Re: [Dri-devel] radeon dri + software suspend/resume [THE PATCHES]

[Michel <mi...@da...>]

On Thu, 2002-07-25 at 23:15, Charl P. Botha wrote:
> On Thu, Jul 25, 2002 at 02:11:29PM -0700, Jens Owen wrote:
> > Can somebody fix the ATI VT switch rather than apply this hack that  do=
esn''t
> > support running 3D apps?
> >=20
> > I'll follow up next week, after  SigGraph is over.
>=20
> What do you mean "fix the ATI VT switch"?  If you mean the crash: we've
> fixed that.  This hack (which has nothing to do with the VT switch crash
> fix) is to support suspending to and resuming from disc. There's (at the
> moment) no other way of doing this.

Actually, I think this is a great change of which being able to suspend
is only a side effect:

- this changes 'direct rendering possible' from a property which remains
constant over a server generation to one that can change over VT
switches. This allows to have DRI on several servers. Only one at a
time, but it's only useful on the currently 'active' one anyway. The
others fall back to software rendering.

- another bonus from this is that the kernel resources are unused when
you switch to console or a non-DRI server, so you can replace them,
which allows for testing different versions with a running server. It
doesn't currently survive the DRI re-init failing in EnterVT(), that can
hopefully be addressed, keeping DRI disabled until the following switch.


All in all, this looks to me to be a great idea which has the potential
to obsolete some of my habits to work around the current situation. (One
drawback being that the recently discovered ability of gdm to manage
several displays would no longer be needed ;) While it has loose ends as
is, let's work on tying those up so we don't have to describe it as a
'kludge' or 'hack'.


Now I just hope I haven't overlooked anything which renders this merely
wishful thinking...


--=20
Earthling Michel D=E4nzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast