R-ten-K wrote:Well for 15MB images @ 24 fps you are talking around 300+ MBps of disk bandwith, if that is where the images are stored. That is not a trivial IO, so it almost looks as if you are going to be IO bound.
It's certainly not trivial, but my target market generally has RAID/SAN setups that can handle that.
R-ten-K wrote:In any case, why are you using OpenGL to do image processing?
Primarily for speed, but also for portability, and to take advantage of the existing graphics primitives within OpenGL. I've written code in C++ that can do some fairly fast processing, but nowhere near as fast as what modern gpus can chew through. To achieve that kind of speed I would have to either use assembly and specialized cpu instructions (losing portability), or have to distribute the processing across multiple cpus (which adds quite a bit of complexity).