Considering a Doom 3/Chaos Theory game engine with old-school techniques on modern hardware.

[Cyan]

back into Doom 3, this is the version I'm currently playing, an open source version of the game -which is algo open source- that supports DirectX 12 and Vulkan, and adds other modern features.

GitHub - RobertBeckebans/RBDOOM-3-BFG: Doom 3 BFG Edition source port with updated DX12 / Vulkan renderer and modern game engine features

Doom 3 BFG Edition source port with updated DX12 / Vulkan renderer and modern game engine features - RobertBeckebans/RBDOOM-3-BFG

github.com

I talked more about it in the Doom 3 thread.

 

[Davros]

So I finished Industria and found it boring

 

[inlimbo]

Sorry to hear. It's got a lot of a rough edges but the potential is there for something cool in the followup.

Been wondering about sector-based lighting tricks lately. Seems like that's underexplored in rasterized real-time lighting. I know idTech 4's games and the like relied on some sector exclusion tricks and I wonder why more games don't.

Seems like that could've been one solution to single-toned gray and blah real-time solutions pre-SSAO (and I still find SSAO a wanting solution) - boxes defining sectors where a custom ambient light is set. Interiors are defined in sectors with a default lower ambient term than exterior lighting, etc. Feels like that could've been as elegant solution as Sucker Punch's fake SSAO solution for Infamous 2: IIRC rather than use any SSAO they simply gave every dynamic object a rectangularish drop/blob shadow texture that sat flush with whatever object that object was resting on. All the benefits of SSAO with none of the post-process, alpha blending affecting drawbacks. They could even selectively enable them where it made sense so you don't see the indiscriminate coverage you'd see with SSAO.

So what am I missing? Was the multipass rendering necessary to accomplish sector tricks like the above too expense at the time? I imagine there are ways to optimize and sorta fake sector tricks like this using textures too, as in a more complex iteration of Sucker Punch's SSAO trick. In any case how viable would this be for a modern renderer and why don't games attempt this more often? I mean I get that baked lightmaps are effectively texture based "sector darkeners" in traditional rendering. They are a non-uniform precalculated representation of what someone could otherwise accomplish by painting into sectors of the world if they really wanted to do so.

A real-time sector based approximation of this seems very viable and desirable for a real-time lighting solution that doesn't leverage raytracing or lightmaps; and I use approximation lightly as I wouldn't expect any sector occlusion to approach the subtlety of global illumination and bounced lighting.

 

[inlimbo]

Ambient Occlusion Fields and Decals in Infamous 2

Standard methods for ambient occlusion in games are either fully precomputed - baking AO onto vertices or lightmaps - or fully dynamic, i.e. SSAO. Both are useful, but we complement them with a hybrid approach that avoids some weaknesses of...

gdcvault.com

Still wish this had been more widely adopted by the industry for dynamic AO. Completely avoids the problem of SSAO outlining objects like they're flush with a wall even though they're standing meters apart, which even HBAO+ can't completely solve IIRC.

 

[inlimbo]

All with a very basic decal system.

Reminds me of all the "real-time" lighting tricks GTAs 3 through San Andreas pulled using just textures and alpha blending.