Cursed radix sort built purely from fragment shader blits inside VRChat's Udon

Michael_Moroz_ · x · 2026-09-03

A developer implemented a fully working radix sort inside VRChat's Udon scripting environment, where compute shaders, atomics, and scatter writes are unavailable — every step runs as fragment shader blits.

Key design details:

On an RTX 5090, the author reports it's only 2x slower than a 32-bit CUDA radix sort — but only for ≤1M elements with 20-bit keys, where everything fits in cache; the advantage disappears at scale.

Related event: Developer Implements Radix Sort Using Only Fragment Shaders in VRChat(2 posts)→

Original post →

More from Fun

Fun channel →