Skip to content

RESEARCH.md

Latest commit

 

History

History
127 lines (104 loc) · 7.08 KB

RESEARCH.md

File metadata and controls

127 lines (104 loc) · 7.08 KB

Research Directions and Required Research

This document lists research topics needed to complete the specs and deliver a working static recompilation pipeline.

Core Research Directions

1) Binary Formats and Loading

  • Identify Switch executable formats used by retail titles.
  • Map sections, relocations, and symbol handling.
  • Determine how overlays or dynamic code segments are represented.

Needed research:

  • Public documentation or reverse engineering notes on Switch executable formats.
  • Tools and libraries that can parse these formats reliably.

2) CPU ISA and Semantics

  • Confirm the exact ARMv8-A subset used by Switch titles.
  • Enumerate required SIMD and floating point instructions.
  • Define precise flag and exception semantics for those instructions.

Needed research:

  • ARMv8-A reference manuals.
  • Instruction usage analysis on sample titles.

3) OS, Services, and IPC

  • Identify essential system services required for boot and main loop.
  • Define a minimal service surface for early milestones.
  • Determine which services can be stubbed without breaking games.

Needed research:

  • Public technical references on Switch OS services.
  • Trace-based profiling of service calls for selected titles.

4) GPU and Graphics

  • Confirm GPU architecture and the command stream interface used by games.
  • Assess feasibility of command stream interpretation vs shader translation.
  • Identify core texture formats and shader features.

Needed research:

  • GPU architecture references.
  • Shader translation tooling that can be adapted.

5) Timing and Determinism

  • Determine which timing-sensitive behaviors must be modeled.
  • Identify stable methods for trace collection and replay.

Needed research:

  • Prior work on deterministic replay in emulation or recompilation.
  • Instrumentation approaches for low overhead tracing.

6) Audio, Input, and I/O

  • Identify the minimal audio path needed for gameplay.
  • Map controller input handling to the runtime.
  • Define file system access boundaries and safety rules.

Needed research:

  • Audio subsystem assumptions in Switch titles.
  • Input service behavior and common usage patterns.

7) Legal and Preservation Policy

  • Define acceptable inputs, distributions, and legal compliance expectations.
  • Ensure documentation is consistent with preservation goals.
  • Follow and update docs/LEGAL-POLICY.md as policy decisions are finalized.

Needed research:

  • Jurisdiction-specific rules affecting preservation.
  • Best practices for open source preservation tooling.

8) Automation, Input Replay, and Media Normalization

  • Define a stable automation loop for intake, build, capture, and validation.
  • Normalize reference media and capture outputs into comparable artifacts.
  • Model deterministic input replay aligned to reference timelines.

Needed research:

  • Capture tooling behavior and determinism guarantees.
  • Input timing and latency characteristics for Switch titles.
  • Video/audio similarity metrics and drift analysis.
  • Operational checklist for title-a24b9e807b456252 validation artifacts (see docs/title-a24b9e807b456252-validation-prereqs.md).

9) LLM-Assisted Decompilation Long-Tail Strategy

  • Define when similarity-guided scheduling beats difficulty-only scheduling for unresolved functions.
  • Identify specialist lanes where domain tools materially improve output quality (graphics macros, math transforms, mechanical cleanup).
  • Determine guardrails needed to keep unattended agent loops policy-compliant and reproducible.

Needed research:

  • Compare opcode-sequence distance tools (for example Coddog-style bounded Levenshtein) against embedding-based retrieval for candidate ranking.
  • Measure whether cleanup/documentation passes improve downstream success on similar unresolved functions.
  • Track attempt distributions and outlier classes (for example large functions, graphics-heavy functions, matrix/vector math).
  • Define retry budgets and escalation rules that reduce token burn without masking hard blockers.

Seed Resources (Reviewed)

Research Deliverables

  • A research summary for each category with sources.
  • A list of confirmed requirements to update the specs.
  • A compatibility matrix based on real title traces.

Open Research Questions

  • What is the minimal instruction coverage needed for a first title?
  • Which OS services are required to reach a game loop without patches?
  • What is the simplest graphics path that still produces correct output?
  • How can we generate reference traces without distributing proprietary content?
  • What is the minimal input script fidelity needed for stable validation?