Hi Eric
>> In the original design and implementation, we do have a parameter named >> 'hint' to pass "predicted >max idle time" to cpu_idle_enter_cbfn. >> I have removed it from the onepager due to following reasons, >> 1) parameter cstate roughly has the same info. >> 2) There are arguments whether we should extend parameter 'hint' as a >> structure to contain more info >other than "predicted max idle time" for extensibility. >> 3) Our code is based on onnv_92 which has no support for deeper C state and >> tickless timer yet. I >have no idea about the tickless timer interface at that time. >> If tickless timer or C-State project provides some interface to query next >> timer expiration time, we >could pass the predicted max idle time to callbacks. >> > >One thing that's nicer about the predicted max idle time, it that's it's >relatively platform independent. I would really like to see us make this >idle callback work common code to the greatest extent possible, so that >it could be leveraged by the other architectures at some point. On same lines as predicted-max-idle (possibly from tickless idle subsystem) being passed to callbacks, Would it be good to tell some measure of outstanding work to the callbacks? That could give some indication on intr, and also get the last time we went to idle how long did we stay in idle be useful measure to pass around?
