Architecture¶
LiDMaS+ is organized as a modular simulation stack with clear separation of concerns:
- Lattice and code construction.
- Noise generation (Pauli and hybrid CV-discrete).
- Decoder plugins.
- Threshold runner and statistical aggregation.
- Plotting and reproducibility scripts.
Decoder plugin model¶
Surface decoders are exposed via a shared interface and registered through plugin infrastructure:
mwpmufneural_mwpm
This enables controlled decoder swaps without changing the experiment harness.
Noise modes¶
pauli: discrete error-rate sweeps overp.hybrid: CV displacement sweeps oversigmawith 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