What to do when analysis takes too long


Take fluid.buftransientslice~ for instance. Use the following params:
order 200, debounce 25, padding 123, skew 0, threshback 1, threshfwd 1, windowsize 20

The analysis takes a few minutes. Should one get a warning message or something else or nothing ? This is just a question I have no answer for.


Hi, I agree it is a bit weird, I had the same feeling in SC. For iterative procedures (e.g. NMF where you can specify the number of iterations) one possibility would be to print the iteration number, but maybe this would pollute the max/sc console too much…

In the specific case of the transient seperator (by far the most expensive thing we have, for certain combinations of settings), the most time consuming phase isn’t iterative in a way we can currently report on (it’s solving some huge matrx system). We need to look again at whether that particular solver is our only option, and should also consisder how give feedback in long cases.

The time consuming phase is non-reportable, but progress through the buffer is very easily reportable. That will only be possible when (and if) we run the processing on a separate thread, however, as if the main max thread never releases the graphics won’t update.