I do have a handle on interpolated writes now - someone at work explained it to me & it’s essentially identical to read (with a bit of a mindflip). Also I think the interpolated read/write can be sped up enormously using a 4 256-value lookup tables (because inter-sample timescale is only 256). That would help with grains too…
But I’m frantic now about improving the controller side - won’t be able to focus on this properly till a better plan emerges on the avr32/bees side of things…