[Epic] Ray-traced global illumination #87

New issue

Open

opened 2026-09-27 05:44:58 +00:00 by backspace119 · 1 comment

backspace119 commented

2026-09-27 05:44:58 +00:00

Owner

Goal: dynamic diffuse global illumination (bounce light) on RT-capable GPUs.

Approach options (pick after a spike)

Probe-based (DDGI-style) - recommended first. Camera-following cascaded irradiance-probe grid; each frame a budgeted subset of probes traces N rays against the TLAS, shades hits with sun/sky/local lights (+ previous-frame probe irradiance for multi-bounce), stores irradiance + distance moments (Chebyshev visibility). Lit pass samples probes for indirect diffuse. Stable, low noise, no heavy per-pixel denoising, scales well to large SL scenes, amortizable.
Screen-space ReSTIR GI. Higher fidelity, much heavier denoising dependence; revisit after TAA + denoiser mature.

Interplay

SL already has reflection probes (our VK probe path handles specular); GI replaces/augments the diffuse ambient term. Needs a clear blend with probe ambient and SL's environment ambient to avoid double counting.
Hit shading needs a material/texture lookup at hit points (albedo from the RT material table - the alpha-test code already samples textures).

Children

Spike: probe grid prototype (static cascade, sun only) + debug visualisation.
Probe update pass (budgeted rays, hit shading incl. local lights, multi-bounce).
Lit-pass sampling + blend with probe ambient.
Cascades + camera scrolling + teleport invalidation.
Settings/quality tiers + GPU timers.

Size: L. Depends on: RT foundation (done), benefits from temporal epic. Risk: light leaking through thin SL walls (mitigate with Chebyshev visibility + probe relocation), cost in dense sims.

**Goal:** dynamic diffuse global illumination (bounce light) on RT-capable GPUs. ### Approach options (pick after a spike) 1. **Probe-based (DDGI-style) - recommended first.** Camera-following cascaded irradiance-probe grid; each frame a budgeted subset of probes traces N rays against the TLAS, shades hits with sun/sky/local lights (+ previous-frame probe irradiance for multi-bounce), stores irradiance + distance moments (Chebyshev visibility). Lit pass samples probes for indirect diffuse. Stable, low noise, no heavy per-pixel denoising, scales well to large SL scenes, amortizable. 2. **Screen-space ReSTIR GI.** Higher fidelity, much heavier denoising dependence; revisit after TAA + denoiser mature. ### Interplay - SL already has reflection probes (our VK probe path handles specular); GI replaces/augments the *diffuse* ambient term. Needs a clear blend with probe ambient and SL's environment ambient to avoid double counting. - Hit shading needs a material/texture lookup at hit points (albedo from the RT material table - the alpha-test code already samples textures). ### Children 1. Spike: probe grid prototype (static cascade, sun only) + debug visualisation. 2. Probe update pass (budgeted rays, hit shading incl. local lights, multi-bounce). 3. Lit-pass sampling + blend with probe ambient. 4. Cascades + camera scrolling + teleport invalidation. 5. Settings/quality tiers + GPU timers. **Size:** L. **Depends on:** RT foundation (done), benefits from temporal epic. **Risk:** light leaking through thin SL walls (mitigate with Chebyshev visibility + probe relocation), cost in dense sims.

backspace119 added this to the Rendering: temporal + RT effects milestone

2026-09-27 05:44:58 +00:00

backspace119 added the

feature

rendering

epic

labels

2026-09-27 05:44:58 +00:00

backspace119 referenced this issue

2026-09-27 06:34:26 +00:00

[Epic] Emissive surfaces as light sources (RT emissive lighting) #91

backspace119 commented

2026-09-27 06:34:26 +00:00

Author

Owner

Scope note: GI hit shading must include emissive radiance (PBR emissive + legacy fullbright/glow mapping) so emissive surfaces contribute soft bounce light. Sharp emissive area lighting is tracked separately in #91.