Quote:
Much depends on the type of processing that's needed. In case of Stereo Tool, I have douzens of different filters that all require the output of the previous filters (so they cannot be run in parallel), but also in some cases the output of the previous step of the same filter (so that also cannot be done in parallel). That makes it difficult to use threading in this case. Although some pipelining is possible - but usually GPU's are used to all do the same thing at a different portion of the data.
Thats the problem I found while reading up on DSPs through the GPU. Audio data is hard parallelize... umm whoops lol. you get what I mean.
Anyways the good thing about handing off processing to the GPU would be reduced load on the CPU, would it be feasible to separate the process that draws the GUI and the signal processing?