Re: The tcpip layer is really frustrating me

Mike Houston wrote:

> Switching to non-blocking sends and/or threading the network layer is 
> needed to better make use of highspeed layers.  If we can overlap the 
> packing/sort and the network send(s), we should be able to almost double 
> the performance of Chromium, at least over highspeed interconnects like 
> IB/Quadrics/Myrinet.  Obviously, this is a non-trivial exercise.  It's 
> pretty easy to setup non-blocking sends in the tcpip and sdp layers 
> using pthreads, but it's difficult to convert spus to make good use of it.

I'm slightly suprised that chromium has gotten as far as it has with blocking 
i/o.  It's certainly not premature to consider a rewrite of that code.

I'm not sure if threading is or isn't going to be a win on top of nonblocking 
i/o.  You can certainly imagine situations where it wouldn't.  In any case, 
the decoupling required to move to single-threaded O_NONBLOCK io should put cr 
in good shape to shift io to one or more additional threads in future if that 
somehow looks like a good idea later on.

Keith






-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com