Closed sowbug closed 1 year ago
Cf. Pirkle, DSSPC++, p.19. Recommend 64-sample blocks to start with; that's just a 1.4msec worst-case delay.
With 1b59f06b8b542306e548f80180db90609e968f55, we are past the point of basic design and implementation. gather_audio()
still has an opportunity for big improvements; it's still working sample-by-sample. But we're good enough (2.54 usec/sample!) to close this issue.
Today it's one tick/timeslice, which is inherently inefficient because we can't amortize the infrastructure cost across multiple samples.