Jump to content
Diego A Grimaldi

New Bifrost features which seem very powerful

Recommended Posts

I don't think FLIP is going to make a huge difference for gases.

 

@Pablo

well i have absolutely no idea, cause i don't figure precisly how thing works in the back.

 

but :

- having the ability to uprez things without lost/modification of dynamics behavior is a redundant / inherant pb with euler (maya fluids / pyro solver)

- fast moving pyro can be touchy , you have auto-resize but i have never trust at 100% this feature

 

maybe FLIP could offer new opportunity when you want to

- keep the velocity field created on a low rez sim and and keep it for the highrez sim (this could be possible with pyro in H ? but i lack XP in this field)

- ability to rebuild the grid per frame to solve only where you need to

 

And it's different from auto-resize where you have your threshold that truncate info in your grid under this threshold. sometime  you loose info where you don't want.

 

There was a feature in squirt that was impressive at the time. It's the fact that you can animate your fluid grid, and for long trail , smoke trail etc ...

that propagate extremely fast in space , having the ability to animate your grid was just pure magic when you come from a Maya Fluid Background !

Well i still love Maya Fluid a Lot ... oldies but goodies ! :wub:

 

@Thomas

yes i agree this looks like a FLIP smoke implementation but i have no precise info !

Edited by sebkaine

Share this post


Link to post
Share on other sites

Hey Emmanuel!

Well the main reason because FLIP was developed is to make a particle fluid solver more stable.

Obviously this could get more complicated but basically is much better to solve pressure on a grid.

Said that in gases the behaviour is different and multigrid can do the job fairly well.

The only good reason to use FLIP is to get millions of particles advected and render them rather then the grid, or with the grid, and getting a sharper result. You can get a very close result just advection get particles.

I think there is nothing in squirt that I can't do in houdini.

I'm not saying that in certain corner cases splatting the velocity field into particles like FLIP, could be useful, but probably it doesn't justify to move from multigrid to FLIP.

Share this post


Link to post
Share on other sites

I'm not saying that in certain corner cases splatting the velocity field into particles like FLIP, could be useful, but probably it doesn't justify to move from multigrid to FLIP.

 

Thanks for your feedback Pablo. My XP with Pyro is not as good as Maya Fluids, but in fact to achieve those effects where you have huge distance Multigrid is the best option.

The advantage is that you can multithread your sims on the farm and most important you can retake only grids where problem occurs.

 

But from a theoric point of view, i would not agree with you , because when you use Multigrids the voxel grid you solve is huge and not optimised.

Autoresize would be the closest technics to an optimised solution, but the unpredictability it introduce when upresing or when you lost nice velocity detail is not worth it imo.

at least in Maya ...

 

So this concept of storing the velocity grid in particles and rebuild an accurate and optimised grid per frame could offer new possibility. for exemple

- the equivalent of autoresize ( or the animated grid in squirt) that actually works without lost of data

- the fact that you can pre-process / post process your sims on 2 levels Particle Level + Field Level

 

I feel that FLIP pyro could offer new opportunities, well not FLIP by itself but the fact of :

- rebuilding the velocity grid per frame

- storing data on point and not voxel

- ability to control on a particle / field level

 

As i only use particle to control fluids in Maya Fluids and Pyro i really find that it's a path worth investigated !

 

ps : my english suck badly hope i'm not too much abstract here !

Edited by sebkaine

Share this post


Link to post
Share on other sites
Guest tar

Out of interest, what would be the equivalent forum to OdForce for Bifrost/Maya stuff? i.e. technically strong, solution orientated, minimal gossip. A+ quality

Share this post


Link to post
Share on other sites

So this concept of storing the velocity grid in particles and rebuild an accurate and optimised grid per frame could offer new possibility. for exemple

- the equivalent of autoresize ( or the animated grid in squirt) that actually works without lost of data

- the fact that you can pre-process / post process your sims on 2 levels Particle Level + Field Level

ps : my english suck badly hope i'm not too much abstract here !

