DOOM3 with dri/r200 drivers: first impression

[Pawel S. <paw...@th...>]

Hello,

I thought I would share my impression of running DOOM3 (250SEK on =20
sale!) using DRI drivers (free!).

SUMMARY:
- hardware TCL broken.
- some lighting problems with software TCL.
- few lockups (save frequently!).
- average 10 FPS on AMD XP1800+, Radeon 8500LE, 640x480.

FOLLOWUP:
My goal was to run DOOM3 with a minimal number of changes since
this is a "production" machine. I used kernel-2.6.10-1.741_FC3 and =20
patched xorg-x11-6.8.1.902-1.

Trying to compile current Mesa against xorg-x11-6.8.1.902 is a major =20
PIA[1]. I ended up using mesa-6.2.1 that comes with xorg and port the =20
R200 part of Roland Scheidegger's S3TC patch to this code. See
http://www.theochem.kth.se/~pawsa/dri/ for few more details.

I could confirm that hardware TCL is broken (at least in the version I =20
used) - does anybody know how to fix it?
http://www.theochem.kth.se/~pawsa/dri/tcl-enabled.png
http://www.theochem.kth.se/~pawsa/dri/tcl-disabled.png

Software TCL appears to have some problems, too:
http://www.theochem.kth.se/~pawsa/dri/lighting-problem1.png
http://www.theochem.kth.se/~pawsa/dri/lighting-problem2.png
- can anybody else confirm that?

I could repeatedly get a hardware lockup when standing in certain place =20
and looking at certain point:
http://www.theochem.kth.se/~pawsa/dri/sure-hang.png
The lockup happened twice also when looking at the moving bridge few =20
corridors later.

Pawel
[1] current mesa CVS appears to use a number of constants mostly =20
related to HYPERZ without defining them. Perhaps I should have followed =20
build instructions exactly. I have a patch that gets me through the =20
compilation (it is not something to commit, rather to give an idea =20
where the problems lie). Still, the linking phase stops with some error =20
related to software shaders, I believe.
http://www.theochem.kth.se/~pawsa/dri/mesa-build-changes-20041207.patch

I think the DRI building instruction phase is all nice and great but =20
the whole process is complex and limits the number of people testing =20
the code to hardcore hackers (or perhaps the goal is to reduce number =20
of reporters that have no idea what they are writing about? In a way, I =20
can understand that...:).