Use this page as the reader-facing entry point. The guides are organized by task and subsystem rather than by individual routines. The machine-readable documentation corpus and corpus review record ownership, consolidation, and the few intentionally retained filename families.
Start with the README, then use the verification contract for the complete command sequence. The toolchain contract, runtime evidence, and debugger workflow explain the pinned tools and the evidence produced by those commands.
Release boundaries are recorded separately for Source Reconstruction 1.0, 2.0, and the 2.1 candidate. The revision profiles describe the USA, Europe, and Japan inputs and capabilities. Code quality and the roadmap distinguish accepted behavior from later work.
Read architecture for execution and data flow, then use source layout and source organization to locate the owning module. RAM map defines runtime storage and data formats defines encoded records and their codecs. Editing in Retrolarium explains how the game is edited.
The runtime is documented by subsystem:
- Startup, NMI runtime, cooperative scheduling, and the machine-oriented scheduler entry table.
- Gameplay runtime, player actions, objects, enemies, and items, timer, and score.
- Room lifecycle, room data pipeline, map interactions, special-room scripts, Demon Mirror runtime, and the ending system.
- PPU update pipeline, title screen, CHR-bank policy, and audio engine.
- Attract and demo flow for the non-gameplay presentation path.
Naming and evidence policy, provenance, and the canonical label rename registry separate confirmed names from hypotheses and imported knowledge. Unknowns records unresolved questions. Licensing separates project tooling and documentation from reconstructed source and private user-supplied inputs.