VolksForth/vfvm

Portable VM for VolksForth

• Goal
• Implementation Idea
• Design questions and thoughts

Goal

• create a portable VM to run VolksForth
• multiple implementations (C, Go, Pascal …)
• 64KB Memory + Banked extension memory
• Stack machine CPU (Data-Stack, Return-Stack)
• Stacks mapped into main memory
• Stack-Pointer as memory mapped register

Implementation Idea

• 16bit alignment
• 16bit opcodes
• opcodes $0000-$00FF -> Forth primitives (dup, swap, drop …)
• opcodes $FF00-$FFFF -> extended opcodes

• opcodes $0100-$FEFF -> implicit jump to subroutine

  • allows easy implementation of ITC and DTC Forth
• memory $FFF0-$FFFF -> memory mapped I/O, Bankswitching etc

Design questions and thoughts

• implicit jump to subroutine opcode(s)

  • would use the Forth Return Stack?

  • with 16bit alignment, should we use opcode bit 0 for implicit return from subroutine?

    • pros: obvious, space and time saved
    • cons: a compiled word would have different values with and without following RTS. Some code may depend on the same word always having the same compile token.
  • that would mean non NEST or UNNEST routines, but NEST and UNNEST would be primitives, UNNEST even a 0-byte primitive if we do the bit 0 thing?
  • would not having a NEST routine influence compatibility with ITC VolksForth flavours? Would a colon definition then look just like a Code definition, with in both cases the CFA pointing to the PFA?
  • would the address of the implicit JSR be the CFA or the PFA of the called code?