Move PointLight to LTC

[dylansechet]

Objective

Move spherical area lights (PointLight with radius>0) to LTC.

This is a follow-up to #23288 and I'm working from that branch, only 7317542 is new.

Solution

I used the method described in this SIGGRAPH course ("Real-Time Line- and Disk-Light Shading" by Heitz and Hill). This PR is pretty much a direct translation of the reference GLSL implementation shared in the course.

This has the downside of breaking anisotropy for area lights, as we don't have the LTC lookup tables for it.

Testing

• Ran many_lights
• Compared to eevee/cycles, see showcase below

This PR seems to improve many_lights performance on my system (igpu), which makes no sense to me given the extra work it is doing. Could someone try to reproduce this ?

---

Showcase

Main

LTC

Cycles

---

[github-actions]

Your PR caused a change in the graphical output of an example or rendering test. This might be intentional, but it could also mean that something broke!
You can review it at https://pixel-eagle.com/project/B04F67C0-C054-4A6F-92EC-F599FEC2FD1D?filter=PR-23400

If it's expected, please add the M-Deliberate-Rendering-Change label.

If this change seems unrelated to your PR, you can consider updating your PR to target the latest main branch, either by rebasing or merging main into it.

[alice-i-cecile]

This has the downside of breaking anisotropy for area lights, as we don't have the LTC lookup tables for it.

We should fix this before merging IMO.

[holg]

Great work and progress IMHO!
LTC area lights would let us replace some multi-light hack with a single area light per luminaire + a photometric profile texture for angular intensity modulation: physically correct and far fewer entities. (eulumdat_bevy)

Question about additional shapes: The current PR covers rectangular area lights. For real-world luminaire simulation,
disk/circular lights are equally important (most downlights and recessed ceiling fixtures have circular apertures),
and line/tube lights cover fluorescent and LED strip fixtures. Would contributions be welcome?
Dunno, to me it seems it would best be the more general approach in bevy_light needed for other