cURL / Mailing Lists / curl-library / Single Mail

curl-library

Re: [PROF RFC 1/2] profiling of libcurl by curl-loader, steady state

From: Robert Iakobashvili <coroberti_at_gmail.com>
Date: Thu, 19 Apr 2007 11:33:06 +0200

On 4/19/07, Daniel Stenberg <daniel_at_haxx.se> wrote:
> On Thu, 19 Apr 2007, Robert Iakobashvili wrote:
> > This is already decreases the number of gettimeofday() per each recv call
> > for about 20%, which is very good !
> Amazing. I wonder if we can dig up more places in the code where we fetch the
> time without actually using the results (due to conditions similar like the
> one Yang found here),

20 % less here means less 1/5 less gettimeofday calls number per each
recv () call.
and this is also the absolute decrease in the syscalls for the same
number of fetches.

My guess is that saved 1/5 of the CPU share consumed by gettimeofday(), not from
the total CPU consumption.

> Did you measure the speed/performance difference in your app when you changed
> curl-loader from curl_multi_perfom() to curl_multi_socket_action()? If so, can
> you tell us any numbers?

If you mean the improvement from curl_multi_socket () to
curl_multi_socket_action()
for hyper (hiper) mode, it was a 15% improvement and the details were published
here:
http://curl.haxx.se/mail/lib-2007-04/0101.html

If you mean the difference between the modes: smooth (curl_multi_perform())
and hyper (libevent - curl_multi_socket_action() and curl_multi_socket_all ()
combination), I am going to test it.

60K
Could you recommend a way (sharing?) do decrease memory consumption
per client in libcurl?
Thank you.

-- 
Sincerely,
Robert Iakobashvili,
coroberti %x40 gmail %x2e com
...................................................................
Navigare necesse est, vivere non est necesse
...................................................................
http://curl-loader.sourceforge.net
An open-source HTTP/S, FTP/S traffic
generating, and web testing tool.
Received on 2007-04-19