Hi Aubrey,

Resending from my old email account which OpenSolaris will forward.  ;-)


This looks very good!

Here is my feedback/thoughts:

1) I would definitely change the names from "pad" to something
   else such as "pur" or "purn".  "PAD" is already used in many places
   for "Power Aware Dispatcher".  Using "pad" for two different
   but related things will be very confusing.

2) cpupm_pad_idling_ops() will likely call a common code function.
   This function will change considerably how it chooses the cpus
   to offline.  The current algorithm (looping sequentially over cpuids
   from 0) will not be a good choice.  It will tend to offline the most
   efficient cpus first or stripe across many cores/sockets.   In order to
   offline the least efficient cpus, we likely want input from NUMA,
   lgroup load average, etc.

3) What is the maximum rate the platform can generate PUR requests?

Regards,
Bill


On 03/30/10 08:22, Dana Myers wrote:
> Hello Aubrey,
>
> On 3/29/2010 11:58 PM, Li, Aubrey wrote:
>> Hi guys,
>>
>> In order to reduce the platform's power consumption, the platform may 
>> direct
>> OSPM to place a logical processor in the idle state where non-processor
>> affinitized work is dispatched. ACPI4.0 introduces the optional 
>> Processor
>> Aggregator device. The Processor Aggregator Device provides a control 
>> point
>> that enables the platform to perform specific processor configuration 
>> and
>> control that applies to all processors in the platform.
>>
>> Here is a prototype to enable Solaris to handle the new ACPI object and
>> event notification. This is also a new mechanism introduced in the 
>> coming NM2.0.
>> http://cr.opensolaris.org/~aubrey/pad_v1/
>>    
>
> Interesting; I'll take a look.  Thanks!
>
> Dana
>
>> Welcome your any inputs.
>>
>> Thanks,
>> -Aubrey
>>    
>

Reply via email to