Subject: Re: interrupt sharing improves performance?
To: Jason R Thorpe <thorpej@wasabisystems.com>
From: David Laight <david@l8s.co.uk>
List: tech-perform
Date: 02/03/2003 22:48:37
> It does require some changes to the network drivers, however, in order for
> it to actually work.  Right now, NetBSD's network drivers check an interrupt
> status register for the device, and return "interrupt not for me" without
> doing additional work if their interface was not the source of the interrupt.

That cycle is a waste of time, you might as well look to see if
there is an rx data (etc), the driver will have to look there as well.
For ring based drivers the ring structure itself can be polled.
This will be much cheaper than an access to the chip.

I got a massive benefit on one ethernet driver by almost never
taking the 'end of transmit' interrupt.  You can 'reap' the
transmit descriptors during tx setup.  Only packet blasting needs it
(oh and when the sending code isn't goimg to do anything else until
you free it's buffer, On that system only NFS did that to us.)

	David

-- 
David Laight: david@l8s.co.uk