Open elgiano opened 2 years ago
It's not so much an issue of correctness – as far as I know the parallel process will still honour the selected range (if it doesn't, then obviously there's a bug) – but of redundant copying that is especially problematic when doing lots of small analyses on a single large buffer.
Anyway, no, I haven't fixed this yet. But I will, honest.
From the Oslo workshop some months ago, I remember that buffer slices were not copied correctly across threads: instead of copying only the portion from startSample to startSample + numSamples, the whole buffer would be copied. Is this still the case?
Thanks!