PDA

View Full Version : RealFlow 5



Mr Rid
04-19-2010, 03:09 AM
Some early test examples of morphing, temperature and crowds-
http://www.youtube.com/user/iosimdk

And yet another fluid simulator-
http://www.magicpics.com/aquarius/aquarius_gallery.html

Red_Oddity
04-19-2010, 06:54 AM
Some early test examples of morphing, temperature and crowds-
http://www.youtube.com/user/iosimdk

And yet another fluid simulator-
http://www.magicpics.com/aquarius/aquarius_gallery.html

I was under the impression Aquarius was a service, not a program that will (ever) be sold outside of MagicPictures

erikals
04-19-2010, 07:05 AM
I was under the impression Aquarius was a service, not a program that will (ever) be sold outside of MagicPictures

yes, looks like it...

Aquarius is an in-house software of MagicPictures,Inc. and there is no schedule for public release.

erikals
04-19-2010, 07:10 AM
RealFlow 5, i liked the temperature previews, esp. the bobbling lava

the other features are also a good start, opens up some new doors.

but nothing would beat a big speed improvement, so, let's see...

Red_Oddity
04-19-2010, 09:18 AM
Doh...it says so on the site..i looked into it a while ago, there's a surprisingly large amount of fluid simulators out there, although most of those are services (some look so-so, like Aquarius, and some are absolutely mind boggling in churning out insanely complex scenes, like Flowline.)

Anyway, RF5 looks really promising, i wonder how easy the C++ SDK will be for extending the existing toolkit and events.

Netvudu
04-19-2010, 09:44 AM
I think Naiad is going to be the new big thing that everybody will drool at. RF will still stay around because of the general user base, though...

aurora
04-19-2010, 09:45 AM
I'm looking forward to getting my upgrade to RF5 but am even more looking forward to getting time to converting my fluid solutions to LW----. I love the look of some of Aquarius's demos but I've had better results with my solvers. Sadly those results, those being mine, were all ran on largish clusters from only 100 nodes up to 90,000 on Kraken and each node multi proc'ed and lots of memory per node. Needless to say getting equivalent results on say only 8 cores is kind of stressing me out.

aurora
04-19-2010, 09:48 AM
I think Naiad is going to be the new big thing that everybody will drool at. RF will still stay around because of the general user base, though...

I'm slightly familiar with Naiad do to Weta getting a some seats a couple weeks ago but have not seen direct results from it. Do you know of any demos on the net?

Red_Oddity
04-19-2010, 10:02 AM
Here's an older FXGuide TV episode showing Naiad : http://media.fxguide.com/fxguidetv/fxguidetv-ep062.mov

aurora
04-19-2010, 10:29 AM
Awesome, thank you. I find it interesting that the front end is going to be open source and paying only for the batch end. I like the tiling feature allowing for infinite domains. I have been studying infinite domains for fluids and most solutions are ugly and often only for shallow fluids. The tiling approach is something I'll have to dig into more!

Netvudu
04-20-2010, 09:45 AM
Aurora, Im sure youll agree with me on how close to Houdini the interface looks. Its much more than a passing resemblance...

What you say about your own solvers sounds amazing. Do you think you will be able to move any of them to a LW-useable state, for us, mere mortals?

aurora
04-20-2010, 10:18 AM
Aurora, Im sure youll agree with me on how close to Houdini the interface looks. Its much more than a passing resemblance...

Yeah and being a Houdini lover I'm in heaven!


What you say about your own solvers sounds amazing. Do you think you will be able to move any of them to a LW-useable state, for us, mere mortals?
Already in the works. I seriously doubt I'll bring the full solvers in as they are for pure scientific work, meaning lots of time and care goes into accuracy, stability, huge datasets(as in billions) , long time runs (sometimes in billions of simulation years) and is designed to run on large clusters. They are written with MPI (MPICH2 specifically) for message passing, and in c and fortran90.

For LWX I'm converting to c++, will most liekly use the built in multi-proc routines in the SDK (or go with OpenMP), will be simplifying things tremendously to give a solver that will work for the normal CG Joe. In the end though it should provide solutions that 'look' the same visually without all the extra anaylsis features added in like the umpteen trillion graphs that get produced to show and prove whats happening in the system.

erikals
04-20-2010, 05:02 PM
how will it be different than e.g. RF, will it calculate faster, or slower?