Midi Import loops single pairs of notes, which is almost never effiencent

[darkychao]

in any case where each note in the pair has a delay shorter than 128 ticks looping them will always save net 0 bytes, as it removes a notevg (3 bytes) and adds loops (2 bytes) and loope (1 byte).
this often leads to notes directly following the pair being notedvg when they otherwise be could be notevg, adding a byte. it also can lead to sets of notes not being grouped into call sections.

[sauraen]

I'm sorry for not responding for so long. At some point I would like to do an overhaul of the optimizer which takes into account the stack and so can do multiple levels of optimization, which would mean there was no penalty for these short loops. The specific situations you described are a bit more complicated than you said, but your general point is correct.