SoC project : Fluid Linux/Windows (2005/08/29)

User-contributed CVS development builds. Please test and give feedback!

Moderators: jesterKing, stiv

n_t
Posts: 0
Joined: Tue Jun 28, 2005 2:07 pm

Post by n_t » Sat Sep 03, 2005 8:13 pm

Pablosbrain,Zsolt,Koba: Yes, displaying the req. memory is probably a good way to do it... As the resolution is taken for the longest axis , computing the memory is not just (res^3*x), though... So especially for non-cubic domains a resolution of more than 200 might often be useful, as you otherwise get resolutions along the other axes that are too low. I have to add this to the user docs in the wiki as well.

RRuiz: I found a render of that scene in a spanish forum btw. :) , cool!

Koba
Posts: 0
Joined: Thu Apr 15, 2004 9:48 am

Post by Koba » Sat Sep 03, 2005 9:06 pm

I'm sorry if I have misunderstood...perhaps I didn't quite explain what I was asking about well enough.

In essence, will it be possible to have seperate resX, resY and resZ settings which can be adjusted independantly (for the domain, obviously)?

I mean why have a hundred void cells above a liquid surface when the spashes only rise up 20 cells or so?

Perhaps you could just direct me to the userwiki - I had a look but I couldn't find anything to clear up my queries.

Thank you...

Koba

Rahu
Posts: 0
Joined: Sun Sep 26, 2004 5:16 pm

Post by Rahu » Sun Sep 04, 2005 2:19 am

It's not working for me. The domain disappears in all the frames that are calculated, but the fluid doesn't change at all :/

n_t
Posts: 0
Joined: Tue Jun 28, 2005 2:07 pm

Post by n_t » Mon Sep 05, 2005 7:31 pm

Koba: it should already be working that way - if you create a domain that has a bounding box size of (10,5,2) and set the resolution to 200, you get a simulation that runs on a (200,100,40). as the cells in the simulation have to be cubic, it's enough to have one resolution and the bounding box. if you specifiy (200,100,100) for the example above, it wouldnt work. i hope that's a clearer answer, now :) ...

poutsa: btw., did you get some nice results so far ?

Rahu: sounds like you dont have a fluid object, or it's not in the domain box...

Koba
Posts: 0
Joined: Thu Apr 15, 2004 9:48 am

Post by Koba » Mon Sep 05, 2005 11:08 pm

Thank you for that clarification.

Koba

poutsa
Posts: 0
Joined: Sat Jan 03, 2004 4:49 pm
Location: Munich (Germany)

Post by poutsa » Tue Sep 06, 2005 12:04 pm

Thanks also n_t!!!!
-Great done here you can find a Fluid Test Animation i made (Bink Exe):

http://www.savefile.com/files.php?fid=6998869


seeYA

Vassilios Boucer

Koba
Posts: 0
Joined: Thu Apr 15, 2004 9:48 am

Post by Koba » Tue Sep 06, 2005 5:00 pm

Ok..here goes.

I'm sure n_t has his own plans for what happens next but I'm sure he won't object to a couple of suggestions/feature requests. I know there are plenty of more important things to do right now (stability, bugs etc) so I hope I'm not being a pain or anything :wink:

All I really want to know is if any of these ideas are feasible extensions to what we have now.

- Tileable domains.

For things like simulations of rain falling on large surfaces of water, it would be great to have tileable domains. I have no idea if this is possible with LBM - in essence is it possible for cell neighbours to wrap round the edges of the domain? I suspect it should be possible as LBM is essentially a CA machine so it should be quite flexible with such things. If it were possible, it would be great if domains could be tiled (impression of a much larger simulation than is actually there)

-Keyframed object moving into a domain

This would work something like this: the bounding box of objects that are marked "fluid" but are not yet inside the domain are checked at each frame to see if they have entered the simulation domain. If so, the algorithm calculates which cells now need to be fluid cells and now they are part of the simulation. These cells could have velocity vectors inherited from the IPO curves. The problem here would be objects that have only partially entered a domain.

-Constant volume feature.

If you have placed a fluid sim over a large body of water that is meant to represent a lake, you don't want a few drops of water to rise the water level. If would be nice (if we get keyframing implemented) to allow fluid to drain out the domain from the bottom (as defined by the gravity vector) so as to keep the overall volume in the domain constant.

-"Air" cells

I assume that atmospheric pressure is assumed as a boundary condition at the surface. Shouldn't it be possible to mark objects as "air" objects so bubbles can be made and such like?

One idea I've just had which may be possible already is to apply a displacement map with an animated wave/raindrop texture to the simulated mesh.

Koba

n_t
Posts: 0
Joined: Tue Jun 28, 2005 2:07 pm

Post by n_t » Thu Sep 08, 2005 7:55 pm

Koba wrote:Ok..here goes.
- Tileable domains.
-Keyframed object moving into a domain
-Constant volume feature.
-"Air" cells
Koba
Hi, thanks for the suggestions! Actually most of that is on my todo list :) , but I for example didnt think there was a use for periodic boundary conditions in graphics (=tileable domains). Might be interesting to try, though... I think esp. keyframed object will be important, but that will take a few weeks still...

Post Reply