I'm impressed by the improvement. I didn't really expect any.
On Tue, Jul 13, 2021 at 09:19:34AM +0100, Anton Ivanov wrote:
> I ran the revised patch series (v2) which addresses your comments on the
> ovn-heater benchmark.
>
> With 9 fake nodes, 50 fake pods per node I get ~ 2% reproducible
I ran the revised patch series (v2) which addresses your comments on the
ovn-heater benchmark.
With 9 fake nodes, 50 fake pods per node I get ~ 2% reproducible improvement.
IMHO it should be even more noticeable at high scale.
Best regards,
A
On 09/07/2021 19:45, Ben Pfaff wrote:
On Fri,
Good point,
I forgot about the rcu.
I will redo the patch as per your suggestions and resubmit it next week.
A.
On 09/07/2021 19:45, Ben Pfaff wrote:
On Fri, Jul 09, 2021 at 06:19:06PM +0100, anton.iva...@cambridgegreys.com wrote:
From: Anton Ivanov
If we are not obtaining any useful
On Fri, Jul 09, 2021 at 06:19:06PM +0100, anton.iva...@cambridgegreys.com wrote:
> From: Anton Ivanov
>
> If we are not obtaining any useful information out of the poll(),
> such as is a fd busy or not, we do not need to do a poll() if
> an immediate_wake() has been requested.
>
> This cuts out
From: Anton Ivanov
If we are not obtaining any useful information out of the poll(),
such as is a fd busy or not, we do not need to do a poll() if
an immediate_wake() has been requested.
This cuts out all the pollfd hash additions, forming the poll
arguments and the actual poll() after a call
