In this post I will describe a small electronics project designed to enable the remote programming of a real Commodore 64 that is switched on and running. To achieve this, several parts are involed, utilising some custom circuitry and communications protocol, a Raspberry Pi, a C program, a 6502 assembler program and a little F# program to finish it off.
This post describes a memory allocation strategy I call a “fixed” size memory pool.
Recently I decided to re-write my virtual machine Drey almost completely in the C programming language, using no external libraries except for the networking via ZeroMQ. This is mostly for fun, to see if I can remember how to program in C again and write a load of low-level stuff (apologies if my C is currently terrible!.)
Drey executes programs written in a much higher level language (typically Scurry), its purpose being to remove all low level details and provide an experience for the author focused on game logic only. As such, the virtual machine implementation must manage the memory itself, not knowing how much of what size memory it will need up front.
Since the system calls malloc/free are relatively slow and will fragment the heap, especially for lots of small object allocations, a memory management system will be required.
(actually, I could probably get away with using malloc since performance doesn’t really matter in Drey, but where’s the fun in that!)