overview
SimpleVM is a compact virtual machine: a fixed instruction set, a stack and registers, a byte-addressable memory, and a fetch–decode–execute loop. It runs programs compiled from a small assembly dialect. The scale is the point — the whole machine fits in your head, which is precisely what real runtimes refuse to do.
motivation
Every language I use bottoms out in some VM or CPU model I'd never opened. I wanted to close that gap from the simplest end: build the interpreter, the memory model, and the instruction cycle myself, and see which performance intuitions survive contact with the dispatch loop.
the problem being explored
- What a machine actually is: state plus a transition function, nothing more.
- Why instruction dispatch is the hot loop, and what it costs.
- How a memory model — segments, bounds, alignment — shapes what programs can say.
- Where higher-level constructs (calls, closures, exceptions) would have to live.
approach
The core is a switch over opcodes in a tight loop, with memory as a flat
byte array behind bounds-checked accessors. A small assembler (Python) turns
a readable ISA into bytecode, so writing test programs is quick. Each
instruction's semantics are documented next to its implementation — the ISA
reference and the source can't drift apart.
what i learned
- The fetch–decode–execute loop is the whole story; everything else is ISA design.
- Bounds checks are where safety, simplicity and speed negotiate in the open.
- Writing assembly for your own machine reveals how much syntax is kindness.
- This project quietly reorganized how I read performance advice — branch prediction, cache lines and dispatch cost stopped being folklore.
repository
Source and notes live on GitHub.