[ODE] Performance patch : not checking collision between disabled bodies

Konstantin Slipchenko slipch at gsc-game.kiev.ua
Thu May 27 11:31:23 MST 2004


NB>       Well, interesting case! But I had experimented this and it works
NB>       just fine.
NB>       As long as the small boxes enabled, right after that they will
NB>       be fed to collision detection code.



Hello Nguyen,

It must depend on different things.
In my case one box standing on a plane (triangle mesh)
have problems if I switch off collision with static
when it disabled. Pushing the box by something which represent
character I enable it on each step. If I can not move it by some
reason suppose too heavy  the box then disabled on each step.
It becomes to jitter and it is very noticeable. The case itself is
not a real problem this can be fight by different way. But as for me
it proofs that if we not set some needed contacts even for one step we
may have problems. And I agree with others on that is not the best and
full solution of the problem.
As was said before in STALKER we enable objects in collision phase -
seems to be better approach.

If we do apply the patch I vote to make it optional.


Thursday, May 27, 2004, 5:20:57 AM, you wrote:


-- 
Best regards,
 Konstantin                            mailto:slipch at gsc-game.kiev.ua

NB> Hi Konstantin,


KS>> Imagine a stack of small boxes which are already disabled.
KS>> Another big box touche them simultaneously from stack side.
KS>> As far as only the big box was enabled during colliding we create
KS>> contacts between big box and small boxes but do not create contacts
KS>> between small boxes. While WorldStep all boxes will be enabled and
KS>> boxes in the stack start to fall knowing nothing about each other and
KS>> static environment. If we have not too small stepsize and world
KS>> gravity the stack would explode in this case.

KS>>   Colliding:
KS>>              ______
KS>>   [disabled]|      |
KS>>   [disabled]|enabed|
KS>>   [disabled]|      |
KS>>   [disabled]|______|

KS>>  WorldStep
KS>>            ______
KS>>   [enabed]|      |
KS>>   [enabed]|enabed|
KS>>   [enabed]|      |
KS>>   [enabed]|______|









More information about the ODE mailing list