clampTriangleCount about nvcloth HOT 1 CLOSED

I don't think it is currently possible to use more collision triangles with the gpu solvers since we don't have a code path that doesn't use shared memory.
Even if you could the performance of using a lot of triangles is not going to be very good.

You can't bypass the limit of collision triangles by splitting the cloth in simpler cloth objects, as (in the general case) it won't reduce the number of triangles against which the simpler cloth objects need to collide.

You could rewrite the collision code to not use shared memory and live with the slow down, but this is not a small task.

from nvcloth.