Hi all.

Some times ago i create Houdini 9.0 network, for automatic shattering poly models. I use foreach SOP and cookie SOP for this.

Simple poly cube, shattered in 300 pieces:


Shattered cube in some test RBD action:


And now. I download 9.1, install and import my scheme, and it don`t work.

Cookie SOP node in Houdini 9.1 produce unnecessary primitives. This primitives is too big for fuse SOP cleaning (or faset SOP). Why cookie SOP was changed???


1. - I never see before, this strange geometry, in cookie SOP node in Houdini 9.0…
2. - How i can find by expression, this unnecessary primitives, for deletion ? This primitives is part of the cookie groups (Ain, Bin, etc)
3. - My old cleaning ways, for deleting degenerate primitives don`t work now - fuse SOP, faset SOP don`t help me resolve this problem.
4. I tune all parms in new cookie SOP. Of course, in some 3D toleranse values, jitter values, i don`t see the “dirt” geometry.

What happen with cookie SOP node ? Old cookie, in Houdini 9.0 work fine for me. My test don`t work anymore, in Houdini 9.1.

Just as a work-around you could perhaps use the geometry (without the extra bits) as a bounding volume, group what you want to keep and delete the geometry outside that volume.

I'm working on a file that I would like to work on in 9.1.
It has geometry that uses the cookie. When I open the file
in 9.1 I get funky looking geometry also.

Just thought I'd chime in so you don't think you are nuts.

I just noticed a fix to the cookie SOP, you might want to download a new cut of Houdini:

http://www.sidefx.com/journal [sidefx.com]

Sean

Actually I already have 9.1.124 and I'm still getting the same problem.

mrCatfish, thanks, i test it (build 168), and see troubles again


for example - second image from houdini 9.0 build 858:


And two files with cookie SOP using - from houdini 9.0 and houdini 9.1

Cookie from h9.1, in some cases, produce strange primitives out of bound, and holes in result surfaces. i never see this results in cookie SOP in houdini 9.0

In Houdini 9.1 the Cookie SOP is a bit more flexible but requires a little bit more babysitting. To get a similar appearance to Houdini 8.2, turn on ‘Assume Input A is Closed’, (cause it is). You can try a jitter value of 0.01.

live_fx
mrCatfish, thanks, i test it (build 168), and see troubles again

I didn't try this in 9.0, but in 9.1 use the Mountain sop instead of the Fractal and the outcome works fine.

Another thing is that the Cookie SOP has generally worked better with two closed surfaces.

Like instead of cookie based on a box and a plane, extrude the plane so its a box too.

Wolfwood
In Houdini 9.1 the Cookie SOP is a bit more flexible but requires a little bit more babysitting.

More babysitting? Does this mean it's not really much more robust?

jason_iversen

Wolfwood
In Houdini 9.1 the Cookie SOP is a bit more flexible but requires a little bit more babysitting.

More babysitting? Does this mean it's not really much more robust?

More babysitting as in a few more options to play with to set up correctly.

Your mileage may vary.

Seems pretty stable to me.

funky result in a “simple” case :

[img211.imageshack.us]

I try enable “assume A” option in “dirt geometry” situations:

“Assume input A” option is off. We see out of bound geometry and “in bound” holes:


“Assume input A” option is on. Outbound geometry is some cleaned, but we have holes and little “outbound” primitives:



WolfWood - is true, with closed surfaces , cookie SOP work cleanly. But in some cases i see troubles again:



My general question:

How i can detect inbound Holes in result geometry?
I can`t stitch or close this holes, but, if, in case, i can find it by expression, i can make some search-cycle expression like
“we have dirt geometry? = true - next step - try other jitter values or other objects positions, return.”

how i can detect this holes and outbound primitives? Big thanks for answer.

( cookie SOP was heavy changed. For what? Sad. )

To detect holes, non-manifold geometry, you can use the Divide SOPs ‘Removed shared Edges’ Toggle.

If this returns no points or prims, than your geo is solid. Otherwise you have non-manifold geometry, (either has holes, or some flipped prims and what not)

jacob clark - big thanks. I try detect holes and outbound primitives with “Divide SOP” node - excellent. It work.
Now i can recreate my network with new solving algorithm. Big thanks to everyone.

… but “shattering” process is very slow now Yes, geometry is clean, but resulting solving time is too long.
cookie SOP from Houdini 9.1 take longer time for solving, instead of cookie SOP from 9.0.


..May be i need new fast processor