MISFIT 2.9
 

I am currently working on 2.8 and the first item on the list is [B]preventing partial results[/B] from being exported/uploaded.

The condition is as follows

The first line in WorkTo.txt

[B]Factor=2014-4-3,[COLOR="Red"]91175807[/COLOR],69,70[/B]

Results.Txt contains the following (somewhere)

[B]no factor for M[COLOR="red"]91175807[/COLOR] from 2^66 to 2^67 [mfaktc 0.20 barrett76_mul32_gs]
no factor for M[COLOR="red"]91175807[/COLOR] from 2^67 to 2^68 [mfaktc 0.20 barrett76_mul32_gs]
no factor for M[COLOR="red"]91175807[/COLOR] from 2^68 to 2^69 [mfaktc 0.20 barrett76_mul32_gs]
[/B]


As you can see we have committed to processing the range 66 -> 70 but 69,70 is still underway.

Extending a range is a feature in MISFIT, but recently GIMPS has been assigning out work that can be quickly turned around and between result uploads another person begins working the same exponent in the range you have committed to. In this condition someone will end up with a "results not needed" message from GIMPS on upload. Preventing the partial results upload will ensure there is no chance of colliding with another TF-er. This will be a configuration item with a default value of ON.


Comments?


Any requests for 2.8?

It sounds great to me!

:tu:

2.8.0 released
 

VERSION 2.8.0
1. Added the ability to hold back completed results for an exponent if a bit level is still being processed. This will ensure GIMPS only gets results when the entire pledge is completed. Default is ON. Results that are held back during an export are logged for forensic purposes.

2. Improvements to logging
2.1 An improved naming convention where instead of random GUIDS all file names now contain an embedded DateTime.Now.ToFileTimeUtc().ToString() component
2.2 File name prefixes have been clarified
2.2.1 COMBINED-RESULTS* now have the name RESULTS_EXPORTED*
(This file name chang means old files are not used in the productivity calculations unless you manually rename them)
use the command line rename utility to manually fix {ren MISFITbackups\COMBINED-RESULTS* RESULTS_EXPORTED*}

3. For forensic purposes the error form now logs all text to MISFITbackups\ERROR_FORM_*

4. Lots of code cleanup

5. The leaderboard is turned off until GIMPS site can provide structured data for me to process.



Get from [url]http://mersenneforum.org/misfit/[/url]

Enjoy

[QUOTE=swl551;386089]...
5. The leaderboard is turned off until GIMPS site can provide structured data for me to process.
...[/QUOTE]

I hope to have a nicely structured XML of the "top producers" info by later tonight. It's mostly just modifying the current job to also spit out some XML in the process.

Once that's ready I'll PM swl551 since we've already discussed it, but I'll also mention it here. I don't know if I'd put a link to it since I don't know how useful it'll be, but anyone's welcome to use it of course. It's a simplified and shrunken version of the report on the site... it'll just have rank, account name and their total GHzDays score.

Format will be (I think) along the lines of:
<topProducers>
<producer rank="1">
<name>curtisc</name>
<ghzDays>12345.67</ghzDays>
</producer>
</topProducers>

[QUOTE=Madpoo;386123]I hope to have a nicely structured XML of the "top producers" info by later tonight. It's mostly just modifying the current job to also spit out some XML in the process.

Once that's ready I'll PM swl551 since we've already discussed it, but I'll also mention it here. I don't know if I'd put a link to it since I don't know how useful it'll be, but anyone's welcome to use it of course. It's a simplified and shrunken version of the report on the site... it'll just have rank, account name and their total GHzDays score.

Format will be (I think) along the lines of:
<topProducers>
<producer rank="1">
<name>curtisc</name>
<ghzDays>12345.67</ghzDays>
</producer>
</topProducers>[/QUOTE]

Okay... kersmash! Here we go...

[URL="http://www.mersenne.org/sp_report_primenet_gimps_top_500.xml"]http://www.mersenne.org/sp_report_primenet_gimps_top_500.xml[/URL]

Give it a try, see if the data and format is any good. It'll be generated every hour along with the stats on the site itself. I added a timestamp attribute to the main element.

As mentioned, it's pretty simple, just the rank, name and ghz-days total. I could add in some of that other 90/30/7 day stuff like the web version has but unless anyone is planning to crawl the XML for that, I'll just omit it to keep the XML lean and mean.

@madpoo

Thanks! I will update MISFIT later today.

Bug holding back results
 

I am pulling 2.8 off the release server and putting 2.7.8 back. If you have already upgraded to 2.8 you can stick with it as long as you turn off the HOLD BACK feature.


The bug is that the hold back feature will put the same hold backs in each directory. This is only a problem if you have MISFIT handling more than one FACTORING directory. If this has happened to you you can recover the original results from the MISFITbackups directory.



Sorry.

2.8.1 released
 

VERSION 2.8.1
1. Fixed problem with holding results back when MISFIT is managing more than one directory.
2. Fixed problem with Wrap-Up feature giving error message.
3. minor fixes and touch ups here and there.


I put significant time into [B]totally[/B] re-writing the EXPORT functionality. Not only did I get that nasty HOLD-BACK bug out, but the code is significantly cleaner now too.

:tu: Thanks a lot!
Will try it at once!

V 2.8.1 worked fine "out of the box". The hold back function works exactly as advertised.
Thanks, Scott! :tu: