Vivian Meazza wrote:

Mathias Fröhlich wrote

On Mittwoch 15 Juni 2005 11:00, Vivian Meazza wrote:
When browsing Mark Harris' web page I got the impression that he
simulates the
This one?

I can't find any reference in the site or the papers there to simulating
airflow. It just renders clouds.

Look into that. I believed that this was integrated somehow. That looks
phantastic and does things like that.
Watch the video on the bottom of that page.

Wonderful - just what we need! However, where's the code? I don't think it's
what we have. The OpenGL stuff for download from Mark's site just renders
cloud. This is "Simulation of Cloud Dynamics on Graphics Hardware"
From his paper:

"5. Hardware Implementation

As mentioned before, we perform all of the numerical
computation for our cloud simulator in the programmable,
floating point fragment unit of a graphics processor."

Hmmm. Have I got this one wrong?
Mark's code (OpenGL) renders up to 51 fixed shape clouds. His lighting and
shading is good. I like the way the aircraft penetrates cloud from an
outside view. Harald's clouds are nicer, more varied, but perhaps not quite
as well lit and shaded. The cloud penetration from an outside view could be
improved (and I'm sure will be). Mark's code does not seem to be under
active development, while Harald's is. I therefore support the retirement of
Mark's OpenGL code.

Of course, if we can go the "Simulation of Cloud Dynamics on Graphics
Hardware" route, then wow! All bets are off.
Anyone know how to program graphics hardware so that any/all cards would


On Harris pages there is two different things.
One is about cloud rendering - physical aspect of scattering of light and technical implementation using dynamic billboards, etc. The other is about simulating the formation of clouds (and simulating fluids in general). The 'Skywork' code available on his site and integrated in FG only handle the rendering, the simulation of formation of clouds is usually non real time anyway. It could be done realltime of course, it all depends of the degree of realism one wants. Other papers show how to do that with non physical rules and still have nice results. See Dobashi and Nishita paper :

About the current lighting of clouds : I admit that it is very simplistic but it has at least the advantage of being light
for the cpu and the graphic card.
Now it is possible to integrate the Harris lighting method with the existing code and have something perhaps not so far from his rendering. This is easy to do, I could make a prototype (err when I have a free moment).


Flightgear-devel mailing list

Reply via email to