Steps

The steps in one collision detection computation are as follows.

1. Contributors give the local collideMgr their objects to be collided.
2. The local collideMgr determines which voxels each object touches, and calls collideVoxel::add for the appropriate voxels.
3. Once the voxels have received all their objects, the collideMgr objects synchronize.
4. startCollision is broadcast to all voxels, which then run their serial collision detection algorithm.
5. The resulting lists of collisions are collected.

In step 1, the contributors use the input format described in Appendix . As usual in CHARM++, because contributors are almost always parallel objects themselves, there can be several contributors per processor, or none at all on a particular processor.

In step 2, the local collideMgr actually accumulates the triangles to be sent off, then finally sends all the triangles for a voxel at once. This local accumulation is much more efficient than sending a separate message for each triangle, and allows us to use the indexed triangle representation. Splitting up indexed triangle lists efficiently is the subject of Appendix .

The message sends in step 2 actually create the voxels, if they haven't already been created. By default, the creation happens on the sending machine- this makes all future communication between this collideMgr and the voxel efficient. If the triangles contributed on a processor are clustered well, most will never be sent over the network.

The synchronization in step 3 is needed because voxels may receive objects from several processors. Since any processor can send a triangle to any voxel, ``Have all triangles been delivered?'' is a global property, obtainable only via a global operation. Recall that in CHARM++, however, other local objects will automatically use the CPU during this single synchronization.

After the synchronization, each voxel runs a separate serial collision detection algorithm and reports the results.