Stereoscopic rendering (red-cyan anaglyph)

[MirceaKitsune]

I was wondering if the engine has any support for stereoscopy, and how to enable it if so. Nothing fancy like the Oculus Rift (yet), just a basic ability to render from two different viewpoints with a given eye separation, in either side-by-side mode / interlaced mode / red-cyan anaglyph / etc. I have a pair of red-cyan glasses, which I love using with games that support anaglyph stereoscopy (like Minecraft)... I can only imagine how TDM's environments would look like with full depth perception added to the mix

[HMart]

TDM has no support for stereoscopic 3D of any kind, Doom 3 BFG on the other end has it, but TDM uses the older Doom 3 engine before BFG. Btw i also have a 3D screen and i play many games in 3D mode, unfortunately, OpenGL games are not supported by Tridef 3D, the software i use.

[MirceaKitsune]

TDM has no support for stereoscopic 3D of any kind, Doom 3 BFG on the other end has it, but TDM uses the older Doom 3 engine before BFG. Btw i also have a 3D screen and i play many games in 3D mode, unfortunately, OpenGL games are not supported by Tridef 3D, the software i use.

 

Ahhh... that sucks. I was half expecting the original idTech 4 engine to support this, or TDM to have added it later. Since I assume this isn't on any immediate priority list, I guess the answer is a definite no then.

[nbohr1more]

Some folks have ported the Occulus stuff to vanilla Doom 3 as I recall, so it's within the realm of possibility...

 

The bigger problem is, can you get the bigger missions to run at 120 FPS needed for the proper VR experience ... ?

[Xarg]

But but frame rate doesn't matter

 

At least, according to Microsoft.

[STiFU]

Some folks have ported the Occulus stuff to vanilla Doom 3 as I recall, so it's within the realm of possibility...

 

The bigger problem is, can you get the bigger missions to run at 120 FPS needed for the proper VR experience ... ?

There is always the option to get a proper 3d screen without those crappy shutter glasses. ;-)

[MirceaKitsune]

I think lower FPS is understandable and acceptable, since you are rendering two different views. We could at worst use this on heavily optimized maps where framerate remains playable.

 

Anyway, stereoscopy is something most engines support these days. The Darkplaces engine even has a lot of modes and options, and I could tweak eye separation and angle to get the best possible effect. So although this is the question every user with a wish might be afraid to ask: Is there a chance that someone might code it?

[SteveL]

I don't think so. It wouldn't work very well without huge changes to the engine.

 

You (or someone) could have a go at a half-fps version without touching the engine. The way the player view gets set up every frame is pretty straightforward, and it's in the game code. You could easily modify it to draw alternate frames from alternate eye positions, if you could then get a 3d monitor to display that in sync with those glasses that block out one eye each frame (showing my ignorance here )

[nbohr1more]

Here's the Occulus thread about Vanilla Doom 3 I found on Reddit:

 

https://forums.oculus.com/viewtopic.php?f=29&t=13130&p=183048#p183048

[OrbWeaver]

You could easily modify it to draw alternate frames from alternate eye positions, if you could then get a 3d monitor to display that in sync with those glasses that block out one eye each frame (showing my ignorance here )

 

Manually syncing in time isn't necessary. The way it's done is to create an OpenGL context with quad buffer support (both front and back buffers for each eye), and then render the two eye perspectives into the left and right buffers as appropriate. The graphics driver is then responsible for displaying the two eye buffers in a manner appropriate for the current hardware (e.g. by syncing them in time with shutter glasses, or providing the two images in whatever format the 3D monitor expects).

[MirceaKitsune]

Manually syncing in time isn't necessary. The way it's done is to create an OpenGL context with quad buffer support (both front and back buffers for each eye), and then render the two eye perspectives into the left and right buffers as appropriate. The graphics driver is then responsible for displaying the two eye buffers in a manner appropriate for the current hardware (e.g. by syncing them in time with shutter glasses, or providing the two images in whatever format the 3D monitor expects).

 

I don't know the coding specifics... except when stereoscopy was implemented in the MineTest engine, and this thread was linked as an example of the code. Maybe it can come in handy for us as well?

 

But from what I'm aware, the engine is responsible for creating and displaying the two views, not the driver. Especially since it must decide how to map it: Either side-by-side, interlaced (alternating per pixel line horizontally), or anaglyph (red-cyan). Since most Quake engine forks (<= idTech 3) added this easily, I don't imagine it being problematic... especially if another idTech 4 fork (BFG?) did it and we can simply copy their section of code!

[OrbWeaver]

But from what I'm aware, the engine is responsible for creating and displaying the two views, not the driver. Especially since it must decide how to map it: Either side-by-side, interlaced (alternating per pixel line horizontally), or anaglyph (red-cyan).

 

In theory, the engine should never have to know the capabilities or configuration of the hardware. It just renders the two views into two different buffers and the hardware determines how to display it to the user.

 

In practice, I have only seen this done for shutter-glasses stereo (where manually syncing in the engine would be nearly impossible). Anaglyph and interlaced stereo are handled manually by the application. But there is no reason why this couldn't be done by the graphics drivers just like shutter glasses (it would certainly make application developers' lives a lot easier).