Discussion
In 3D computer graphics there's 2 main algorithms used: Triangle-rasterization & ray-tracing. Triangle rasterization was more common until recently (its what OpenGL standardizes) since it can be cheaper, whereas ray-tracing is faster at rendering more complex scenery.
That said there's research into bringing some of the major benefits of raytracing to cheaper GPUs, & one technique I've learned recently (via https://www.youtube.com/watch?v=3so7xdZHKxw ) is Radiance Cascades!
1/3?
Radiance Cascades renders lighting into a texture, taking samples around points on a grid. With coarser grids taking more samples. We can then interpolate between these samples & sum the different levels to calculate the lights reaching each point, possibly caching results into a texture.
With this we can recreate the effect of single-bounce ray-tracing with a couple cheap preprocessing steps!
It was innovated in part by Grinding Gear Games in Auckland, 2024.
2/3!
The reason why this works is because:
1. The more close a surface is to a light the fewer samples need to be taken to find said light.
2. We can split rays between a light & a surface at any arbitrary point along it, as long as we take into account opacity.
Yes, I find ray-tracing interesting as well since it relies not only on faster hardware but also clever software. Whilst triangle-rasterization demonstrates just how powerful transformation-matrices can be!
3/3 Fin!
Install bonfire.cafe
Get the full app experience
