GridGizmo question

[DiuS]

Hello. Tell me please. How many times is usually accelerates render grid gizmo.

I tested on two scenes. The first is the atomic bomb. Grid resolution of about 60-million.
I tested only 10 frames. And maybe the reason is that. The explosion did not have time to expand. But without Grid gizmo frame handles about 3.5 seconds. With Grid gizmo about 7 seconds.

The second scene is a rocket launch. Resolution of about 100-million cells. Grid gizmo animated. At the beginning of the scene it is a small cube. Its size is about 1/20 the size of the grid. At the end of the simulation the cube grows almost to the size of the grid.
The speed of one frame without Grid gizmo about 3.50 minutes. The speed with Grid gizmo 3.30 minutes.

Normal in both cases turned out. Grid gizmo animated. And simulation clipped by grid gizmo. But what is happening with time?



And the second question. What other method of increasing speed, you know? Except as reduce the resolution / decrease step per frame / quality.

[Ivaylo Katev]

the grid gizmo is not an acceleration tool, it's a tool added for convenience in the cases when you need internal simulation for some object. without a grid gizmo you can do this by inverting the geometry, but this will lead to bad rendering, and you have to make two copies, one for simulation and one for rendering. with the grid gizmo this is one click task.

[DiuS]

But it says here:https://docs.chaosgroup.com/display/P213M/Grid
* This way you could fill irregular shapes with liquid, or generally speed up your simulation by chopping off empty cells when you have an irregular fluid shape, eg a rocket launch (this will not reduce RAM usage though).


And about the acceleration speed of the simulation. What else can you recommend. (For test simulations)

[Svetlin.Nikolov]

Can you drop a screenshot of what your gizmo looks like with the atomic explosion? The gizmo *should* increase speed, as it skips processing of parts of the grid. There could be a bug related to the shape of the gizmo or the fact that it's animated.

[DiuS]

http://piccy.info/view3/7380186/de76...2d33436a22206/

[Svetlin.Nikolov]

Hmm, is the geometry open?

[Svetlin.Nikolov]

You absolutely have to make sure the geometry is fully closed.

[DiuS]

Yes, it is exactly closed.
I selected gizmo mode becomes polygons and chose GraphiteModelingTool> Selection> Open
Nothing happened.

[DiuS]

I tested explosion in the middle of the simulation.
Move the slider to frame 180.
Pressed the button Restore. The simulation was about 27-30 seconds per frame.
Next moved to frame 180 again.
Disabled gizmo and hid it.
Pressed the Restore button. The time was about 30-35 seconds
Like or the same time or a little faster. But why such a weak difference

I thought that if the gizmo is about two times less than the grid simulation, then the time will be saved at least twice.

[Svetlin.Nikolov]

Well, at most twice There are parts of the simulation that depend on the base size of the grid, rather than on the active cells - a good example for that is the time it takes to save the cache file to the hard disk.

[DiuS]

It is sad.

And one more question. If fewer objects or less polygons will interact with the simulation. Can this trick speed up the simulation?
Maybe you know other tricks to speed up the simulation. Such as disabling extra channels. (For the test calculations)

[Svetlin.Nikolov]

Yes, reducing the polygon count that interact with the simulator will certainly help, removing unneeded channels will help too. But the biggest gain you can get is from reducing the Steps per frame and the conservation quality.

[DiuS]

Thanks a lot!

[Svetlin.Nikolov]

Good luck