In fact, the line breakdown is as following:
wc -l if_wm*.{c,h}
    3238 if_wm.c
    5051 if_wm_hw.c
    2099 if_wm_hw.h
     134 if_wm_osdep.h
     433 if_wmreg.h
   10955 total

if_wm_hw.c and if_wm_hw.h are dedicated to tuning and represent around =
7100 lines. I think it speaks for itself. The *generic* part of the =
driver is around 3700 lines.

Again, the source code is available from FTP with anonymous access at =
216.170.216.166 If some people find it useful, please send feedbacks, =
otherwise just ignore it.

Pascal.

-----Original Message-----
From:	Thor Lancelot Simon [mailto:tls@rek.tjls.com]
Sent:	Wed 12/17/2003 2:08 PM
To:	Pascal Renauld
Cc:	tech-kern@netbsd.org
Subject:	Re: A new wm driver
On Wed, Dec 17, 2003 at 01:31:59PM -0500, Pascal Renauld wrote:
> Well, it's just that we had some issues with specific hardware and =
felt=20
> that the hardware was much better supported in the em driver. The =
current=20
> NetBSD driver is 3500 lines against 11400 for the one coming from =
FreeBSD.=20

3X the total line count, in a device driver, sounds to me like a pretty
persuasive argument *against* using the larger driver.  ;-)

> The lower layer part of the driver (5000 lines) is dedicated to tuning =
of>=20
> the different boards.

I'm sure there's some work to be done here, because I've already noticed
significant performance differences between, for instance, the PCI-X =
versus
the standard PCI adapters in the same machine, seemingly to do with =
their
default tuning for different PCI bus configurations.  But I can't =
imagine
that that could possibly justify a vastly more complex and enormously
larger driver that doesn't actually provide any features the current
driver doesn't also provide...

Thor