The second is that I can do "SIMD" bounding box checks for collisions.
You might also wonder why I did bounding-box compares instead of using hardware collision detection. Two-fold: I use multiplexing for some of the targets, so the mapping of MOBs to actual individuals changes frame to frame. The other is that, while hardware collision detection is fast compute-time wise, it's slow game-time wise.
- On Frame N, we compute the state that will display on Frame N+1.
- On Frame N+1, the state actually gets displayed. Hardware collision detection happens during display.
- On Frame N+2, we can finally react to the game state we set up on frame N.
Edited by intvnut, Wed Sep 15, 2010 9:53 AM.