It is currently 19 Sep 2014, 11:46

All times are UTC + 1 hour [ DST ]




Post new topic Reply to topic  [ 150 posts ]  Go to page 1, 2, 3, 4, 5 ... 15  Next
Author Message
 Post subject: PhysX Fluid Simulation in Softimage ICE
PostPosted: 21 Jan 2012, 23:36 
Offline
User avatar

Joined: 10 Aug 2011, 12:35
Posts: 126
Location: Kharkov, Ukraine
Hi here,
I'd like to public some experiment with physx fluids, it is custom ice node which wraps sph fluids.
Image
Attached archive contains samples, installation guide and workgroup. There are no docs or toolbars, all workflow similar to usual particle simulation except solver, you need just to replace "Simulate Particles" with the "PhysX Simulate Fluid" compound. Builded and tested under 2012 sp1, win x64 only. CUDA card ( hi-end fermi is preferable ) required. Also, maybe you will need to install this
Some notices:
-if you get red "particle GPU creation failed" try to decrease max particles number parameter or drop down Radeon from your system and buy Nvidia :) this attribute allocates ( once ) memory in GPU ram at the beginning of evaluation, so it is great tool to detect how many particles you can use with your videocard
- Max Speed attribute - this has VERY critical influence on the overall performance, set it as low as possible
- Approximate Fluid Box - rough bounding box in which your simulation happens ( big values slow down speed, low values leads to bad case when particles tops to collide with obstacles )
- Statical obstacles are implemented as triangle ( actual shape in bullet-terms ) mesh, they load once at the beginning of evaluation and cannot be changed during this eval
- Dynamical obstacle implemented as convex ( only at this time, I think it is possible to get it as triangle meshes in future ) and node check this port at each frame. when variable topo = false - only SRT of obstacle are traced by the solver, if true - all topology changes ( include simple points movement ) are mirrored in the Px world
- if it is possible - do all your sims near center of the world ( 0\0\0 ) and at normal scales ( in comparison to home grid size )
I did some test on GTX 480 ( 1.5 GB ), and at this time it is possible to run about 2M particles, which allocates about 0.9M videoRAM. CPU mode is also very fast.
So have a fun with the incredible fast water :)
-Regards,
Oleg


Attachments:
pxFlow.jpg
pxFlow.jpg [ 47.77 KiB | Viewed 11373 times ]
pxNode.jpg
pxNode.jpg [ 35.79 KiB | Viewed 11373 times ]


Last edited by Mr.Core on 21 Jul 2012, 19:06, edited 1 time in total.
Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 21 Jan 2012, 23:38 
Offline
Administrator
User avatar

Joined: 04 Jun 2009, 12:15
Posts: 4047
Wow, looks fascinating. Thanks for sharing! :ymhug:


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 21 Jan 2012, 23:43 
Offline

Joined: 09 Jun 2009, 23:47
Posts: 88
Hey, that's fantastic! Mr. Core strikes again!
thanks a bunch.


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 00:54 
Offline

Joined: 22 Mar 2010, 18:43
Posts: 411
Wow, great stuff! Can't wait to play with this...


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 01:33 
Offline
User avatar

Joined: 09 Jun 2009, 14:49
Posts: 424
Location: London, UK
haha! think I got it working , soon as you crank it p though its not long before

Runtime Error!

R6025

-pure virtual function call

:D im only on a Geforce GTX 560 ti though so probly cant go much far on gpu memory. but still I just have to find the limit and stick to that.

awesome! ^:)^ will post some watery stuff soon!

_________________
Gossip is what no one claims to like, but everybody enjoys.


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 01:55 
Offline
Moderator
User avatar

Joined: 21 Jul 2010, 00:33
Posts: 585
Wow!! This is freaking amazing... So insanely fast, if this gets serious, it will take liquid simulations to a whole new level...
:ymapplause:

Are there any serious drawbacks of its GPU nature? Or is it possible for something production ready to come out of this?

_________________
Gustavo Eggert Boehs
Blog: http://www.gustavoeb.com.br/blog | Training: http://gustavoeb.com.br/blog/?cat=16&lang=en_us


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 11:31 
Offline
User avatar

Joined: 10 Aug 2011, 12:35
Posts: 126
Location: Kharkov, Ukraine
>>Are there any serious drawbacks of its GPU nature? Or is it possible for something production ready to come out of this?

I think that there are several drawbacks but they related mostly to its game-orient nature, so it has not a bunch of settings, it has fixed spatial hash grid size ( some internal structure for better performance when searching for neighbors ) and speed stay over quality. On the other hand, it is really fast, and if you dont need some advansed stuff like lagoa's plastic\tearing structures, physx is not bad choice. Considering rayfire uses physx too this fact is a good proof that game engine can be used in production. And of course if you will use CPU mode ( which is also very quick ) you aren't limited in memory.


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 15:52 
Offline
User avatar

Joined: 09 Jun 2009, 11:59
Posts: 619
Location: warsaw
a-we-so-me!

demo scenes working okay here (gtx285)

_________________
http://vimeo.com/origin/videos


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 23:21 
Offline
Administrator
User avatar

Joined: 04 Jun 2009, 12:15
Posts: 4047
Could you please implement "CUDA accelerated Rigid Bodies" next? =))
But seriously: just had a chance to actually play with it for a little while
and it seems to work like a charm... Beautiful work, Mr. Core!
;)


Top
 Profile  
 
 Post subject: Re: PhysX Fluid Simulation in Softimage ICE
PostPosted: 22 Jan 2012, 23:31 
Offline
User avatar

Joined: 10 Aug 2011, 12:35
Posts: 126
Location: Kharkov, Ukraine
>>Could you please implement "CUDA accelerated Rigid Bodies" next?
=))
btw, bullet node finally uses CUDA :)
I thought about it and one thing that stopped me - no triangle ( actual ) shape for dynamic rigid bodies.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 150 posts ]  Go to page 1, 2, 3, 4, 5 ... 15  Next

All times are UTC + 1 hour [ DST ]


Who is online

Users browsing this forum: No registered users and 5 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group