Well, unfortunately this is not what is happening in FLIP, the velocity field is the same as in multigrid, basically you splat the velocities from the particles into a grid, calculate pressure and update the velocities and copy voxels velocities back to particles. So if your grid resolution is not big enough this process introduce lots of diffusion in the velocity values because a voxel contains too many particles.

An this is the reason because SPH is still around, it only use particles and for small scale fluids it retains much more details than FLIP.

For instance, you can simulate an ocean using FLIP, but simulate foam and spray using SPH.

  • Like 2

Share this post


Link to post
Share on other sites
Guest tar

thanks Igor - I'm reading your tests for adaptive voxels as 10% quicker sims, and, 60% of the memory required.

Share this post


Link to post
Share on other sites

thank you Marty,

 

Yes Adaptive sim, sometime can be same sim time as non adaptive, its all depend of your scene, more depth in your simulation faster and less memory on Adaptive side.

 

Igor

Share this post


Link to post
Share on other sites

@Pablo

well i have absolutely no idea, cause i don't figure precisly how thing works in the back.

 

but :

- having the ability to uprez things without lost/modification of dynamics behavior is a redundant / inherant pb with euler (maya fluids / pyro solver)

- fast moving pyro can be touchy , you have auto-resize but i have never trust at 100% this feature

 

maybe FLIP could offer new opportunity when you want to

- keep the velocity field created on a low rez sim and and keep it for the highrez sim (this could be possible with pyro in H ? but i lack XP in this field)

- ability to rebuild the grid per frame to solve only where you need to

 

And it's different from auto-resize where you have your threshold that truncate info in your grid under this threshold. sometime  you loose info where you don't want.

 

There was a feature in squirt that was impressive at the time. It's the fact that you can animate your fluid grid, and for long trail , smoke trail etc ...

that propagate extremely fast in space , having the ability to animate your grid was just pure magic when you come from a Maya Fluid Background !

Well i still love Maya Fluid a Lot ... oldies but goodies ! :wub:

 

@Thomas

yes i agree this looks like a FLIP smoke implementation but i have no precise info !

 

I think you guys are on the right track. Aero (or aerodynamic) sims in Bifrost are FLIP-based, therefore are particle-driven, not based on a predefined grid. The sims, therefore, are not bound to any particular domain, unlike grid-based systems.

 

There are two different simulations happening in the aero solver: the FLIP particles and the render particles. The FLIP particles drive the movement, are affected by velocity/collisions/accelerators, and the render particles are advected through the FLIP sim. You can have a low-res FILP sim and high-res render component, or vice-versa. Eventually, we'll be able to deform the render particles independent of the FLIP particles, so we can add detail, curl noise, etc. We can't do this just yet, but it's on our roadmap.

  • Like 1

Share this post


Link to post
Share on other sites

Quick question; is bifrost using the GPU?

 

No, not for computations. Not yet, at least.

 

However, we are heavily leveraging Maya's updated Viewport 2.0 OGS for display purposes. You can see this in those demos, showing fully lit and shadowed aero voxels or 50 million particles in the viewport without a hit in performance.

Share this post


Link to post
Share on other sites

Thanks for your answer Adrian, so you'r goal is definitly to unify Liquid and Pyro under the same lagrangian approach !

I'm very curious to test Pyro FLIP stuff in Bifrost to see the +/- vs grid based pyro ! :)

Share this post


Link to post
Share on other sites

Adrian , 

 

I would be curious to know if bifrost implement by default an air field ?

 

Because we were testing bifrost for liquid stuff and the default result looks imo more realistic than the default FLIP result i had in H.

 

I have this feeling that in H FLIP without home made airfield tend to look more like viscous paint than water.

Maybe my lake of experience with H was the problem :) ... but i was able to get very realistic water in a matter of minutes with bifrost ...

 

Cheers 

 

E

Edited by sebkaine

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×