Naiad

[Solitude]

Thx for the ifno, Nanocell... I totally agree that it will be (likely) a better investment, and that the workflow itself will be much better as well. I went to see a Siggraph presentation by Robert Bridson, and from what I remember most of the stuff he showed/talked about was how a lot of what he is doing is going back to relying on particles. (my memory sucks, and took in a lot from other presentations as well) I'm sure it was a hybrid, but at the very least it looked very, very SPH.. even showing a particle-only render that was awesome. I still totally look forward to the Naiad way of all the open-source stuff, but I think or would at least hope Realflow would follow suit in some way if Naiad is competitively priced. The realflow .bin format isn't totally closed (it ships with a format spec), so anybody familiar with writing a a binary reader/writer and a plugin for their app it really isn't all that trivial... (that's just input/output though). For RF5, who knows what they will do in terms of opening up the app there will at least be a C++ sdk now. It's nothing compared to what Naiad is planning, but either way I'm excited to see all of the fluids stuff picking up whether it be Houdini, RF, or Naiad.

Edited March 15, 2010 by Solitude

[lukeiamyourfather]

Its worth mentioning that Naiad isn't the only simulation package that's going to use a hybrid solver. RealFlow 5 has a grid solver that can interface with the SPH solver in some pretty awesome ways. I do however wish RealFlow had an open source interface, and/or a more portable solver. Cheers!

[lisux]

Its worth mentioning that Naiad isn't the only simulation package that's going to use a hybrid solver. RealFlow 5 has a grid solver that can interface with the SPH solver in some pretty awesome ways. I do however wish RealFlow had an open source interface, and/or a more portable solver. Cheers!

Yep that is truth, this is really a great improvement for RF, and the new awaited SDK will be fundamental to open it more and allow people to make their own plugins.

RF will still be competitive, sure.

[Solitude]

Just speculating, since I don't have it open, but doesn't Houdini also use some form of Hybrid already as well for the voxel fluids? I remember seeing a few "surface tracking" things in the liquid solver, and also surfels, which appear to be particles used for deriving the surface of a liquid. ..?

[jason_slab]

Just speculating, since I don't have it open, but doesn't Houdini also use some form of Hybrid already as well for the voxel fluids? I remember seeing a few "surface tracking" things in the liquid solver, and also surfels, which appear to be particles used for deriving the surface of a liquid. ..?

yip, at the moment the sand solver uses a hybrid approach, but it's not working to well. maybe, or hopefully the next version of houdini will have some improvements in this area

jason

[Diego]

This may be an interest reading to find out more about Naiad : http://www.fxguide.com/article606.html

[Solitude]

This may be an interest reading to find out more about Naiad : http://www.fxguide.com/article606.html

I just wet myself. Thanks!

[anamous]

This may be an interest reading to find out more about Naiad : http://www.fxguide.com/article606.html

Based on that article and the GUI screenshots ("BOPs", "FOPs" ), I fail to see the extreme difference between this and Houdini's solvers. Yes, it's probably much faster etc. but how is its concept any different from DOPs? I know we all love to rant about DOPs (and often rightly so, especially when it comes to speed and elegance) but this looks like a similar framework with better performance, a slicker GUI, minus interaction with SOPs/etc.

I know next to nothing about Naiad so far (very sparse information base out there, and lots of marketing talk), but what I know so far doesn't warrant a "revolutionary" tag.

cheers,

Abdelkareem

[nanocell]

Based on that article and the GUI screenshots ("BOPs", "FOPs" ), I fail to see the extreme difference between this and Houdini's solvers. Yes, it's probably much faster etc. but how is its concept any different from DOPs? I know we all love to rant about DOPs (and often rightly so, especially when it comes to speed and elegance) but this looks like a similar framework with better performance, a slicker GUI, minus interaction with SOPs/etc.

I know next to nothing about Naiad so far (very sparse information base out there, and lots of marketing talk), but what I know so far doesn't warrant a "revolutionary" tag.

cheers,

Abdelkareem

You are correct in that it does looks similar to DOPs, and probably functions in a similar way as well. The thing that I found to be "revolutionary" is the volumetric 'compositing' capabilities. Naiad has some pretty cool tools for manipulating volumes (shrinking, growing, inverting, and other fancy field stuff). Now, I'm not saying that Houdini can't do it; it is just infitely more cumbersome to achieve the same thing. Houdini's volume handling mechanism needs more ... flexibility? I like how Naiad doesn't limit your simulation to some predefined "volume box". The "volume box" just expands to wherever the simulation needs to go (although you can limit it with a killplane or collider or something, if the need arise).

I guess Naiad is what DOPS should be .

Edited April 8, 2010 by nanocell

[lisux]

Based on that article and the GUI screenshots ("BOPs", "FOPs" ), I fail to see the extreme difference between this and Houdini's solvers. Yes, it's probably much faster etc. but how is its concept any different from DOPs? I know we all love to rant about DOPs (and often rightly so, especially when it comes to speed and elegance) but this looks like a similar framework with better performance, a slicker GUI, minus interaction with SOPs/etc.

