Just for reference it is also worth looking into the built-in "mt" app_plan - in case you don't want to roll your own scheduler. Other built-in plans include: ati, cuda, nci, sse3
-- Janus On 2010-10-29 06:34, David Anderson wrote: > You need to tell BOINC that the app uses multiple cores: > http://boinc.berkeley.edu/trac/wiki/AppMultiThread > > Otherwise it thinks it's single-threaded and starts one job per core. > > -- David > > On 25-Oct-2010 4:38 AM, Olivier Romand wrote: >> Hey people, >> >> First of all Id like to introduce myself. Im Olivier ROMAND, fresh graduated >> student and now working for Renderfarm.fi. Im pretty new in the BOINC world >> and must say that Im quite impressed by the possibilities of the whole >> solution. >> >> Renderfarm is a small but growing project that uses BOINC to provide an Open >> Rendering Environment. We are supporting the rendering engine Blender, also >> open source. >> >> Enough talking for now. I have a small problem and I can't understand wether >> I understand wrong or if my settings are incorrect. >> >> I would like to tell server side that the clients have to process 1 wu at a >> time, using ALL the cores, rather than using 1 core/wu. This problem isn't >> really new since I found many sources dealing with this issue: >> >> -http://www.archivum.info/[email protected]/2010-06/00008/Re-(boinc_projects)-1-WU-per-machine-not-per-CPU-core.html >> >> > -http://boinc.berkeley.edu/trac/wiki/ProjectOptions#Joblimitsadvanced >> In our case, this is pretty important since rendering is quite a heavy >> process. For example if a machine has 4 cores with 4Gb of RAM, it would only >> be able to use 1Gb per CPU rather than 4Gb for the all WU (blender supports >> multithreading). >> >> I've been playing around with the configuration file config_aux.xml but it >> results that my client (still) launches 1Wu/core ... >> >> I wonder if this a normal behavior ? Would you guys know more about that >> matter ? In particular could you take a look at my config file ? >> >> The version of the server used is the stable branch and the config_aux.xml >> file used is as follow: >> >> <config> <max_jobs_in_progress> <!-- We are interested to send many wu in >> advance .. --> <project> <total_limit> <jobs>5</jobs> </total_limit> >> </project> >> >> <app> <app_name>blender</app_name> <!-- ... but we want to process 1 wu for >> this specific application --> <total_limit>1</total_limit> </app> >> >> </max_jobs_in_progress> </config> >> >> On the client I have set 100% of the processors but I still get 1 wu/core. >> >> Would you be able to light up my lantern ? >> >> Thanks for your support >> >> Olivier ROMAND renderfarm.fi _______________________________________________ >> boinc_dev mailing list [email protected] >> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, >> visit the above URL and (near bottom of page) enter your email address. > _______________________________________________ > boinc_dev mailing list > [email protected] > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev > To unsubscribe, visit the above URL and > (near bottom of page) enter your email address. _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
