[SOLVED] Disabling bodies from simulation

[Gianluca]

Hello, (and happy new year)
I need to totally disable some bodies during the simulation for a certain interval and then re-enable them. In this period, I want that the NewtonUpdate speed up as if the bodies simply does not exists.

My first idea was to set a Null Collision to bodies and to set a Null all callbacks regarding the bodies.
In that way, I suppose that the NewtonUpdate should simply skip the bodies and do nothing.
I tested the execution time of an empty NewtonWorld, of a NewtonWorld with 100 bodies active, and of a NewtonWorld with 100 bodies disabled as above.
There is no time difference between the last two conditions. Why ?? If there is no collision on bodies, and if there is no callbacks to call... why NewtonUpdate takes exactly the same time as if the bodies has collision and callbacks set ???

What can I do in order to disable all bodies in a NewtonWorld and reach more or less the execution time as a truly empty NewtonWorld ???
(I would prefer to avoid the destruction of the body)

Thanks,
Gianluca.

[JernejL]

You can try freezing it and set a material which never collides with any other.

Or just delete the body, i use this and it works ok.

[Gianluca]

I tried freezing it and using a material that never collides... but anyway, the 80% of execution time is spent on collision handling... and the time needed for one step of NewtonUpdate is the same as the bodies are not freezed and with the default material.
That's way, my second test was to set also a Null Collision and to nullify any callback.
But the execution time remain the same.

Ok, at the end I can destroy the body... but I prefer to avoid that solution.

Anyway, I'm wondering what's the difference between an empty Newton world and a newton world with bodies without collisions and callbacks.
Why in the latter case netwonupdate spend time on collision handling and dynamics updating ??

Thanks,
Gianluca.

[JernejL]

i do destroying & creating of bodies at runtime, it works really fast.

[Julio Jerez]

My first idea was to set a Null Collision to bodies and to set a Null all callbacks regarding the bodies.
In that way, I suppose that the NewtonUpdate should simply skip the bodies and do nothing.
I tested the execution time of an empty NewtonWorld, of a NewtonWorld with 100 bodies active, and of a NewtonWorld with 100 bodies disabled as above.
There is no time difference between the last two conditions. Why ?? If there is no collision on bodies, and if there is no callbacks to call... why NewtonUpdate takes exactly the same time as if the bodies has collision and callbacks set ???

that was true of sdk 1.xx, in 1.xx inactive bodies where keetd in a separate list, and each body had a pointer to the list node.
Sicne the list where of the same kind, it was constanat time to move a body from one list to the next.
that made newton take almost zero time in a update popultated with bodies with NULL collision or sleeping bodies.
Newton 1.xx did not have a Force And TorqueCallback loop, activation and diactivation were handled by moving bodies from one list to the next.

with Newton 2.00 one for the main issue is mutithreading, so the above scheme made activation and dactivation happen in a critical section
and what became the become the biggest performnace bottlenect on the engine.
Newton 2.00 uses a much simpler approach, the body has a flag that indicate if it is sleepping or not.
then there is a loop that go over the list calling force and torque callback (even if the body is sleeping) and take care of the sleepping state of the body.
The loop is runned uncoditianlly on each call to newton update. and it is uses as the hartbeat of the simulation.

As it is now there is not way the newton 2.00 can ignore bodies.
what we can do is add the funtionality to remove a body from the simulation,

something like:
NewronRemoveBody(world, body)
NetwonInsertBody(world, body)
this can maybe be used to move bodies from one work to anotherr which is a funtionality Newton never had.
and it will handle NULL bodies, because make a body with Null collision deos no mean it sodu no get force and torque callback.

does that sound like a solution to you?

[Gianluca]

Thank you Julio... but my idea works.
When I posted there was a bug in my code that slow down everything.
But now it works.
Essentialy, if I freeze a body and I set a null collision and I unset the callbacks... NewtonUpdate as the same speed is the body would be destroyed

I edited the title to [SOLVED]... but I forget to write how.

However, If you would add the two function you mentioned... it would be even better.

Thank you,
Gianluca M.

[Julio Jerez]

Ok I will add those functions this satuday, it should be very simple.

I did not now why I did not think of that.
That will be of great help for people doing mutiplayer worlds with mutiple worlds
if they can move a body from one world to the other

[Gianluca]

Thank you very much.

A final favour, the project where I use newton is shared amongs variour laboratories... so, before I can update the newton I need to get newton libraries for all platforms... there are strict rules that I have to respect, and I cannot update if the same version is not available on all platforms (win, lnx, mac).

Can you compile the mac version as soon as possibile ???

Thank you very much,
Gianluca.

[Julio Jerez]

I will compile the Mac this weekend.

Th ereason is is delayed is that need to make the demo framework usin the Mac GUI since I abandome wxWidget.
however I still can compile all the libraries and post the SDK.

[collerblade]

Julio
We need this functions. I could not find in the new version (2.30). Have u implemented them?

Code: Select all

NewtonRemoveBody(world, body);
NewtonInsertBody(world, body);

And i have another question. How can i awake the slept bodies?


Thx CollerbladE

[Julio Jerez]

Oh with the Opensource stuff I fogot.
I will add it this weekend.

[KingSnail]

Bumping this because I dont think this is implemented yet?
Have you forgotten about this?