I know next to nothing about Naiad so far (very sparse information base out there, and lots of marketing talk), but what I know so far doesn't warrant a "revolutionary" tag.

cheers,

Abdelkareem

AFAIK the procedural approach is more or less the same as DOPs. Flowline also use a similar approach. So both of them uses something similar, but the solver is much better in Naiad or Flowline.

All the stuff about revolutionary GUI, approach, etc ... is just marketing.

But is a good thing to SESI to notice how other companies that only do simulation software are using something similar to DOPs, that proves that the concept is good. Now we only need a solver that can compete in terms of speed and amount of data simulated against flowline and niad.

[anamous]

Naiad has some pretty cool tools for manipulating volumes (shrinking, growing, inverting, and other fancy field stuff). Now, I'm not saying that Houdini can't do it; it is just infitely more cumbersome to achieve the same thing. Houdini's volume handling mechanism needs more ... flexibility? I like how Naiad doesn't limit your simulation to some predefined "volume box". The "volume box" just expands to wherever the simulation needs to go (although you can limit it with a killplane or collider or something, if the need arise).

By volumes, do you mean voxel based simulations or collision SDFs? Does that mean that voxel grids are flexible in XYZ dimensions, and voxels are added and removed on demand? What do you setup then, the size of a single voxel?

But is a good thing to SESI to notice how other companies that only do simulation software are using something similar to DOPs, that proves that the concept is good. Now we only need a solver that can compete in terms of speed and amount of data simulated against flowline and niad.

That's my point. In the long run, if I have to choose between two options, I'd rather have more flexibility and less speed than the other way around, simply because algorithm optimizations can always happen as we go, whereas switching to a completely new architecture for more flexibility takes a lot of time (see nucleus for reference...). And trust me, optimizations are coming.

cheers,

Abdelkareem

[lisux]

And trust me, optimizations are coming.

cheers,

Abdelkareem

We trust you, we know it

[static]

Just speculating, since I don't have it open, but doesn't Houdini also use some form of Hybrid already as well for the voxel fluids? I remember seeing a few "surface tracking" things in the liquid solver, and also surfels, which appear to be particles used for deriving the surface of a liquid. ..?

Houdini will be using a FLIP solver (similar to the sand solver) for their liquids in 10.5, or so I've heard. Not sure what they're using now since I don't use it, because it's so ungodly slow.

Edited April 9, 2010 by static

[nanocell]

By volumes, do you mean voxel based simulations or collision SDFs? Does that mean that voxel grids are flexible in XYZ dimensions, and voxels are added and removed on demand? What do you setup then, the size of a single voxel?

The voxel grids dynamically expand as is needed by the simulation, so yes they are 'flexible in XYZ dimensions'. I would expect that all voxel representations are implemented in the same way so you can, at any stage in your network, expand or erode or composite your SDF and the voxel grid should adapt accordingly.

Each "body" in Naiad (ParticleLiquid, or Volume) have settings on the container where you specify the "cell" scale to adjust the size of the cells/voxels. So it can be different for each volume in the SIM which is very handy especially when you start working with very large collision SDFs (like as terrain or mountain).

<snipped>

Edited April 11, 2010 by nanocell

[old school]

A Flip and Pic solver gains it's speed due to the fact that it doesn't need to substep to remain stable while an SPH solver needs sub-stepping as Marcus mentioned.

Open unconstrained fluids do well with Flip but when constrained in a bottle or glass and undergoes extreme transforms from frame to frame it needs to be sub-stepped and looses it's benefits compared to SPH. Think of a fluid inside a bottle that is keyframed in a character's hand with hard hits and huge transforms. That is where SPH still has a leg up on Flip, for now that is.

SPH in Houdini does ok when distributed on the local machine as there are a few parts in Houdini still not multi-threaded (POPs esp). Just follow the Old School blog on distributing sims on your local machine to gain performance. Nothing to compare to the sim sizes you can achieve with Naiad though.

Passing SPH through POPs is still something to help tweak the simulation. You can add particle like behaviour to SPH and blend the weighting between the two behaviours.

Edited April 10, 2010 by old school

[mightcouldb1]

Thanks for these great explanations guys. It's really helping me understand how all of this wacky fluid stuff works.

Edited April 10, 2010 by mightcouldb1

[goldleaf]

This is really, really cool. Can't wait for that horizon to get here, so we can dig in.

For the moment, Naiad is Linux/OS X, right?

[nanocell]

This is really, really cool. Can't wait for that horizon to get here, so we can dig in.

For the moment, Naiad is Linux/OS X, right?

Yep, and chances are it would probably stay that way.

mv /windows /dev/null

[pclaes]

mv /windows /dev/null

Hahaha

[lisux]

mv /windows /dev/null

Jaja nice one!