Subject: Re: RFC: framebuffer device support in NetBSD
To: Andy Isaacson <adi@hexapodia.org>
From: Martin Husemann <martin@duskware.de>
List: tech-kern
Date: 09/14/2002 00:41:47
On Fri, Sep 13, 2002 at 04:30:07PM -0500, Andy Isaacson wrote:
> It would be fairly trivial to
> implement a "wscons-fb" driver, but you'd not be wanting to do that
Just to make it realy clear - *I* do not want this.
I was only trying to show that the current architecture provides everything
we need - with minor ommisions (ok, if you count the i386 exception and
it's vga text mode cards even some bigger ommisions, but I don't care for
those).
> many important features such as DRI (access to the hardware 3D
> acceleration) depend on the current driver architecture.
Yes and no. There are very ugly bits that could be easily cleaned up (for
example there is realy no reason to scan a PCI bus, or all PCI busses,
just to mmap an aperture and get an io region). But I agree, I count myself
in the camp that says it's ~impossible to create a device independend, yet
powerfull enough abstract interface between kernel and X server that gives
us comparable performance to what can be done now with DRI and the current
interface. If all the effort going into this whole argument would go into
implementing DRI NetBSD would benefit much more.
Martin