver, but I'm not sure if he's integrated this into mfaktc or not.

Regarding Stages=0: yes, basically mfakto supports this as well. However, mfakto has a lot more kernels than mfaktc, and only few of them can handle multiple bitlevels at once. If you set Stages=0 and the assignment really has multiple bit levels, then mfakto will select a kernel that can handle multiple bitlevels, but may be slower than a kernel that would be chosen otherwise. Therefore I recommend to run the bit levels one by one for mfakto. I'm currently working on a montgomery kernel that hopefully is faster than all kernels before, and this one can run Stages=0.

Can MISFIT also be used to manage worktodo and results files that reside on some network share without actually managing mfakto/c (because they run on Linux)?

And yes, I'd like to have GPU72 integration as well

Last fiddled with by Bdot on 2012-11-27 at 08:40
Bdot is offline  
Old 2012-11-27, 16:45   #54
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

2·401 Posts
Default

Bdot,
Thx for the info. I knew mfaktO implemented locking but did not know the mechanism so thanks for the info. MISFIT can easily support that processes. Getting mfaktC using the same process would be ideal, but I leave that to others to drive. Of course IO collisions are possible but the timing has to be so specific that it is statistically unlikely. Flashjh and I have worked together to try to re-create conditions to account for some of the items he mentioned in an earlier post. So far we not able to reproduce the results. I'd more likely to add support for the file-lock system if it were in mfaktC since the ROI is greatest if everyone is doing the same thing.

I personally use Stages=1 as I like to see the results posted at the end of each run. Others don't. MISFIT supports either configuration.

MISFIT can manipulate files on remote machines as long as the OS supports emulating a windows share. SAMBA being an prominent implementation. Others have confirmed MISFIT (early versions) runs fine under MONO on linux too. Sans the two functions that call WIN32 api directly. Not sure if the current version still works fine on MONO.

GPU72 integration is still on hold until Chalsall provides integration specs.

Thanks for your comments.
swl551 is offline  
Old 2012-11-29, 15:36   #55
swl551
 
swl551's Avatar
 
Aug 2012
New Hampshire

80210 Posts
Default MISFIT 2.1.0

VERSION 2.1.0
1. Added configuration to display the "working" factor column based on phases. This is the default method of operation for mfaktO/C. See their readme for more info.

When showing factors in phases the format is
working_range|max_bit_level|exponent

When showing factors without phases the format is
full_range|exponent (as viewed in releases prior to 2.1.0)

2. Complete re-write of balance work process to simplify code, reduce file I/O, and introduce error detection and reporting via the MISFIT alert panel.

3. Added menu option to view GIOM log directory

4. Added ability to dbl-click mfaktX locations which then opens Windows Explorer for that directory.


Get from http://www.mersenneforum.org/misfit/
Attached Thumbnails
Click image for larger version

Name:	misfit210.PNG
Views:	146
Size:	41.9 KB
ID:	8920  
swl551 is offline  
Closed Thread

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
questions about misfit jschwar313 MISFIT 1 2016-02-02 00:32
Help with Misfit TheMawn MISFIT 18 2015-06-21 22:00
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

All times are UTC. The time now is 21:10.

Thu Apr 9 21:10:21 UTC 2020 up 15 days, 18:43, 1 user, load averages: 1.39, 1.52, 1.53

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.