I remember writing this once upon a time,.....

Now, trying to build recursion into the AS/360 w/ out a stack pointer and last op code position register. that cut down the usable register count by 2 per call and with 14 registers, there was no way to make recursive calls a useful algorithm design w/ out altering program memory in dangerous ways.

Regardless, could you theorize better ways to write more efficient byte code instructions based on chip architecture rather than general purpose design? Can you work in FPGA??