On 13/01/2025 14:19, Ilya Maximets wrote:
> On 1/13/25 11:45, Roi Dayan wrote:
>>
>>
>> On 10/01/2025 19:38, Ilya Maximets wrote:
>>> On 1/10/25 18:35, Ilya Maximets wrote:
>>>> On 1/9/25 12:35, Roi Dayan wrote:
>>>>> From: Gaetan Rivet <[email protected]>
>>>>>
>>>>> Count the number of updates done (insertion + deletion) to each caches.
>>>>>
>>>>> Signed-off-by: Gaetan Rivet <[email protected]>
>>>>> Acked-by: Roi Dayan <[email protected]>
>>>>> ---
>>>>
>>>> Hmm. What's the point of counting if these counters are never read?
>>>>
>>>
>>> Also, did you run some kind of performance testing to see if these
>>> counters affect performance in any way? They are all on a hot path,
>>> so performance evaluation is necessary. We'll need at least a few
>>> words about performance impact in the commit messages.
>>>
>>>> Best regards, Ilya Maximets.
>>>
>>
>> Hi Ilya,
>>
>> We planned to use the counters from another framework/module we wanted
>> to add that will expose the counters to investigate performance but it
>> was not done yet and we had these standing commits and thought others
>> might benefit from having the preparation commits. We can send them
>> again later with the bigger series when ready.
>
> OK. I see. It's better to include the changes together with the code
> that is using them.
>
> Other counters are being reported via pmd-perf-show command, so adding
> a section there might also be an option.
>
> Best regards, Ilya Maximets.
We thought about adding to pmd-perf-show after you pointed this but we also
don't have specifics about impacting more than minor the performance.
We prefer to revisit this later when we have a more complete series and after
some basic testing.
Thanks,
Roi
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
