# Computation Fluid Dynamic for RealTime Simulation

This topic is 4809 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

COMPUTATION FLUID DYNAMIC FOR REALTIME SIMULATION Almost flows occurred in nature are turbulent flow, for example smoke, spring, wind … To animate them, solving Navier stoke equation with high Reynolds number is necessary. Solution of the equations of fluid mechanics on computers has become so important to reach realtime simulation. CFD ( Computation fluid dynamic) is a discretization method in space and time, It can solve the Navier stoke equation by approximate method. By using C++ language with supporting of DirectX SDK, HLSL (High-level shading language) the ability of realtime calculation come true. A simple sample is made to demonstrate the ability of CFD – Which is run on the single computer (AMD 2.0 GHz, 1G Ram, GeForce FX 5200 128MB). The result is not bad with Fps from 16 up to 20. Figure1: Front view from far distance Figure2: Front view from near distance Figure3: Front view with high angle Figure4: Right view Figure5: Back view Figure6: Left view You can download this sample at here http://s25.yousendit.com/d.aspx?id=1NC3L233JQBFF0OXJGRUN8I92N

##### Share on other sites
I'm assuming these are all 2d flows right? you're not the first to do this in real time.

##### Share on other sites
even for 3d fluids he's not the first doing it. But some details regarding the performance would be cool(the demo doesn't work for me :( )

regards,
m4gnus

##### Share on other sites
I know that even 3D I'm not the first person to do it, I post it here not to say I'm the first person, I post to get your opinion about my work. I care about RealTime it means the speed. Can you run Demo and give me your opinion so I can improve my work, thanks for your help

@timw You are right here is 2D but I think it enough for this case.

##### Share on other sites
your post sounds like you have something extremly cool and new algorithm and then you talk about navier strokes so it sounded like you weren't knowing that this was already implemented by others...but i'm also curios about the performance...especially with 3d fluids( i hope they are included in the demo)

regards,
m4gnus

##### Share on other sites
np I'm not dissing, can you set up a web page with some fps measurments? the video is fussy I'm having trouble downloading it.

##### Share on other sites
Nice work. You could submit it to Image of the day.

##### Share on other sites
I'm getting an error when trying to run the demo. The error is: "The procedure entry point GetRawInputDeviceInfoW could not be located in the dynamic link library USER32.DLL

Lizard

##### Share on other sites
That looks really good. I'm curious, what grid resolution are you using for the simulation? I don't think it's quite practical in real time yet, but it might implement well on PS3/Cell (although I haven't checked out the official documentation for that yet).

The GetRawInputDeviceInfoW error seems to be caused by including all the DirectX DLLs when you only seem to need d3dx9_25.dll for it to work (at least on my system).

##### Share on other sites

I try to run the calculation only and it take 0.0065s per 1 step, but in this demo I have to use 5 step per 1 frame to give a good animation ( the fluid flow very slow ). So total 1 frame it take 0.0325s.

I run my demo without smoke, the result is 30 fps and 16-20fps with smoke. once again my PC is AMD 2.0 GHz, 1G Ram, GeForce FX 5200 128MB.

Also I try to run in the PC: P3 1.7 GHz, 512M Ram, GeForce FX 5200 128MB, the speed is 3-5 fps.

PC: P4 Dual 3.4 GHz, 2G Ram, GeForce FX 5900 128MB, the speed with smoke is 30 fps.

@ZQJ: I used 80*80 cells for this simulation.

1. 1
Rutin
32
2. 2
3. 3
4. 4
5. 5

• 13
• 9
• 9
• 9
• 14
• ### Forum Statistics

• Total Topics
633328
• Total Posts
3011376
• ### Who's Online (See full list)

There are no registered users currently online

×