For folks who are curious about this sort of thing and want an approachable starting point, I would recommend:
https://www.goodreads.com/book/show/44882.Code
It would be way cool to see an actual application which wanted this sort of speed optimization --- the last significant assembly language program I can recall using was WriteNow, which was ~100,000 lines of assembly and to this day is my favourite word-processor (well, the NeXT version --- the Mac, even v2.0 suffered in comparison for not having access to Display PostScript and Services).
Really wish that there was a writeup of it at folklore.org --- unfortunately, it only gets a single mention:
https://www.folklore.org/The_Grand_Unified_Model_The_Finder....
(or that there was an equivalent site for the early history of NeXT)
replies(1):