Hello Alan,

Did you have time to check the below?

Regards,
Adel



On Thu, Nov 3, 2016 at 2:53 PM +0100, "Adel Boutros" 
<[email protected]<mailto:[email protected]>> wrote:

Hello Alan,


Issue created and test case attached in the JIRA. You need to have a running 
broker for the code to run. The functor just closes the connection and it takes 
1.15 seconds for the "on_connection_closed" to be called. The timer is started 
when the functor code is called.


https://issues.apache.org/jira/browse/PROTON-1339


PS: If the fix is patchable and appliable on 0.14.0, I would like to apply it 
locally.


Regards,

Adel

________________________________
From: Alan Conway <[email protected]>
Sent: Wednesday, November 2, 2016 6:32:00 PM
To: [email protected]
Subject: Re: [Proton-c] [0.14.0] [0.15.0] Event_loop injection is very slow 
compared to timer schedule

On Wed, 2016-11-02 at 17:28 +0000, Adel Boutros wrote:
> Hello,
>
>
> In Proton 0.12.2, we were able to inject events from other threads to
> the messaging_handler thread by calling timer.schedule indefinitely
> and queues storing messages to send.
>
>
> With Proton 0.14.0, We have implemented
> proton::thread_safe::event_loop() which should replace the above
> timer.schedule.
>
>
> We have noticed our unit tests are 1 000 times slower with the event
> loop mechanism

Yikes! Please open a JIRA with a test program, I will look at it right
away. I am currently working on improvements to the underlying IO
driver so this is well timed. Better performance metrics is also a hot
topic right now.

> Are you aware of such performance regression? How can we analyze it?

Probably not a regression but just a foolish bit of code in the inject
path, I imagine it will be easy to find and fix with such a large
delta. I'll try to get it fixed for the next release.

Cheers,
Alan.

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to