What is the denormal problem and how does Max/MSP get around it?
The denormal problem is the situation which arises when calculations involving extremely small numbers use a very large proportion of the CPU on x86 processors. These small numbers are most commonly the result of some system with feedback, e.g. a delay line, reverb, or IIR filter such as biquad~, allpass~, comb~, lores~,etc. The MSP filter objects and the times~ object (*~ ) on windows have special code to set these small values to zero. As a precaution, if one were to use tapin~/tapout~ or send~/receive~, one should insert at least one copy of times~ (*~) in the feedback chain to avoid these small numbers.
