ft-primitive-bench

name: ft-primitive-bench description: "FTPrimitiveBench methodology for fault-tolerant quantum computing benchmarking. Provides systematic approach for evaluating QEC protocols under hardware-motivated noise models including Pauli bias, measurement bias, and spatio-temporal non-uniformity. Use when: (1) analyzing fault-tolerant quantum computing performance, (2) benchmarking QEC codes under realistic noise, (3) comparing decoders for surface code, (4) studying logical primitive operations (memory, lattice surgery, Hadamard, phase gate), (5) hardware-aware quantum architecture co-design, (6) noisy stabilizer simulation workflows."

FTPrimitiveBench - Fault-Tolerant Primitive Benchmarking

Systematic benchmarking methodology for studying how logical primitives interact with hardware-motivated noise in fault-tolerant quantum computing.

Core Concept

FTPrimitiveBench standardizes the link between noise-model specification and logical primitive construction. It extends memory-only benchmarks to active logical computation, where the interaction between noise structure and primitive implementation matters.

Key Insight: Structured noise (Pauli bias, measurement bias, spatial non-uniformity) affects logical primitives in qualitatively distinct ways, shaped by the interplay between noise model, primitive type, and decoder choice.

Noise Model Families

1. Pauli Bias

• Asymmetric rates for X, Y, Z errors
• Common in superconducting qubits (dephasing dominant)
• Can be exploited by biased-noise codes (XZZX surface code)

2. Measurement Bias

• Different error rates for measurement vs gate operations
• Critical for syndrome extraction fidelity
• Affects lattice surgery and flag-qubit protocols

3. Spatial/Spatio-temporal Non-uniformity

• Position-dependent error rates across qubit array
• Time-correlated errors (drift, crosstalk)
• Captures real device heterogeneity

Logical Primitives

Workflow

Step 1: Define Noise Model

Specify hardware-motivated noise parameters:

• Pauli error rates (pX, pY, pZ)
• Measurement error rate (pM)
• Spatial variation function f(x,y)
• Temporal correlation model

Step 2: Select Primitives

Choose logical primitives to benchmark based on target architecture:

• Memory-only: baseline code performance
• Active computation: include lattice surgery, logical gates

Step 3: Choose Decoder

Select decoder appropriate for noise structure:

• MWPM (minimum-weight perfect matching)
• Union-Find decoder
• Neural network decoder
• Custom decoder exploiting noise bias

Step 4: Run Simulation

Execute noisy stabilizer simulation at HPC scale:

• Vary code distance d
• Measure logical error rate pL
• Track error propagation through primitives

Step 5: Analyze Results

Compare logical error rates across:

• Different noise models
• Different primitives
• Different decoders
• Code distances and depths

Hardware-Aware Co-Design Principles

1. Noise-adapted codes: Choose codes matching hardware bias (e.g., XZZX for dephasing-dominant devices)
2. Decoder selection: Match decoder to noise structure (biased-noise decoders for biased channels)
3. Primitive scheduling: Order operations to minimize exposure to dominant noise
4. Threshold estimation: Compute thresholds under realistic noise, not uniform depolarizing

Key Findings from Paper

• Structured noise affects primitives in qualitatively distinct ways
• Uniform depolarizing model fails to capture real device behavior
• Noise-primitive-decoder interaction determines logical error rate
• Results extend beyond memory benchmarks to active computation

Code Reference

Official implementation: https://github.com/kan-shuwen/FTPrimitiveBench (verify URL from paper)

Activation Keywords

• ft-primitive-bench
• fault-tolerant benchmarking
• QEC benchmarking
• logical primitive analysis
• hardware-motivated noise
• noisy stabilizer simulation
• surface code benchmark
• lattice surgery benchmark
• quantum error correction evaluation
• 容错量子计算基准测试
• 量子纠错码评估