from what I remember it's where the message body gets created.

The packet impl buffer should be pooked while the message not (at least not
yet)

I want to do another major improvement next release where I use ref counts
on messages and also keep them pooked.  But paging was a bit issue now.


I can take a look tomorrow. I'm saying all this from memory.  I will need
to review the code if there is any duplication not necessary now.

Using the iPhone.


Clebert



On Tue, May 9, 2017 at 8:04 PM Michael André Pearce <
[email protected]> wrote:

> Hi
>
> During some testing we decided to turn on a jvm profiler to see what
> allocations were occurring under load causing gc pressure.
>
> We see approx 25% (it varies obv over time) is an on heap byte[] caused by
> copyMessageBuffer method in PacketImpl.
>
> Is there a reason why we create a new buffer and copy the data causing a
> new on heap byte[]
>
> Cheers
> Mike
>
> Sent from my iPhone

-- 
Clebert Suconic

Reply via email to