Sorry, I didn't think this through. The bottom line is that there can't be a generic OpenCL plan class. If you have an OpenCL app that can use either NVIDIA or ATI, you must create two separate app versions, one for NVIDIA, one for ATI, with the existing plan class names.
Furthermore, your application has to call boinc_get_opencl_ids() to see which GPU instance it must use (this info is passed from the client via the init_data.xml file). The reason for this: The client needs to know exactly what resources a job will use before it starts the job. So each app version is associated with a particular resource usage (e.g., 1 NVIDIA GPU and .5 CPUs). This doesn't allow app versions that sometimes use an NVIDIA GPU and other times use an ATI GPU. -- David On 01-Dec-2011 7:32 AM, Timo Strunk wrote: > Hi everybody, > > I implemented a generic OpenCL plan class. We talked about it on the > projects mailing list here: > http://lists.ssl.berkeley.edu/mailman/private/boinc_projects/2011-November/009132.html _______________________________________________ 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.
