LightmapGI: Use Geomerics technique for retrieving irradiance from SH

[BlueCube3310]

Changes the way directional lightmaps are calculated by using the technique described in https://web.archive.org/web/20160313132301/http://www.geomerics.com/wp-content/uploads/2015/08/CEDEC_Geomerics_ReconstructingDiffuseLighting1.pdf

This implementation is based on https://github.com/kayru/Probulator/blob/4a97a2b021eb2ca7ef696f4ddf36ba9a9432cbb6/Source/Probulator/SphericalHarmonics.h#L136-L151

Calculating SH lighting this way has several benefits, namely it avoids overdarkening caused by the resulting irradiance being negative, as well as enhances the contrast in the colors.

Results:

Before	After

Issues:
The contrast seems too high, probably due to incompatibility in the coefficients,
Dimly lit or completely occluded areas exhibit significant artifacting:

TODO:

• Fix the aforementioned issues,
• Implement for Mobile and Compatibility,
• Optimize the code with glsl magic

[Calinou]

I wonder if this technique would make it easier to implement specular lobes for directional lightmaps, as I mentioned in #63016 (comment).

[Calinou]

Testing this PR rebased on top of master 9ee1873, using the MRP from #96114 (review):

Edit (2025-02-10): Updated screenshots to fix overly saturated lightmap colors since #102203 was fixed.

Before

After

The directional aspect looks much closer to real-time lighting now 🙂

This appears to work well to address #102300 too, seeing the directional lightmap's brightness now better matches the non-directional one.

[jcostello]

Does this fix or improve this? #63016

[Calinou]

Does this fix or improve this? #63016

No, this should be implemented separately. I've attempted to implement this in a branch but didn't get it working correctly yet: https://github.com/Calinou/godot/tree/lightmapgi-directional-draw-direct-specular-lobes

[jcostello]

After

In the after, some areas look darker that in the before or the non-directional

[Calinou]

In the after, some areas look darker that in the before or the non-directional

I've updated screenshots to match master (and this PR rebased on master) and the issue appears to be gone now, unless you were referring to something else.

[jcostello]

@Calinou are you sure you updated the images? I still see the issue in your images.

In the static directional, you can see the bottom of the image darker than the non-directional or the real time. I think this issue was fix on master with #102424
This is why I'm asking if you updated the images.

The image in that PR doesn't exhibit the darker area in the directional static

[Calinou]

@Calinou are you sure you updated the images? I still see the issue in your images.

Yes, check the comment's edit history (in the top-right corner of the comment) to see the previous images. They were oversaturated, but the new ones aren't.

However, I initially didn't understand what you were pointing. I didn't notice the darkening in the middle of the image.