Hey! An excellent thoughtful thread! And the answer is...? ...
Nicolás Alvarez wrote: > El Martes 14 Jul 2009 09:41:10 Carl Christensen escribió: >> with boinc credits the simple rule is there's no pleasing anyone. I [...] > Even more: think what would have happened if they had released the GPU app > before doing the *first* CPU optimization (that is, if they had released the > 1h GPU app while having a 4h CPU app), or if they had released the GPU app > after doing both CPU optimizations. > > Is there any good reason for credits to be different depending on what was > released first? Sorry, I've no magic answer. My thoughts go back to that we should award for what amount of system resources are *utilised* (NOT in terms of time, but) in terms of some value that is usefully used. For example, the number of floating point operations used, the number of integer or logic operations used, the amount of storage used, and perhaps others. Then, for what is actually used, award an appropriate amount of cobblestones * "present currency value". (The existing FLOPs count system but extended to cover and value all resource types.) BUT... Also modify the Boinc Client scheduler to balance resource share in terms of the rate of cobblestones awarded (completely independent of time). At the moment, there appears to be a lot of confusion everywhere due to the mishmash of units used. Some units are units of time, some units are absolute counts, some vaguely are variable rates... And the resource share is purely in units of time even though the rewards are NOT counted in time units! So... Can we have a flop 'worth' some value of cobblestones? Take for example a 1GHz x86 CPU (or whatever the reference CPU is). BUT ALSO apportion the other cobblestone values according to the relative performance for such a reference CPU for those tasks (FLOP, IOP, LOP, RAM bandwidth, etc). As further exotic hardware becomes available, the credits arguments will only get ever more extreme until we have a consistent fundamental fix. My thought is that we must have a semantic shift so that what is usefully utilised is rewarded, and not just *time spent* (perhaps busyly uselessly spinning wheels) on whatever hardware. Regards, Martin -- -------------------- Martin Lomas m_boincdev ml1 co uk.ddSPAM.dd -------------------- _______________________________________________ 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.
