On Wed, Apr 20, 2016 at 8:18 PM, Grigori Goronzy wrote:
> On 2016-04-20 02:20, Nicolai Hähnle wrote:
>>
>> This is just a slight massaging of the patch you sent previously. What
>> happened to the discussion we had about how to do this properly?
>>
>
> This already provides good value as-is and it
On 2016-04-20 02:20, Nicolai Hähnle wrote:
This is just a slight massaging of the patch you sent previously. What
happened to the discussion we had about how to do this properly?
This already provides good value as-is and it is (IMHO) reasonably
clean, so why not include it for the time being
This is just a slight massaging of the patch you sent previously. What
happened to the discussion we had about how to do this properly?
Nicolai
On 19.04.2016 18:13, Grigori Goronzy wrote:
Small IBs help to reduce stalls for workloads that require a lot of
synchronization. On the other hand, if
Small IBs help to reduce stalls for workloads that require a lot of
synchronization. On the other hand, if there is no notable
synchronization, we can use a large IB size to slightly improve
performance in some cases.
This introduces tuning of the IB size based on feedback on the average
buffer wa
