Re: [RFC v2] perf: Rewrite core context handling

[Ravi Bangoria]

On 24-Aug-22 12:57 PM, Peter Zijlstra wrote:
> On Wed, Aug 24, 2022 at 10:37:36AM +0530, Ravi Bangoria wrote:
> 
>>> Now, I suppose making that:
>>>
>>>   {-1, NULL, NULL}, {cpu, NULL, NULL}
>>>
>>> could work, but wouldn't iterating the the tree be more expensive than
>>> just finding the sub-trees as we do now?
>>
>> pmu=NULL can be used while scheduling entire context. We can just traverse
>> through all pmu events of both cpu subtrees.
> 
> But imagine the case where we have 50 event for a PMU that can only
> schedule 8. Then we have to iterate 42 events for naught instead of
> directly jumping to the next PMU.

Yes, that needs to be handled. And, IIRC, you proposed maintaining a list
of leftmost event from each pmu subtree.

Thanks,
Ravi