[rep_lodsb]

The 80186 already implemented the loop for multiplication and (unsigned) division in dedicated logic that could do 1 bit per cycle.

The microcode only had to set up the registers, test for overflow, and undo the final subtraction if it underflowed (since it now used a non-restoring algorithm).

Starting with the 486 there is likely no longer any microcode involved, and newer chips can handle multiple bits per step. See end of article, the first algorithm to do this comes from 1957!