Please add ability to exclude particular GPU per application basis in same way as it currently done on more global level. This will allow users to partition their hardware resources in most effective manner. Current GPU hardware, being merged in single class (let say NV or ATI) actually has very different compute abilities inside this class. Possibility to exclude device on per application basis will allow to use slow device for app with short tasks and fast device for longer tasks that will make host setup more error-protected. As side bonus it will allow new type of app incompatibilities workaround.
Currently such partition can be done only in very clumsy way by installing few separate instances of BOINC client on the same host with anonymous platform mechanism in play. Proposing ability will considerably simplify setup for multi-GPU hosts. -- Raistmer the Sorcerer _______________________________________________ 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.
