Solver Sop - Allow Caching To Disk Location

[fell00]

Hello,

I'm running the Solver Sop.

I checked "Cache Simulation" as well as "Allow Caching To Disk".

Where can we find this files on disk?

I'm running some boolean operations on deforming geometry in the Sop Solver and it uses up all my RAM eventually, so now I'm trying "Allow Caching To Disk" but it doesn't say where it saves these files. I don't see them in the HOUDINI_TEMP location, I don't where they could be..

 

[fell00]

In case anyone's confused as to what I'm referring to, here's a screenshot. 

[underscoreus]

If I were you I'd avoid using that checkbox and instead just use a filecache node to cache the output of the solver.

As a personal note though, booleaning deforming geometry inside a SOP solver sounds like the kind of horror you'd see in a psychological thriller.
If at all possible I would strongly suggest seeing if any of the parts you just described can be avoided since all of them combined could be a monster consuming a lot of resources.

If possible maybe you could share your HIP file?

[fell00]

Hello,

Yes, it pretty horrible.

Here's an example hip file. The hip file runs fine, but in my case I have a lot of high poly geo.

It's basically a progressive boolean. Doing the boolean each frame based on the results of the previous frame.

I have been able to use a File Cache but it takes a long time.

Can you think of any other way to do this?

solversopbool_odforce.hip

Edited April 4, 2021 by fell00

[fell00]

The example I uploaded works for non-deforming geometry but for deforming geometry using the solver sop won't work anyways because it changes shape every frame.

I found a solution though.

To boolean deforming geometry you can use a trail sop on your cutter geometry to grow it.

So you have a growing cutter that follows behind your original cutter.

Then convert the cutter to sdf and back to polygons. Good to cache it.

Then boolean with that cutter and cache it out.

I think I also prefer this method with static geometry than using the solver sop.

Edited April 4, 2021 by fell00

[underscoreus]

Hey @fell00,
Great that you found a solution! I'll attach a hipfile here with the way I would go about something like this. It essentially boils down to not using booleans at all but rather use the "vdb combine" node. Doing boolean-like operations but with VDB's is in my experience WAY more stable, however, it comes at the cost of having to compute the volumes, both converting the geometry to a volume and then back again at the end. But, the times I've needed something like this it has been a preferable alternative to dealing with the booleans unreliable nature.

Hope it can help!

deforming_boolean_alternative.hiplc

[fell00]

Thanks for posting, yes I agree vdb's are very stable, more than booleans.

I considered using vdbs but in my case I had the high poly geo already, modeled, uv'd and textured and didn't want to redo those.

I had simulated the deformation before the slice using a bullet simulation with glue and soft constraints then a rdbdeformpieces node post sim. That process required polygon geometry.

Then the slice.

I thought that if I converted the deformed geo to vdb and back to polys, if I redid the uvs it would change the look.

 

Edited April 4, 2021 by fell00

[underscoreus]

Yeah, that is the downside to my method. Attributes would have to be transferred back from the original geometry onto the new geometry and you could end up losing some details depending on how many small details are in the original geo and how high res you make the VDB.