rerun of the classical libev benchmark against libevent 2.0.10

Marc Lehmann schmorp at schmorp.de
Sun Jan 30 17:52:19 CET 2011


On Mon, Jan 10, 2011 at 07:57:29AM +0100, Marc Lehmann <schmorp at schmorp.de> wrote:
> I just had a few minutes of free time and re-did the benchmark against
> libevevent-2.0.10.
> 
> Seems the epoll fixes they incorporated brought libevent 2 performance on
> the same level as libevent 1., while the same thing lead to libev 4 being
> slightly faster then the version i benchmarked in 2008 or so.
> 
> http://data.plan9.de/2dat.t0.png
> http://data.plan9.de/2dat.t1.png

Just a F'up: indeed, the performance of libevent 2 is slightly worse
than libevent 1. The reason why the performance was better in the alpha
versions was that libevent 2 has two epoll backends, one that breaks
existing code, which is faster, and which was the only backend available
in the alpha versions.

The release versions of libevent2 apparently use the "safe"
backwards-compatible model again. The reason that libevent 2 is slightly
slower then 1 is therefore that it uses basically the same algorithms and
implementation as libevent 1, with lots of added mutex locks.

Since libev has gained speed since the libevent alpha benchmarks I am
confident though that it beats even the experimental "unsafe" backend - of
course, if anybody wnats to verify, feel free (I am too lazy to find out
how to get the alternative non-default backend).

-- 
                The choice of a       Deliantra, the free code+content MORPG
      -----==-     _GNU_              http://www.deliantra.net
      ----==-- _       generation
      ---==---(_)__  __ ____  __      Marc Lehmann
      --==---/ / _ \/ // /\ \/ /      schmorp at schmorp.de
      -=====/_/_//_/\_,_/ /_/\_\



More information about the libev mailing list