Make sense. So all that's missing is a more efficient way to feed a (constant
Originally Posted by bridgman
I assume) shader program to the GPU, or an easier way to kick said shader into
action once another frame is ready.
I think so. First-cut code tends to be simple and safe -- you don't want to start putting lots of optimization paths in until the basic code does what you want all the time.
Code gets big and hard to understand very quickly once you start adding all the special case optimizations.