Separate CDF Renderer from IBL Shadows and use for realtime filtering

[MiiBond]

I'd like to improve the realtime filtering and prefiltering of IBL's using the CDF maps that I'm using for IBL shadows. To do this, I'm separating the CDF renderer out from the IBL Shadows Pipeline.
Then, as a first step, I'm using these maps to improve the realtime filtering of irradiance.

As a test, here's an extreme IBL with a single sunlight rendered in a raytracer.

Here's how it looks with spherical harmonics:

Here's the old real-time filtering:

And the new real-time filtering:

I still may not have the math quite right, especially for the weighting, but the early results are very promising. I haven't changed the number of samples yet but we should be able to lower them from what they currently are.

TODO:

1. Using importance sampling with the CDF maps should be an option for prefiltering IBL's as well.
2. Should using these CDF maps be a separate option and not directly replace the existing filtering? The main downside is that it requires the CDF maps to be generated first. After that, the speed/quality should always be a win.
3. I haven't modified the WebGPU irradiance function yet because it doesn't seem to be used. What I mean is that, when I modify the irradiance() function in the glsl shaders, it seems to affect the WebGPU rendering as well. I really don't understand why.
4. I haven't tested anything other than diffuse irradiance yet.

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[Popov72]

If we can improve performance AND the quality of real-time filtering, I'm all for it!

However, I think it should be a new path that we can enable/disable, because replacing what we currently have would be a breaking change, and I know that for the commerce sandbox, we should be sure that the new code doesn't break anything => or, at least, we should discuss this internally and decide if we can allow a breaking change here.

I'm also wondering if we could pre-calculate the two CDF textures and integrate them into an .env file? I think we have a version number in the .env files, so it would be easy to support a new format, where these two textures would have been precalculated. This way, the new code path will be automatically selected at runtime when we load these .env files, and we won't have to suffer the generation time.

cc @sebavan

One thing about your PR: is it possible to undo all the formatting changes in the shader code? A lot of the changes only concern formatting (undoing them would simplify reviewing), and I also find the new formatting difficult to read, with types appearing on the next line...

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[MiiBond]

Agreed. I've made it its own code path now. Set scene.useEnvironmentCDFMaps = true

Yes, the CDF maps could absolutely be part of the .env file, along with irradiance. If the prefiltered maps are already created, however, the CDF maps aren't needed anymore except for other features like IBL shadowing.

About the shader formatting, VSCode keeps doing that and I need to remember to disable "format on save" while working on shaders. Do you guys use different settings to avoid this?

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[MiiBond]

I'm still having problems getting the WebGPU version to actually load the WGSL shaders. If I purposely type in bad code in packages\dev\core\src\ShadersWGSL\pbr.fragment.fx, it will still run fine. Putting the bad code in the glsl version results in compile errors.
Is there something special that needs to be done to make Babylon load the WGSL shaders?

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s).
To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[bjsplat]

Snapshot stored with reference name:
refs/pull/15878/merge

Test environment:
https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15878/merge/index.html

To test a playground add it to the URL, for example:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15878/merge/index.html#WGZLGJ#4600

Links to test babylon tools with this snapshot:

https://playground.babylonjs.com/?snapshot=refs/pull/15878/merge
https://sandbox.babylonjs.com/?snapshot=refs/pull/15878/merge
https://gui.babylonjs.com/?snapshot=refs/pull/15878/merge
https://nme.babylonjs.com/?snapshot=refs/pull/15878/merge

To test the snapshot in the playground with a playground ID add it after the snapshot query string:

https://playground.babylonjs.com/?snapshot=refs/pull/15878/merge#BCU1XR#0

[bjsplat]

Visualization tests for WebGPU (Experimental)

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15878/merge/testResults/webgpuplaywright/index.html

[deltakosh]

Fantastic work! Thanks a ton!

[bjsplat]

WebGL2 visualization test reporter:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15878/merge/testResults/webgl2playwright/index.html