View Full Version : ATI/AMD card owner debug help

08-02-2008, 06:25 AM
This is a revisit on an old demo I did back in January:

It seems people with some ATI cards (X1600 possibly X1800) were having trouble with my demo. I was hoping some owners of these cards (or just other ATI/AMD card owners) could try the demo again.


If the lighting is still broken, could you do me a huge favor and see if updating this shader file lightingLIDefer.shd with this change on line 174:

lightIndex += 0.5/256.0;

(just before the comment // Lookup the Light position)

and see if this fixes the issue. (Note the file is read only when it is uncompressed - so you will have to unset the read only flag on the file properties)

08-02-2008, 08:02 AM
I could when I get home, but it will be on a 3650? IIRC...

08-02-2008, 12:17 PM
Does you application compatible with linux? I can't compile it with the makefile you provide.
Some errors like this:

../Framework3/OpenGL/OpenGLApp.h:23:38: erreur: X11/extensions/xf86vmode.h : file not found

I get some errors with your framework too.

I'd glad to test it as a poor owner of a Radeon X1600 mobility. :)

08-03-2008, 12:18 AM
Looks fine on 4850

08-03-2008, 02:54 AM
Thanks everyone so far.

@ dletozeun - In theory I think it should compile on Linux as I don't think I do anything special - and I use the http://www.humus.ca/ framework which does compile on linux. However - I have no current means of updating the makefile for testing. If you cannot, I will see what I can track down.

08-04-2008, 04:11 PM
What do you want me to look for? I ran it on the Gf's Radeon 3650 under XP with latest 8.6 or 8.7 drivers and it ran fine... But the lights have a circle around the other perimeter maybe the textures are supposed to be like that?

08-05-2008, 03:44 AM
There is a screen shot on the website if you are wondering what it is supposto look like:

(Screenshot is from the static lights option)

If it looks much different from this, you could email me or post a screen shot. (email is in the readme.txt)

08-05-2008, 04:20 AM
It seems to be ok on Radeon 3870 X2 with 8.7 drivers. In the precision test there are two horizontal red lines, probably on GPU boundaries.

08-05-2008, 01:20 PM
Here are some SS of what I see. First one is the non modified version and the second is the change you wanted done...

http://img511.imageshack.us/img511/6002/notmodifiedhq1.th.jpg (http://img511.imageshack.us/my.php?image=notmodifiedhq1.jpg)

http://img229.imageshack.us/img229/6374/41504214zz3.th.jpg (http://img229.imageshack.us/my.php?image=41504214zz3.jpg)

08-05-2008, 04:52 PM
The error seen by Mars_999 is very similar to this (http://www.opengl.org/discussion_boards/ubbthreads.php?ubb=showflat&Number=242629#Post242629)

08-07-2008, 08:08 PM
Thanks for that - looks like I will be putting a note about ATI in the readme.txt (at least until the driver bug is fixed)

08-08-2008, 03:49 PM
Best of luck! I would bet GL3.0 drivers will be running better than current ones... Sad.

08-09-2008, 06:34 AM
@ dletozeun

I have uploaded a new version that I tested compiling on Linux.
The missing include errors you are getting are probably caused by you not having the X11 development headers installed.

The new version link:

08-09-2008, 12:53 PM
Thanks sqrt[-1] but this not even better. For X11 you are right. In fact I had the X11 dev packages installed but not extensions one... so I had fun to find the good one.

Then, I have many errors with the framework and I don't talk about warnings (that should be due to last compiler, I had more warnings with my old projects too).

I have errors especially in OpenGLExtensions header, something like invalid use of GLvoid, I don't understand. So I tried some very bad hardcore things, integrating GLEW to load opengl extensions and remove the recalcitrant lines in the OpenGLExtensions header.

This resolved this problem and now I have problems with the App.cpp at this line (894):

glLoadTransposeMatrixfARB((GLfloat*)projectionMatr ix);

As you see I tried to cast to GLfloat* since I had a conversion error from float* to GLfloat* (Yes this is weird). But In fact I realize that projectionMatrix is a vec4? Which looks like GLSL data type. I don't know how the framework manage shaders, but it looks like you write function directly in the source file in a GLSL syntax...

So, sorry, I did went more further waiting the 1.3 version :)

08-09-2008, 04:56 PM
Thanks, for all your efforts in trying - I did not expect you to do anything other than type "make" and say - "I get these errors".

My Linux development experience is very limited - but all I know is I can download the above zip, type "make" in the deferred lighting directory and it compiles and works. I am using PCLinuxOS 2007 so perhaps I have made some assumption somewhere?

That GLfloat->float cast error makes no sense to me - the mat4 type is from the Humus framework that has a float conversion operator. It almost seems like you do not have the OpenGL headers installed - but that makes no sense at all.

I might post on the Linux forum and see if any other Linux people can point out the error.