Kd-Tree based SDF evaluation experiment

During my last LIVE stream Sean Sartell asked me in the chat about SDFs for polygon meshes. So I did an experiment on computing a Kd-Tree for the well known Sponza 3D model, to see how much it'd help evaluate SDFs quickly vs a brute force approach. It is fast enough that I can raymarch it, do some soft-shadows and cheap occlusion, quite like for my Shadertoy mathematical/procedural examples.

I wrote an article for it:

SDF,Kd-Tree,acceleration structure,shader,GLSL,raymarching,