For those of you who are trying to crunch more than work unit at a time on WCG's now GPU project Help Conquer Cancer here are some app_info.xml files to help you to do so.
Printable View
For those of you who are trying to crunch more than work unit at a time on WCG's now GPU project Help Conquer Cancer here are some app_info.xml files to help you to do so.
Do you have anything about "--device" parameter?
No, sorry. I'm not familiar with that.
What is that in reference to? The double hash (--) is used to signify a command line option of a program like boinc.exe, but boinc.exe, boinccmd.exe nor boincmgr.exe use that option.Quote:
Originally Posted by Robert7
"-- device 0/1/2/3" is used in cmdline in app_info to select GPU, but I don't know - works it for HCC or not?
Sorry but I'm at a loss on this one. Maybe post your question on the WCG forum (Here)
XS has a thread, where they have the app_info.xml files for many of the sub-projects at WCG. This allows you to combine GPU crunching for HCC, with other CPU sub-projects.
http://www.xtremesystems.org/forums/...app_info-files
I started crunching this project again today after failing to get any more POEM tasks. I noticed the app_info I had no longer works because there is now a new version of Help Conquer Cancer. I have it working for AMD. Here it is.
I downloaded that app it run only three tasks but they are faster in deed the problem was that my computer received a message saying:
"This project doesn't support anonymous computers"
Can anyone tell me what is the problem and if I can fix this.
You have to run a app_config.xml file now at the WCG Project, something like this is working for me: It works for either ATI or NVIDIA Cards, adjust to fit your needs, .5 CPU & 8 WCG GPU Wu's works good for me on my 3770k CPU & 7970 GPU's
<app_config>
<app>
<name>hcc1</name>
<gpu_versions>
<gpu_usage>.50</gpu_usage>
<cpu_usage>.125</cpu_usage>
</gpu_versions>
<max_concurrent>8</max_concurrent>
</app>
</app_config>
Thanks STE\/E.
Just an FYI. You must update your BOINC client to 7.0.42 to use the app_config file. If you try it with a lower version you'll get the error message Duke was seeing.Quote:
Originally Posted by STE\/E
:o That is the version I am using for sometime nanoprobe. Thanks for the advice. I have posted on WCG Forum also and Lawrence come into my help. I will restart the computer tonight and see if everything is OK but I think it will be. :)
WCG doesn't want people using an app_info.xml file. In fact, due to a bug in recent versions, BOINC ignores that the platform is not allowed and a fix is in place which will keep future versions from working with it if the project doesn't allow it. The app_config.xml is allowed though, regardless of the platforms allowed by the server, so that's the best option.
Then again, you may be able to include the platform element in the app_info.xml file to get it to work anyway. If that works, don't tell DA!
I'm helping out Team China in the WCG Challenge since our Team isn't entered ... :rolleyes:
Good idea Ste\/e. Team China has been a good friend.Quote:
Originally Posted by STE\/E
Yeha, they are a lot of friends. Very nice to have so many friends.
There was also a bug found in the WCG server code that was allowing the app_info platform to work. The techs have already stated the the code will changed to remove the bug and crunchers will eventually be forced to upgrade their BOINc client to run the app_config set up. I've been playing with it and so far I've found that you may be able to drop an app_config file into any of the project folders and assign cores to run them. I have an i7-860 that is currently running 4 cores on WCG ( 3 on SN2S and 1 on CEP2) and 2 cores on Skynet POGS. Later today I'm going to see if I can get the other 2 cores to run Docking or Lattice or maybe one of each . :-?Quote:
Originally Posted by Slicker
Now that sounds superbly interesting! You mean DA finally gave us a way to circumvent DA's braindead scheduling? So, if I have 8 projects all at equal Resource Share on an 8 Core box, I can force it to only run one task per project so I can get the desired distribution of effort. Rather than having one naughty project over-cache and suck away all the resources as the client determines that's the only way it can make deadline. Leaving the other seven starved of their fair share. :)Quote:
Originally Posted by nanoprobe
It's working, sorta. Instead of Docking or Lattice I'm now using the other 2 cores to run SIMAP. I only have the cache set for 2 days and it doesn't load an equal amount of tasks so each project will get 2 days of work, it just totals them all up for the 2 days. I use Boinctasks so I just suspended the tasks that were in cache that already had 2 days worth and let the other projects fill up with 2 days worth 1 at a time. I don't know what setting the overall cache higher would do but I doubt it would spread out the load among the projects evenly. Even so you could set a higher cache and with a little micro-management fill up each project and forget it for however many days of cache you desire. Not perfect but it's not that hard to manage.(so far :rolleyes:)Quote:
Originally Posted by Mumps
Now I don't have each project with an equal resource share so that's something else I'll have to play with.
Now that it's possible to assign a specific project to x amount of cores maybe there is a way DA can make it so each project gets the amount of tasks per cache setting. Where's the suggestion box? :D
While it's possible to force the client to only run 1 task per application, where's nothing stopping the client from downloading a full cache for this application, meaning your 8-core box has 7 idle cores...Quote:
Originally Posted by Mumps