[g2.1.pre-62-gfd3128a] Dynamic Light Bugs

[Rachael]

Unfortunately I cannot update my driver any further than it is now. My card is considered "legacy" even though it was released well after Windows Vista. This is the latest driver that it allows me to download for the 4850:
http://support.amd.com/en-us/download/d ... 7%20-%2064

Also I am using no 3rd party stuff - I've cleaned out my GZDoom folder of custom drivers and ICD's.

[Graf Zahl]

Too bad that I have no means to debug that stuff on a system with an old AMD graphics card. It's probably something really stupid but without being able to check nearly impossible to find...

Sometimes I have to wonder how AMD constantly manages to release such shitty drivers. That stuff even works on the Intel HD 4000 in my system with both the GL3 and the GL4 render path.

[FishyClockwork]

*sigh* Though this might be a useless post, I'd just like to say that I too cannot update my driver any further.

As far as AMD is concerned, April 2014 is the latest version for Win XP systems (at the moment of typing). I really wish I could be of some further help.

[Graf Zahl]

This 'legacy' nonsense is another reason why I refuse to buy AMD hardware - aside from the general issues with their OpenGL drivers. It tends to render perfectly good hardware useless.

[shadow520]

This is going to sound really stupid, but I'm not really sure how to do that. If you could let me know what to do works, I will be more than happy to test this for you.

[FishyClockwork]

I'll assume you have a shortcut leading to your gzdoom.exe.
Just right click on the shortcut, click on 'properties' and add at the end of the 'Target' field '-gl3'.
Like so:

Code: Select all

"C:\Path\To\Your\GZDoom\gzdoom.exe" -gl3

Click 'Apply' and then 'Ok'.

Alternatively, you'd have to start command prompt and enter the following

Code: Select all

"C:\Path\To\Your\GZDoom\gzdoom.exe" -gl3

After you start GZDoom either through the shortcut or command prompt, go see if the bug still persists.
Of course once you're done, if you did the shortcut method then you should undo the shortcut change(remove -gl3).

[shadow520]

I followed FishClockwork's instructions and I did not encounter the bug that I did earlier; the dynamic lights are now functioning.

[Graf Zahl]

Ok, that's bad news for some people stuck with older drivers. This really looks like a driver bug with AMD and if they refuse to fix it for 'legacy' drivers there's not really much I can do...

Reminds me a bit of the fog bug that has plagued ATI's drivers in the time when shaders were more a curiosity than a usable feature.

[Rachael]

Well - it worked in 1.9. What changed between 1.9 and 2.1 in the GL3 rendering path? I am aware that older rendering paths have been removed but has the GL3 one really undergone that much change?

[Graf Zahl]

The dynamic lights are a special case.

The old code worked with multiple texture passes, the new one with shaders.

Actually the shader based code is relatively old and always worked perfectly with NVidia drivers. It's just that uniform arrays, both of the direct and buffer kind seem to be broken somehow with older AMD drivers. I was somehow hoping that they got at least the buffer implementation right, but no such luck, it seems.

The texture based lighting code has been removed for GZDoom 2.x, because I thought that all recent hardware would be able to handle uniform buffers as defined by the GL spec. Seems I was mistaken. The real problem here, though, is AMD declaring this older hardware 'legacy' a bit too early and stopping support for it.

I'll be honest, though: Since this works as expected on all NVidia and Intel GL 3.x hardware and on all GL 4.x hardware, if not using Windows XP, I'm not bringing back the texture based lighting, as it'd thoroughly mess up the code.
If someone with some programming skills and an older AMD setup manages to debug this and find a workaround, I'll gladly take it - but I have no means to do any tests myself due to lack of affected hardware.

[NeuralStunner]

If someone with some programming skills and an older AMD setup manages to debug this and find a workaround, I'll gladly take it - but I have no means to do any tests myself due to lack of affected hardware.

And looking at history, we know how (un)likely that is.

[Rachael]

Just tested with a new setup, same system, but different OS

New setup:
Display Drivers: WDDM 1.1 ATI Catalyst 13-4 (Beta)
Operating System: Windows 8

The result is the same.

[Graf Zahl]

I'm not surprised. Without the ability to debug on the affected hardware there's little chance to find out what specifically goes wrong here.

[Graf Zahl]

Thanks to Eruanna I was finally able to debug this on an affected machine. Turned out that AMD did not properly implement the glBindBufferRange function on this older driver (it should imply glBindBuffer but it clearly did not) and had already deprecated some hardware before fixing it so these cards never received a fixed version...

It seems to be corrected in more recent driver versions or the bug would have been far more widespread...