@amigakit
On reference to the efficiency I was commenting on the data flow within the Mediator.
Now that's interesting statement. Last I checked if CPU has to access the sample data, it must be in system memory. To transfer this to the audio card buffer requires moving it with CPU, since no DMA can be done directly.
Sure you can do tricks such as moving data to gfxcard memory and DMAing it from there, but really what's the benefit here? You still need to move the data using CPU to gfxcard buffer then...
Thus, direct paula audio is more efficient even when considering "data flow".
The only real benefit I can think of is sound quality, and easy mixing of several audio sources (cd-rom audio for example).
