OpenMW Arch Linux ARM Build

[Time4Tea]

Looks like you are just using the CPU and not your GPU.

I can get 25fps outside with a RPi2 @ 1080p

Ok. Yes, I am suspecting I may not have the correct X11 drivers installed for the Mali GPU. What CPU/GPU is the RPi2 using?

Regarding the videos: they seem to play fine in smplayer, so it doesn't seem to be a general issue with ffmpeg.

[BaronPampa]

https://forum.pine64.org/showthread.php?tid=8665&page=7

A patch for the red sky is mentioned in this thread.

[Time4Tea]

I have done a bit more digging on this and found some interesting things. glxinfo -B gives the following:

Code: Select all

$ glxinfo -B
name of display: :0
display: :0  screen: 0
direct rendering: Yes
Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Panfrost (0xffffffff)
    Device: Mali T760 (Panfrost) (0xffffffff)
    Version: 20.0.4
    Accelerated: yes
    Video memory: 4002MB
    Unified memory: yes
    Preferred profile: compat (0x2)
    Max core profile version: 0.0
    Max compat profile version: 2.1
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 2.0
OpenGL vendor string: Panfrost
OpenGL renderer string: Mali T760 (Panfrost)
OpenGL version string: 2.1 Mesa 20.0.4
OpenGL shading language version string: 1.20

OpenGL ES profile version string: OpenGL ES 2.0 Mesa 20.0.4
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 1.0.16

which suggests that hardware acceleration on the Mali is enabled ('Accelerated: yes'). I tried disabling the panfrost kernel module and ran OpenMW again. The issue with the videos is still there, but there is no problem with the textures - all the textures look fine and there was no need to use the 'texture-decompress' option. However, the FPS was truly abysmal at around 1.

So, it seems the GPU acceleration was being used in the previous run; however, I think support for the Mali GPU is still being developed and perhaps the OpenGL support isn't quite there? The second run seemed to have much better OpenGL support, but obviously terrible performance, as it was just running on software.

I guess OpenMW isn't playable on armv7 Chromebooks just yet, although hopefully they will improve support with future versions of the driver.

[psi29a]

OpenGL version string: 2.1

Looks like the equivalent of RPi's VC4 GPU. Yeah, all things considered, it could be that the driver/module isn't as performant yet. Keep an eye on development though. Perhaps get in contact with the developers? If you can show that the LLVMPipe rendering renders correctly while their driver shows a red sky for example; perhaps they can fix it?

[AnyOldName3]

There's stuff in there about desktop GL, but I thought we were discussing getting things working through gl4es earlier?

[Time4Tea]

OpenGL version string: 2.1

Looks like the equivalent of RPi's VC4 GPU. Yeah, all things considered, it could be that the driver/module isn't as performant yet. Keep an eye on development though. Perhaps get in contact with the developers? If you can show that the LLVMPipe rendering renders correctly while their driver shows a red sky for example; perhaps they can fix it?

Ok, I will try to reach out to them and make them aware of the issue. What minimum version of OpenGL does OpenMW require?

Also, about the videos, what library is the engine using to play them?

[Time4Tea]

https://forum.pine64.org/showthread.php?tid=8665&page=7

A patch for the red sky is mentioned in this thread.

Thanks for the link, I'll take a look at it. Although, as I mentioned in my other post, I suspect the issue is with the GPU driver, which is still under development.

There's stuff in there about desktop GL, but I thought we were discussing getting things working through gl4es earlier?

I've been looking at getting OpenMW working on the hardware, using whatever method is available. I thought we had established earlier in the thread that OpenMW doesn't/shouldn't work on GLES anyway?

[psi29a]

Ok, I will try to reach out to them and make them aware of the issue. What minimum version of OpenGL does OpenMW require?

OpenGL 2.1 which is what the RPi's VC4 is capable of, and yours too I believe.

[BaronPampa]

Regarding the patch - I know that it's used by Android guys to make the red sky look normal. It should have the same effect on your chromebook. OpenMW also has the red sky issue on the Pinebook Pro, an Arm laptop. It's confirmed on the forum that patch solves the issue. All it does is disabling ShadowManager::disableShadowsForStateSet function, whatever it's responsible for. Maybe red sky could be patched in the OpenMW somehow? I suppose that AnyOldName3 would need to take a look.

I've dug it up:
https://raw.githubusercontent.com/xyzz/ ... erge.patch

[Time4Tea]

Regarding the patch - I know that it's used by Android guys to make the red sky look normal. It should have the same effect on your chromebook. OpenMW also has the red sky issue on the Pinebook Pro, an Arm laptop. It's confirmed on the forum that patch solves the issue. All it does is disabling ShadowManager::disableShadowsForStateSet function, whatever it's responsible for. Maybe red sky could be patched in the OpenMW somehow? I suppose that AnyOldName3 would need to take a look.

I've dug it up:
https://raw.githubusercontent.com/xyzz/ ... erge.patch

Ok, I will give the patch a try and see if it helps. Although, the red sky isn't the only issue. Many of the other textures are also redder than they should be (for example the map at the bottom-right of my screenshot). Also, there seems to be an issue with the tree textures that have transparent areas, where the sky seems to be showing through, but not objects or terrain that should be in the background.

I think the underlying issue is with the GPU rendering, rather than with OpenMW, since the game renders fine with a software renderer. So, patching OpenMW might help, but it seems like it's not treating the true cause.