Closed catfact closed 2 years ago
FWIW, the approach SuperCollider takes is to apply a similar (better) zapgremlins
function to filter I/O, but only once at the end of each processing block. thus preventing gremlins(/denorms) from recirculating, with minimal steady-state CPU cost.
but if possible it would be nice if we could know that all signals on internal busses are always safe.
found the underlying source of bad values instead
seems to fix reported noise issues on recent build.
not ready to merge (and maybe not really the correct approach; e.g. gshould sanitize earlier);
at minimum needs benchmarking and probably optimization pass to offset added overhead