The theory behind it is rather more complex than the equations that implement it - which is why it can be done in such a small amount of code.
It's a similar idea to those tiny raytracers and the basis of many demo effects: a relatively simple equation, iterated many times, can produce a complex and even realistic scene.
Now, in the spirit of personal computer magazines of 1980's, you should pause the Youtube video, type in the source code from the screenshot, and run it. Good luck.
Yes, it compiles in to the solver and then can use its own source as the initial state. The guy who writes these (this is one of his many entries to IOCCC) is a genius!
In some examples, I get the feeling that the conservation of volume is not correctly modeled. And in the clock example, how can the fluid escape the hourglass?
I'm not certain if volume was lost or gained, but it looked possible at some points. Some of the other modeling seemed a bit 'off' but for so little code in ASCII it was pretty amazing.
Run the program on its own sourcecode, Wait til the system stabilizes, take that ASCII state as output, compile THAT, and have it be the colorized version. That's how Mel would do it.
http://en.wikipedia.org/wiki/Smoothed-particle_hydrodynamics
The theory behind it is rather more complex than the equations that implement it - which is why it can be done in such a small amount of code.
It's a similar idea to those tiny raytracers and the basis of many demo effects: a relatively simple equation, iterated many times, can produce a complex and even realistic scene.