Subject: Re: A new wm driver
To: Jason Thorpe <email@example.com>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
Date: 12/17/2003 13:11:18
>> Did you not bother to read the rest of my message, or something?
>Err, I was agreeing with you. Sorry if my message was poorly worded.
Ah. That's OK, then. I thought you were disagreeing, but I couldn't
see where why. :-/. Sorry if I seemed testy, rather than merely confused.
As I think I said in aprivate email:
I've done nontrivial throughput measurements with the em driver. I
found that (compared to the bge driver or em driver), the em driver is
very good in terms of reducing CPU load on the host; but less good at
filling the wire, particularly on the dual-port version (where I would
like to run both ports with unidirectional 120Mbyte/sec TCP streams.)
I believe Jason's driver does better at the high-throughput
environment, at the cost of somewhat higher CPU load.
I, personally, would not want all the `tuning' of the em driver
enabled on NetBD by default. YMMV. But having knobs to select
`super-cruise' versus `afterburner' modes would be a Very Nice Thing.