Hi,

On Fri, May 11, 2012 at 3:23 AM, Lauri Kasanen <[email protected]> wrote:
> Hi
>
> I have an experimental branch at 
> https://github.com/clbr/monkey/commits/mallocless that removes all dynamic 
> memory allocation from the hot path. On a typical siege run, this dropped the 
> amount of malloc calls from 400k to 700 :)


It looks good to me, if is thread safe it can be applied, please
provide a standalone patch in the ML..

>
> Will help memory fragmentation and possibly speed. I didn't measure any speed 
> difference, but monkey may be hitting some other bottleneck on my system.
>
> On the other hand, because of the pre-allocation of the client_session pool, 
> idle ram use grows ~2.7mb (with the default ulimit -n = 1024). Even so, I 
> think it's completely reasonable, and since we know the maximum amount of 
> sessions, makes much sense to preallocate instead of doing it anew for each 
> connection. Repeated work for no gain, that.
>

Pre-allocated buffers does not help tooo much, i did it once and the
performance did not grow too much but the code increased a bit...
but.. we can manage a conditional mechanism or flag for that at
compile time... if we gain something lets the user choose..

regards,

> - Lau
> _______________________________________________
> Monkey mailing list
> [email protected]
> http://lists.monkey-project.com/listinfo/monkey



-- 
Eduardo Silva
http://edsiper.linuxchile.cl
http://www.monkey-project.com
_______________________________________________
Monkey mailing list
[email protected]
http://lists.monkey-project.com/listinfo/monkey

Reply via email to