TriggerVolume as view frustum culling

[aqnuep]

Hi,

I know this sounds like a pretty awkward idea but I am interested whether trigger volumes would be appropriate for view frustum culling.
I mean I have all my newton bodies and instead of doing the view frustum culling within the rendering engine, I would rather create a trigger volume that has the same extents as the view frustum and decide what to draw whether any of the bodies has contact with the trigger volume aka the view frustum.
What do you think? Would this be a feasible approach or this would be just too slow for real life usage?

[Julio Jerez]

you can certainly do that, but sooner or later you will change your mind, because then by using the physics as culling device that imply tha every thong have to be a pysiocs body.
even a particle, a decal, anything since the physics only understand bodiesm, and Bodies are not lightweigh objects.

I recomend a 2d multiresultion grid for your frustum culling, and you add a proxy as you representation of the visual object.
In general these day you do not really have to cull too many objects since video cards are so good, so your mutgrid cells do no really have to be too presice.

This is opposit to a mutigrd for a physics engine because a mutidrids that give a positve that is false in a test,
leads to a much more expensive test than one that gives the same resutl in a mutigrid for rendering.
This because the output of e Frutum culling for render jst submin t he object to the graphics for rendering, this is linear on the number of objects in the output.
but a output of a mutigrid for physics goes to a secund phase called narrowphase where each object is tested against every other object in of the putput,
this is quadratic on teh number of objects in the output.
this justify that a mutigrid for rendering can have a very course resultion.

[aqnuep]

Yes, that's what I was worried about as even though the broadphase is done using AABBs in Newton, the narrowphase collision detection may still have a performance impact.
Btw, I know how to make hierarchical culling using rough octrees or whatever other space partitioning just I was interested whether this would be feasible.
The fact that I need some sort of newton body or collision for my objects wouldn't really matter, however the narrowphase really makes me think that maybe this whole idea is not a very good one.

Thanks for the information!