>-----Original Message----- >From: Martin Wilck [mailto:[EMAIL PROTECTED] >Sent: Tuesday, February 26, 2008 6:28 AM >To: Pallipadi, Venkatesh; [email protected] >Subject: Re: [Discuss] _CSD support > >Pallipadi, Venkatesh wrote: > >> _CSD is the way BIOS coomunicates the C-state dependencies >across cores >> in the system. The way _CSD is defined, OS communicates with >BIOS about >> its capability through _PDC or _OSC and BIOS can then return methods >> like _PSD and _CSD. >> _CSD is not supported in Linux right now. >> Having said that, _CSD is not really interesting on Intel >CPUs at this >> time as C-states are hardware coordinated. That means, OS >can make each >> CPU enter C-state independently (assuming each core is >independent) and >> hardware will do the required coordination underneath. > >Yes, but from what I gathered, power savings will only be significant >(at least on Quad CPUs) if the OS puts all cores in a package to sleep >at the same time. sched_mc_power_savings=1 tries to do exactly that - >but how does Linux determine which Cores belong to a package, >if _CSD is >unsupported? Just by sibling/phys core ID ? >
sched_mc_power_savings determines things based on phys core ID. On the processor discussed here, there is no difference with cores in the package going into C-state. The package level P-state is the one that helps as frequency/voltage of all cores in the package will be the minimum requested by any core in that package. That's where sched_mc_power_savings help. Thanks, Venki _______________________________________________ Discuss mailing list [email protected] http://mail.lesswatts.org/mailman/listinfo/discuss
