Fluid test

someone knows how to tell me as he has been made the screenshot fluid test called “fluidtest01-DirectX Renderer#30.jpg”
is it the fluid shader of the nvidia modified of which he spoke foxonfire to the forum?
would I like to be able to realize something of the kind you can help me?
thanks to everybody

Hello paolopix,

I had some success with the deformbytexture DX9 effect shader getting it to look like water, you could start there. I managed to modify it for a glitch like shader, with little knowledge of programming.

Havvvve fun.


hi all,

@paolopix: i did the screenshot(s) you mentioned and yes, i’m actually working on that kind of fluid-shader, which has already been discussed (navier-stokes algorithm). but although the actual output looks quite promising, i’m still facing some problems with the fluid’s physical behaviour. the realisation of such a fluid consists of several unique fragment programs, which have to be combined in a correct order. another problem is that RTT (render to texture) isn’t implemented yet within v4. so you have to do some kind of workaround with Renderer>DXTex>Queue(Texture) and that doesn’t make it easier…
maybe anyone in here has made other experiences?
for a good start i would recommend the paper provided by nvidia and also gregsn’s wave-shader. if i get it work correctly one day i will post it here…

good luck!


i’ve implemented the navier-stokes solver already with a kind subframe iteration system. it was working correct, but the speed was low, around 12 fps with 512x512 and 30 subframe iterations. then i tried to replace the diffusion steps with different approaches. but no time atm to work on it…

thanks for the answers,
I don’t have knowledge for the programming.
and the only experience with effects water, is the example wave02 using the effect pixwave.fx it foresees in this site, but it is another thing.
I wait for the publication of this shader.

@tonfilm: which values did you use for grid scale/grid spacing? i’m still not sure, wether i defined these parameters correctly. is dx in the nvidia paper meant to be the xy-dimensions of a single cell (1/texture size)? in that case, i wonder wether the output of e.g. the jacobi iteration program can be correct. if my textures are 512512, dxdx then would be 0.000004? wouldn’t that be useless for any further operations?
maybe i interpret some things wrong here…

thx for any help,


Flux that fluid shaders looking good!