long sim times for sparse volumes?

[johnLIC]

I've noticed that my sometimes pyro smoke simulations seem to take a long time to sim toward the end of the frame range, when there is very little density left due to dissipation.  This is counter-intuitive to me.  Is this generally how it goes with dissipating smoke?

Edited March 20, 2015 by johnLIC

[edward]

There's normally an opacity limit that Mantra stops marching through once it's been hit. If you have lots of low density voxels, then all the voxels through the ray have to be visited.

[willow wafflebeard]

There's normally an opacity limit that Mantra stops marching through once it's been hit. If you have lots of low density voxels, then all the voxels through the ray have to be visited.

 

edward i think he meant simulation slowness.

things to check:

1. how large is your bbox towards the end.

   - if its large try to use the resizefluiddynamic and set a higher value for cutoff

2.check your resolution, by size means no matter how big the bbox, it will retain the resolution.

3.check your sourcing resolutions

things to try:

4.upres workflow

5.try this for adaptive step size https://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&p=158433&highlight=&sid=4fa3d0680b1b5fc3ef61cd3e7032f10b

 

[edward]

edward i think he meant simulation slowness.

 

Whoops, I'll go back to my cave now.

[Annon]

The simulation isn't sparse btw..

Low density (or no density) will take just as long as high density in the sim, so when it dissipates you're not going to see it speed up unless the resize starts cutting the domain down.

[old school]

Turn on the velocity field on the Pyro object to see what's up in the container.

 

It's the velocity in the container that, along with the density as a mask that basically determines where the smoke solver is doing work.

[johnLIC]

I hadn't considered sourcing resolution.  I've been going into the field that points back to the pyro resolution and adding " * .25" at the end.  But my source geo is static, so if it was going to slow the sim it would do it from the beginning, right?

 

Thanks for all the tips.  I will have a look at the velocities in my volume. 

 

So, it looks like adaptive voxel sizes are not a reality yet, huh?  That would be cool.

 

So the upres workflow is worth investigating?  I remember trying to add wavelets post-sim in Maya, using that batch of plugins that are supposed to be Houdini-ish, that I can't remember the name of right now - Peter Shipkov's stuff? - and it seemed like the upres took as long as the low-res sim, and looked very much like wavelets, in a bad way.

[johnLIC]

Soup!  Those are the Maya plugins I was trying to remember.     sOup? One of those letters is capitalized....

[old school]

Upres is indeed worth investigating. There is a shelf tool designed to add an upres solver to existing smoke/pyro/liqud simulations using a few different techniques.

 

It imparts it's own look as you eluded to. For faster moving smoke sims it adds interesting stretching detail to the simulation that may or may not be to your taste.

 

You can run up some decent sized main pyro sims though so using upres to get higher resolution detail is more for look these days, and simulated detail on top of a cached base pyro/smoke/liquid sim.

[Rio.Dominia]

U can use the Gas Resize Fluid Dynamic  node   and in the field tab , change the field cutoff value bigger. So the Solver will kill the smoke faster than before