Static copy

Copied from docs/en/architecture/bytecode-and-vm.md in https://github.com/poppy-team/aipo-lang (MIT). Pinned to revision ce47c4e070404fc41722ccd8e2b7b808b99062a8, blob 6c6c5de8b91116f335617993bc0371f786a6d4c7. The source repository remains canonical; this copy is refreshed through a sync pull request, not live.

Bytecode & Virtual Machine

Aipo executes natively on a deterministic bytecode virtual machine, optimized for low memory footprint and high-throughput instruction dispatch.

---

The .aibc Binary Format

The Aipo compiler can serialize bytecode chunks directly into an .aibc binary archive:

• Magic Header: 4-byte signature AIBC followed by format version integer (currently 1).
• Constant Pool: Integer literals, floating-point numbers, normalized strings, and function prototypes.
• Debug Metadata Table: Bytecode instruction offsets mapped to source code lines and columns.
• Code Section: Compact binary encoding of virtual machine instructions.

---

Virtual Machine Architecture (aipo-vm)

1. Value Representation (Value)

The core Value enum is compact (48 bytes) with heap variants reference-counted via Rc:

• Primitives: Int(i64), Float(f64), Bool(bool), None
• Reference-counted heap variants: String(Rc<str>), List(Rc<RefCell<Vec<Value>>>), Dict, Set, Bytes
• Execution entities: Closure, BoundMethod, TaskHandle, HostHandle

2. Dispatch Loop Optimization

• Instructions are decoded directly via flattened match branches, eliminating bulky 72-byte Result allocations on every CPU cycle.
• Inline monomorphic caches for struct field resolution, accelerating property lookups by up to 28%.
• Formal bytecode verifier asserting local slot bounds, upvalues, and frame depth limits prior to execution.

3. Integrated Disassembler

The aipo disasm command disassembles bytecode instructions and maps each mnemonic directly to its originating source line, streamlining profiling, debugging, and security audits.