Per Vognsen
New Matthias Müller video just dropped. https://www.youtube.com/watch?v=XmzBREkK8kY
18 - How to write a FLIP water / fluid simulator running in your browser

YouTube
Dec 31, 2022 at 5:56pmWeb
Show threadPer VognsenDec 31, 2022
Does anyone know if the particle density estimation trick he uses at the end for speeding up the pressure solver is standard in FLIP solvers? It's been a long time since I read the Bridson book, but I don't think I saw it there.
Show threadDan PiponiDec 31, 2022
@pervognsen IIUC that's the adjoint of linear interpolation. I thought that was a standard approach for computing densities. It's the thing you can trick libraries like TensorFlow into doing via AD.
Show threadPer VognsenDec 31, 2022
@dpiponi Right, sorry, I didn't mean the density estimation itself (I agree that's just the scatter/adjoint of lerp) but using that estimate to add an extra pseudo-overrelaxation term.
Show threadDan PiponiDec 31, 2022
@pervognsen That's a nice trick but I don't know it. (But don't read much into that as the last time I worked in fluids I didn't use FLIP)
Show threadPer VognsenDec 31, 2022
@dpiponi Hmm, looking at the code, his rho_0 for that k * (rho - rho_0) is the average fluid density in the sim grid. So it's going to have similar non-physical effects that you get trying to enforce global conservation of energy, mass, etc by reinjecting it evenly. But probably okay for small closed systems or if you tune it by hand.