Architecture

LiDMaS+ is organized as a modular simulation stack with clear separation of concerns:

  1. Lattice and code construction.
  2. Noise generation (Pauli and hybrid CV-discrete).
  3. Decoder plugins.
  4. Threshold runner and statistical aggregation.
  5. Plotting and reproducibility scripts.

Decoder plugin model

Surface decoders are exposed via a shared interface and registered through plugin infrastructure:

  • mwpm
  • uf
  • neural_mwpm

This enables controlled decoder swaps without changing the experiment harness.

Noise modes

  • pauli: discrete error-rate sweeps over p.
  • hybrid: CV displacement sweeps over sigma with GKP digitization.

Statistics layer

Threshold outputs include:

  • logical error rate,
  • confidence intervals,
  • defect and correction-weight diagnostics.

The workflow supports reproducible Monte Carlo runs with deterministic trial seeding.

Project layout

include/   # public headers and interfaces
src/       # simulator and decoder implementations
examples/  # reproducible runs and plotting scripts