Archive > Bugs prior to release 2.3

Radeon HD vs UFO: AI - never ending story...

(1/2) > >>

ncpnc:
Hello there! :)
I'm playing with this game for about a month now, and I'm trying to stick as close is possible to 2.3 branch, so I recompile game myself almost every day with fresh sources ;)
Unfortunately it seems that Raden HD series and this game doesn't like each other. I have seen a lot of graphical weirdness changing day after day until now, and they were always related to glsl and/or shadows:

- invisible earth on geoscape
- black textures
- glowing textures (like shaders on LSD ;)
- shaders disappearing after bazooka hit
- game slowdown with glsl shaders enabled BUT no shaders visible on map (disabling shaders in options tripled FPS without any graphical difference)

And that's not all! Enabling multisampling always make game crash while still in the menu (I have to alt-tab to find some cryptic error mesagebox to find that game actually crash).
So I've just give a try to fresh sources (hey, 2.3 is without -dev, yay!) What I find is:

- still no shaders visible but big fps penalty when enabled
- trying to begin mission with crashed ufo ends with nothing (go back to geoscape)

Are those OpenGL drivers by ATI/AMD so much badly written? :)

Some system spec:
- Phenom X4 9850
- 4GB RAM
- dual HD4850 (CrossfireX)
- Windows Vista x64
- Ati catalyst 10.5 driver
- build: r30523
- game compile flags from make_UfoAI_win32.ini:

--- Code: ---[ufo2map]
cores=15
clean=1
fast=4
mtune_ufo2map=amdfam10
Ox_ufo2map=O3
sse_ufo2map=sse4a
L1ls_ufo2map=L1-line=64
L1cs_ufo2map=L1-size=64
L2cs_ufo2map=L2-size=512
[ufoaisrc]
debug=4
default_UfoAI=4
mtune_UfoAI=amdfam10
Ox_UfoAI=O1
sse_UfoAI=sse4a
--- End code ---

Can I do anything to help track those graphic issues?

Mattn:
you should post your ufoconsole.log

ncpnc:
Attached to my first post.

arisian:
The issue with the failure to load the maps has nothing to do with your graphics card; if you look in the log, it says it can't find the proper tile to assemble the map.  Have you tried re-building the maps? 

The issue with models not being displayed properly is much more likely to be caused by the rendering code interacting with the ATI drivers; can you be more precise about what you're seeing there?  As usual, there's nothing in the logfile that indicates any sort of error.  If the only issue is that you can't easily see a difference, that may just be due to the fact that the changes are relatively subtle at the moment.  The most easily visible change is the soft-glow around the edges of emissive sources; for example, the bright green parts on the UFOs should appear to "glow" with GLSL and Postprocessing both enabled.

There's a much bigger visual difference in the renderer_work branch, but that code is nowhere near ready for prime-time yet; hopefully the 2.4 release will have pretty real-time shadows and lighting.

Mattn:
that was an issue in oriental.ump - it's fixed with r30538 - feedback is welcome

Navigation

[0] Message Index

[#] Next page