Dear all
I know 2 works on GPU and PD. André developed something with PD and GPU.
http://www.ime.usp.br/~ajb/wiki/artigos/article-icmc2012-ajb-mqz.pdf
Chuck (Charles Henry) also developed something and he helped Andre to develop his code.
http://www.uni-weimar.de/medien/wiki/PDCON:Conference/GPU_audio_signals_proc...
AFAIK, there is a lib that allows you to copy the processing functions to GPU pipeline and the data to be processed. This lib can identify if there is a GPU available in the system and present the information about it. I think that it really would be a good thing to DSP processing on RPI.
Bests
F Schiavoni
@Hardoff : the patch i'm using in the video requires 16 ms of latency, because it uses phase vocoding. You can drop to 10 ms without it, and I guess even lower would be possible. I use the very first version of the Pi, with half the RAM the new model has. If we can get the GPU to compute the audio i hope that we'll be able to get to really reasonable latencies (to me 6 to 8 ms is really enough to play live). Also, i use the regular Pd fro the debian repos. I can't tell you whether Miller's or Katja's version work better.
Cheers,
Piere.