I did this for my non apple 6502 machine - the issue I always had was how much space to put in between subroutines so I could change them later and not have to redo all the jumps. I think after a while I just did one 256 byte block per subroutine and screw memory efficiency (not for the last time in the following forty five years). The only time I crunched it back down was for a “use almost all 8K for the decimal digits of E” (which was fun and much faster than Pi but not totally useful as I couldn’t print them or save them to tape).