Originally Posted by j_s
computer geek talk...
the dsp chip on the graphics card will know how to do some very specific functions, like transforming 2 and 3 dimentional matricis, and other hard and weird math. every chip has different instructions, even for doing similar things. so a program that uses it would have to know all the nasty details.
for a music program had to know how to use the dsps, it would mean a lot of programming work, and for every new chip, the program would have to be rewritten.
so, the best approach would be to create a software layer between the dsp and the music program. then the music program can have a function that says "apply reverb", which is passed to the abstraction layer, which then tells the dsp to do a bunch of crazy math, and passes the answer back. if a new dsp chip was developed, the abstraction layer would need to be rewritten, but the music software would stay the same.