Issue on PSSM and Physics

Mar 25, 2009 at 10:45 AM
Hi, thanks for providing this great framework!

I found that in some case of CalculateCameraSplitFrustum (PssmEffect.cs Line 882),
the input parameter "nearPlaneDistance" is equal to "farPlaneDistance" which caused an ecxeption.
The case was both equal to 0.1.

In PhysicsComponent, when I increased the "Simulation Speed", the objects become to penetrate each other very easily.
Finally I found that in PhysicsComponent.cs Line 481, the target step time would increase as the simulation speed.
I tried to make the target step time unchanged, the objects become stable again (Not penetrate each other easily).
It would increase physics update count, but result is more accurate.

I like this framework very much, keep the good work!
Mar 28, 2009 at 7:05 PM
Thanks for trying out the framework!

I had just ran into the near plane distance = far plane distance exception as well.  D'oh!  I'll be looking at this to determine what is going on.  I'm planning on removing the requirement of passing the bounds of scene to be rendered (PssmEffect.BeginPssm(BoundingBox bounds) and not requiring the whole scene to be drawn for each split.  These two changes may make this exception a thing of the past.

I'll look into the simulation speed issue with the PhysicsComponent.   You are definitely correct.  Thanks for information!!!
Apr 13, 2009 at 9:24 AM

Currently I try to match CollisionShapes to some buildings in game.
I found that CollisionBox and CollisionSphere are hard to match them all.
(I would not use CollisionMesh due to performance issue, as I hope to place quite a large number of buildings)
It would be nice if CollisionCapsule / CollisionCylinder is added.