The FHP3 model is outdated, from the physics point of view, and superseded by the Lattice Bolzmann models for to decades. However it remains a simple, powerful, accessible model for desktop simulations of subsonic turbulences.
At the turn of the millenium, I have thoroughly optimised the computations of the particles collisions using a large array of modern methods : SIMD, strip-mining, SMP, boolean reduction...
Using the fastest and latest P3@550MHz, I once reached a peak update speed of 150M sites/second, enough to see vortices develop in real time. More information is found (in french) in my Masters thesis :
http://ygdes.com/memoire/
More than 17 years later, computers have evolved to a point where this algorithm would totally scream, thanks to the ubiquitous 128/256-bits registers and on-die SMP. And I have since then learned a lot more about programming and architecture...

It'