Why Hash Grids Beat Sparse Trees for Particle Simulation, From the Author
Michael_Moroz_ · x · 2026-09-10
Following up on his VRChat fluid simulation, the author explains the technical choice: hash-based particle simulation beats sparse tree grids because sparse trees require indirections to find neighbors, while hash grids' only downside is hash collisions — and with latency as the bottleneck, hash grids win.
He's also improving the isosurface ray tracer with more PBR-style rendering and basic screen-space effects, currently only in bounded mode; making it performant in boundless hash-sparse mode remains an open problem.
More from Fun
- User delegates flight refund to ChatGPT, gets back $465.25 in travel credit in 10 minutes — TheMoonMidas · 2026-09-10
- Dev leaves GPT 6 Astra to design a camera for days, gets a beautiful over-engineered monster — jakedahn · 2026-09-10
- OpenAI thought Navier-Stokes would be smooth sailing — it's hitting turbulence — pmddomingos · 2026-09-10
- Demo: comparing iPhone Duo vs iPad mini via Astra through ChatGPT's MagicPath plugin — thesaraharminta · 2026-09-10
- A 78-second AI food comedy: Pig Girl crashes a Viking winter feast — Substantial_Depth744 · 2026-09-10
- Pedro Domingos mocks OpenAI for treating cross-lab paper collaboration as a problem — pmddomingos · 2026-09-10