Forums » Bugs

Linux OpenGL 4 lighting artifacts

12»
Dec 05, 2014 cellsafemode link
OS: Linux Debian AMD64 (unstable)
Driver: Open source Radeon Version: 1:7.5.0-1
Kernel: 3.17.0-rc2 (i need to update i know)
Video card : 01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Tahiti PRO [Radeon HD 7950/8950 OEM / R9 280]
glxinfo : https://drive.google.com/file/d/0B6yF3M_pR-ojLXlSWnFkdGM2a1k/view?usp=sharing

Description : Light from sources are coming through models and appearing on the far side. Light occlusion is being handled per-face and not smoothly passing over an object. etc. See video below.

Video : https://drive.google.com/file/d/0B6yF3M_pR-ojV3RsOUVNWTZTU0E/view?usp=sharing

Xorg log : https://drive.google.com/file/d/0B6yF3M_pR-ojeUxlemxueFNqTzQ/view?usp=sharing
Dec 06, 2014 abortretryfail link
Got a ton of artifacts here on my AMD HD6870 w/Linux 3.17.4 Mesa from git a couple weeks ago. This was after starting the game using the GKGL driver then switching to the GL4 driver.

http://imgur.com/a/kJ6qY

This reminds me of what it looked like when the old Intel driver check was disabling shaders. There's no bump maps on stations, dynamic lights, etc.

Extra info:
http://ix.io/60 opengl4err.log
http://ix.io/61 opengl4info.log
Dec 06, 2014 abortretryfail link
Ho'dup. It looks like that only happens when switching drivers while the game is running. If i set it to use GL4, then quit the game and let it re-load fresh, it works great. Images coming up.
Dec 06, 2014 abortretryfail link
Starting the game with Mesa debugging turned on didn't say much. Just this:

[arf@Brimstone ~]$ MESA_DEBUG=1 vendetta
Mesa: User error: GL_INVALID_OPERATION in glUniform(type mismatch)

Here's pictures: http://imgur.com/a/gVqJL#0

Looks great and doesn't run like shit! Though, the ice doesn't seem to be affected by the lighting.
Dec 06, 2014 abortretryfail link
Ion Storms and fog sectors still have some issues though.

http://imgur.com/a/qAVCC#0

This looks pretty rad though.
Dec 06, 2014 raybondo link
Thanks for the reports.
abortretryfail, I don't see any issues with that screenshot ;) it looks fine to me.
Dec 08, 2014 abortretryfail link
I always forget to blow on the cartridge first.

It wouldn't be so bad except for the occasional crash that happens when jumping in to a storm sector. I hope those crash reporters made it through.
Dec 08, 2014 abortretryfail link
This also looks kinda weird.

There seems to be "holes" in the lighting on this asteroid. (object ID 78 in Jallik C-12) No matter what direction you look at it, some of the texels always get rendered black.

http://imgur.com/a/QMWlu#0
Dec 09, 2014 Kierky link
Ship light occlusion?
Dec 10, 2014 mr_spuck link
The storm radar effect doesn't fade out. I'm crashing when jumping in and out of storms too.

Video mode selection is more limited, but IIRC people complained about that with the DX11 driver already.

The Full scene glow setting doesn't seem to do anything anymore.
Dec 10, 2014 raybondo link
arf, those black pixels are caused by the surface normal actually pointing more than 90 degrees away from the lightsource.
pretend that the surface is actually bumpy and the light is right at the surface, causing the far sides of the bumps to be unlit.
Dec 10, 2014 abortretryfail link
Yeah, that's what I expected, but the same bumps are unlit no matter what angle I'm looking at them from. If I move my ship (and the mining beam lights) to the opposite side of the black bumps, shouldn't they be lit?
Dec 10, 2014 Kierky link
The full scene glow effect option keeps reverting to "off". There's no way to enable it under this driver :(
Dec 11, 2014 Kierky link
and that storm crash is pretty damn annoying.
Dec 13, 2014 abortretryfail link
A little alcohol on the card edge and the storm gibberish is gone. Either that or your update fixed it, Ray. ;)
Dec 15, 2014 cellsafemode link
I'm having an issue in testing this long term. I have fps limited to 120 max, and it will sit there pegged like normal and then graphics will lock up for just a moment and then back to max. The frequency of the lockup increases dramatically when near other ships and roids making using the driver in any situation with combat not really practical.

This problem doesn't occur at all with the regular opengl renderer.

Raybondo / Incarnate: is there a particular apitest or other test from the Phoronix test suite we can run correlating to the new features being used in the new renderer that will help shed light on whether the driver is to blame rather than something wrong with the renderer code?

Edit: in particular, if this renderer uses anything that this site's matrix shows up as red for Mesa intel,nvidia,radeon hardware support. That would basically put us in the "use proprietary or stick with old renderer" boat. http://mesamatrix.net/
Dec 15, 2014 abortretryfail link
You know, I've had a similar problem since this patch. It seems like the graphics pipeline just stalls for half a second once in a while. My GPU's fan even spins down when it happens. It seems to only happen when there's "stuff" going on like ships jumping in/out or people shooting at each other.

I haven't been able to nail it down to VO or the driver doing it though.
Dec 16, 2014 abortretryfail link
Sitting in Edras I-8 watching the scaffolding lights blink causes the stall/lockup pretty consistently.
Dec 17, 2014 incarnate link
Cellsafemode: Basically the only extension added here was GL_ARB_ES2_compatibility (although that extension inherently includes a number of features). It's listed as all-green for Mesa (aside from software). Beyond that, the "OpenGL 4" GKGL driver is basically still 2.x.

The only thing that makes it "OpenGL 4.x" is the fact that GL_ARB_ES2_compatibility became mandatory in OpenGL 4.1. But it was optional for a lot of people before then (Mesa obviously, also Intel's own drivers often support 3.x with GL_ARB_ES2_compatibility, etc).

Do you guys see errors appearing in VO's own OpenGL logs?

Anyway, we'll do some more testing on Linux+AMD soon. Our build machine has a 5870 in it, and we have an extra 7970 sitting around somewhere. A total stall sounds like a driver problem, but, we'll see.
Dec 17, 2014 abortretryfail link
I'll check the game's opengl4err.log next time I'm playing on that machine. The funny part is it sort of appears to make the game's sense of "time" hang as well, so when it unjams a bunch of stuff that would have happened during the stall all happens at once. Things like sounds and ship damage and the like.

I wish I had a way to get a snapshot of what's running when it happens, but I don't know of any tools that do that. :(

Other than that periodic stalling, it runs pretty well and looks great. Last week's patch seems to have improved my framerate in addition to fixing the glitchy fog/storm effects.

Edit: My opengl4err.log file only has a few video modes in it like the one cellsafemode posted.