On Mon, 3 Mar 2003 17:11:29 +0000
David Laight <david@l8s.co.uk> wrote:

> > rtk(4) appears to be very reasonably implemented to take advantage of
> > some useful ethernet chipset features. I don't know where you get the
> > idea that it is a hack designed to benchmark well.
> 
> IMHO the 'early TX start' and 'early RX interrupt' are not 'useful
> ethernet chipset features'.  They are designed soley to improve
> latency at a cost of cpu cycles, this may be appropriate to the
> desktop, but are not relevant to servers.
> 

unless the server is part of a compute cluster. Depending on the
applications being run, EP vs non-EP, low latency in the internal
interconnect is sometime very important. But just as often is not as
important as adding more cpu's instead of spending money on better
interconnects. Admittedly, this may be a fringe group of user, but it is
something to think about if we do in fact make a change that _significantly_
impacts ethernet latency for such common hardware. It's a fringe group,
AFAIK dominated by linux, so it may not matter anyway. 

http://www.nas.nasa.gov/Research/Reports/Techreports/1997/nas-97-023-abstract.html

-- 
Harry Waddell
Caravan Electronic Publishing
-----------

"The louder he spoke of his honor the faster we counted our spoons." - Ralph
Waldo Emerson