Published September 21, 2026 | Version v3.0.0
Software Restricted

Qector Mega: IonQ Superion 256 Walking Cat Decoder (v3.0.0 Frozen Release)

Authors/Creators

  • 1. iD01t Productions

Contributors

  • 1. iD01t Productions

Description

Qector Mega: IonQ Superion 256 Walking Cat Decoder (v3.0.0 Frozen Release)

Distribution Package Manifest & Technical Release Summary

  • Access Tier: Restricted (Proprietary / NDA Evaluation Only)
  • Distribution License: LicenseRef-Proprietary (Redistribution Strictly Prohibited)
  • Metadata Attribute | Specification
  • :--- | :---
  • Release Version | 3.0.0
  • Author | Guillaume Lessard (ORCID: 0009-0000-3465-3753)
  • Affiliation | iD01t Productions
  • Programme | qector-decoder-v3
  • Target Hardware | IonQ Superion 256
  • Target Architecture | Walking Cat (arXiv:2604.19481)
  • Python Support | CPython ≥3.9 (x86_64 / AMD64), via PyO3/maturin
  • Build System | Cargo (Rust 1.75) + maturin ≥1.4

Abstract

qector-mega v3.0.0 is a frozen, production-grade decoder release calibrated for quantum error correction on the IonQ Superion 256 Walking Cat architecture. This release integrates PyO3 Python bindings, high-performance Rust source modules across CPU (AVX2/AVX-512/VNNI vector extensions), CUDA, OpenCL, and FPGA (SystemVerilog RTL) backends, GNN pre-decoder modules, gRPC/MCP server bridges, and formal verification theories (Lean 4 and Isabelle/HOL).

Core Verification Invariant

$$H \cdot c \equiv s \pmod 2$$
Where $H$ is the parity-check matrix, $c$ is the recovered correction vector, and $s$ is the observed syndrome vector.

Supported Quantum Codes

Code Designation Code Parameters [[n,k,d]] Configuration Details
Q70 $[[70, 6, 9]]$
Standard parity configuration (Mean latency: $45.51\text{ }\mu\text{s}$ / $p99$: $63.69\text{ }\mu\text{s}$)
Q102 $[[102, 22, 9]]$
Bivariate-bicycle block code configuration (Mean latency: $107.30\text{ }\mu\text{s}$ / $p99$: $141.63\text{ }\mu\text{s}$)
Q54 $[[54, 2, 10]]$ Optimized high-rate sparse code configuration
Gross-144 $[[144, 12, 12]]$
Gross bivariate cyclic configuration ($t=5$ error correction radius)

 

Decoder Suite

Includes Blossom, FastUnionFind, SparseBlossom, BP-OSD, LookupTable, and a hybrid Cascade decoder, alongside dedicated IonQ Superion real-time decoding paths (ionq_superion_decoder, superion_ultra) calibrated for sub-110 µs hardware clock cycles.

GPU Kernel Throughput — Surface Code Ladder (Tesla T4, OpenCL)

Distance (d) Batch Size Kernel Time (ms) Throughput (syn/s)
$d=3$ 65,536 0.23 225,760,447
$d=3$ 1,048,576 3.47 296,070,078
$d=5$ 65,536 0.14 311,933,592
$d=5$ 1,048,576 2.33 437,421,001
$d=7$ 65,536 0.43 130,809,857
$d=7$ 1,048,576 14.14 73,808,337
$d=9$ 65,536 0.99 61,750,916
$d=9$ 1,048,576 15.32 68,094,745
$d=13$ 65,536 2.04 31,125,862
$d=17$ 65,536 4.35 14,636,573
$d=21$ 65,536 6.57 9,870,540
$d=25$ 65,536 12.90 5,051,143
Peak measured forward syndrome extraction throughput: 437 million syndromes/sec ($d=5$, batch $1,048,576$).

Runtime Stability vs. Physical Error Rate ($d=7$, batch $= 131,072$)

Physical Error Rate (p) Kernel Time (ms) Throughput (syn/s)
0.001 1.13 105,205,029
0.010 1.82 66,956,345
0.080 1.90 66,500,625

Logical-Error-Rate Data

Tested by Guillaume Lessard ($p=0.001$, seed=42):
Distance (d) Shots Errors Logical Error Rate
3 896,273 1,000 $1.116 \times 10^{-3}$
5 1,000,000 114 $1.14 \times 10^{-4}$
7 1,000,000 10 $1.0 \times 10^{-5}$

 

Citation Format

@software{lessard2026qectormega,
  author    = {Lessard, Guillaume},
  title     = {Qector Mega 3.0.0: IonQ Superion 256 / Walking Cat Decoder (Frozen Restricted Package)},
  year      = {2026},
  publisher = {iD01t Productions / qector.store},
  version   = {3.0.0},
  note      = {ORCID: 0009-0000-3465-3753. Proprietary evaluation bundle for IonQ Superion 256.}
}

Files

Restricted

The record is publicly accessible, but files are restricted. Log in to check if you have access.

Additional details

Related works

Is documented by
Software documentation: 10.5281/zenodo.22684231 (DOI)
Is source of
Dataset: 10.5281/zenodo.22678419 (DOI)

Dates

Submitted
2026-09-15
Software