Bullet vs Houdini RBD

[Erik_JE]

@Erik, now that I thought about it, are you sure convex hull option for rbd sims wouldn't speed up? When I make a hand made convex hull for the same shape with much low res of course, it sims so much faster. Probably 20-50x at least. I think this option would be useful.

But I didn't see any operator to procedurally generate convex hulls. Is there one? (Maybe a VEX SOP?)

You can use Tetrahedralize to generate convex hulls however the faster speed is more likely because of the lower resolution.

EDIT: Speaking of Tetrahedralize it would be nice if it respected constraints if someone ever wanted to go crazy and implement som FEA stuff.

Edited January 9, 2012 by Erik_JE

[magneto]

Thanks will try Tetrahedralize today. Sounds like a good idea.

[edward]

You mean there are still a few people using ATI cards?

I certainly see where you're coming from. On the other hand, I personally hate vendor lock-in. Ah, boy I miss the 3Dlabs guys sometimes.

[rdg]

Ah, boy I miss the 3Dlabs guys sometimes.

Matrox? What about Matrox?

[cyybermax]

You cannot just compare Houdini RBD, Bullet, PhysX, etc.. according speed. Its about the ratio between quality and speed!!! And in this is Bullet BEST! But in some cases you need precision and this is place for Houdini RBD, but If you have massive scene, you use game solvers. Its like a SPH vs. FLIP for fluids, SPH is good for small technical scene(glass of water), because you need the highest quality, but If you want to simulate river, you run FLIP solvers.

And its not about "PhysX use GPU", Its about algorithms, because bullet use very fast methods and many of them cannot be rewrite for GPU, because they arent parallel. This is reason why Bullet programmers dont rewrite all code for GPU, but only some part.

Main problem with bullet is, that Bullet includes many solvers(RBD, Cloth, fluids), but It has terrible support for interactions between solvers.