Skip to content

Document layered QEC abstractions and end-to-end workflows #516

Description

@qciaran

Summary

Update PECOS architecture documentation and examples to describe the agreed layered QEC model, lowering boundaries, experiment/noise attachment points, and end-to-end workflows.

Parent: #509
Architectural discussion: #508

Motivation

crates/pecos-qec/docs/levels-of-abstraction.md already distinguishes abstract codes, geometry, circuits, and programs. The RFC expands this into a coherent refinement pipeline reaching target-native timed execution and clarifies that code-capacity, phenomenological, and circuit-level models are experiment overlays rather than different code types.

Documentation should guide both users choosing the right abstraction and contributors deciding where new functionality belongs.

Scope

  • Revise the abstraction-level document after the RFC reaches an initial decision.
  • Document responsibilities, inputs, outputs, and invariants for each persistent representation.
  • Explain lowering and source/provenance maps.
  • Explain where geometry, decoders, Pauli frames, detector models, noise, and calibration data belong.
  • Distinguish logical programs from executable device programs to avoid overloading "program."
  • Add diagrams and API examples only where they materially clarify the relationships.
  • Document qecdb as an abstract-code catalog input.
  • Include the HDL-inspired frontend model if that exploration is accepted.

End-to-end examples

At minimum, document:

  1. Local stabilizers -> validated abstract code -> code-capacity analysis.
  2. Abstract code -> syndrome-extraction plan -> synthesized/verified circuit -> circuit-level simulation.
  3. QEC implementation -> target mapping -> native timed schedule with provenance.
  4. qecdb record -> abstract code -> independently selected PECOS protocol.

Acceptance criteria

  • Documentation matches implemented public APIs and passes existing docs-example tests.
  • Each abstraction has a concise "use this when" description.
  • Noise/experiment models have explicit attachment points.
  • At least one complete executable example retains traceability from abstract checks/logicals to emitted measurements and results.
  • Outdated examples using superseded type names or boundaries are migrated.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions