Re: trimesh error

Hi all,

thanks for your replies, I have made small summation of replies.

Irrisor writes:

Looks
like too low contact count. Raise the maximum allowed contacts to avoid
this (this applies to OPCODE and GIMPACT btw).

Yes, that was a problem. I have set max_contact to 10, and it was a way
to low for OPCODE and trimeshes. When I raise it to 32 simulation are
looking like it should.

Jon Watte (ODE) writes:

However, it's still a good idea to check the
data. Both for correctness in orientation/translation (for example, ODE
uses WXYZ order for quaternions, if it matters), and for correctness in
building (If you use the "simple" build functions, your triangle verts
should be float x, y, z, q -- yes, a fourth, unused component per
vertex!), winding order, etc.

Im using dGeomTriMeshDataBuildSingle1(...), and verts have fourts
components per vertex. BTW. is there an simple way to switch in VS
between GIMPACT and OPCODE library in my project. For now I have made
two solution configuration with two different ODE library builds (one
with OPCODE and one with GIMPACT), added #if dTRIMESH_OPCODE when
defining StridedVertex and StridedNormal. Now when I want to switch to
other trimesh collisions, I switch to appropriate
solution configuration and change config.h file. But is there any way
to omit need of changing config.h file?

That being said, I'm not sure how much testing
plane vs trimesh has gotten. It's not the most common combination.
Typically, trimeshes are solid and unmovable -- not just in ODE, but in
most game engines. Actors are treated by proxies (capsules, boxes,
etc).

Im not making a game engine. Its a plugin, and part of it is physics
simulation. I know that proxies are best solution, but I would also
give possibility
to use meshes in it.

Mark
Williams writes:

In my experience GIMPACT trimesh collisions are far more stable than
OPCODE's. I'd suggest using them instead.

Bram Stolk writes:

I would try GIMPACT.

I have used it, and even with 10 max_collisions points result looks ok.
Now I have to make decission what is better/faster: OPCODE with 32
max_collisions poitns or GIMPACT with 10 ? Someone said that GIMPACT is
about 20times slower that OPCODE, so maybe stay with OPCODE and give it
64 max_collisions points-this still should be faster than GIMPACT ?

Thanks for all of you, and yours help. It really improve my
simulations.

Krystian

_______________________________________________
ODE mailing list
[email protected]
http://ode.org/mailman/listinfo/ode