This repository contains the source code for the Quantum Operating System (QOS), as described in the paper QOS: Quantum Operating System, to appear in Usenix Symposium on Operating Systems Design and Implementation (OSDI) 2025.
-
Qernel Abstraction:
- QOS implements diverse mechanisms across the stack, including circuit optimization, error mitigation, and spatio-temporal multiplexing on QPUs.
- The Qernel abstraction acts as a common denominator for composing the QOS mechanisms in a single stack.
-
Error mitigation:
- To address the low fidelity of current QPUs, the error mitigator composes complementary error mitigation and circuit optimization techniques in a single pipeline.
- Supports circuit cutting (wire and gate cutting), qubit freezing, and qubit reuse.
-
Performance Estimation:
- To guide multi-programming and scheduling decisions on heterogeneous and noisy QPUs, QOS implements performance estimation.
- QOS does not simulate circuits (exponential overheads) to estimate fidelity. Instead, it uses numerical- and regression-based policies.
-
Multi-programming:
- To increase QPU utilization, QOS co-locates multiple circuits on a single QPU (possibly from different users, i.e., supports multi-tenancy).
- QOS increases effective utilization, i.e., spatial and temporal utilization, by taking into account circuit runtimes.
- To mitigate destructive interference between circuits, which degrades performance, QOS implements compatibility functions that quantify how well any two circuits fit together.
- QOS implements buffer zones between circuit allocations, to further avoid destructive interference.
-
Scheduling:
- There is an inherent tradeoff between execution quality (quantum fidelity) and latency (i.e., waiting times/JCTs) when scheduling circuits on heterogeneous noisy QPUs.
- QOS implements multi-objective scheduling policies that strike a balance between these conflicting objectives.
- Description: Implements the techniques from the paper "A Case for Multi-Programming Quantum Computers" by Das et al., MICRO 2019.
- Description: Contains Redis database configuration and calibration data from (now deprecated) QPUs.
- Key Features:
- Provides calibration data for benchmarking and evaluation.
- Includes Redis-based configuration for managing quantum workload metadata.
- Description: Contains data from QOS's evaluation.
- Subdirectories:
/benchmarks:- Description: State-of-the-art quantum benchmarks, including:
- Adder, Bernstein-Vazirani (BV), GHZ, Hamiltonian Simulation, QAOA, QSM, TwoLocal, VQE, and WState.
- Features:
- Configurable circuit structures, e.g., QAOA with regular, power-law, and barbell graphs for Max-Cut problems.
circuits.py: A utility for generating new benchmarks or fetching existing ones.
- Example:
from circuits import qaoa circuit = qaoa(nx.barbell_graph(5, 0))
- Description: State-of-the-art quantum benchmarks, including:
data:- Description: Contains evaluation data, though not all datasets are up-to-date with the paper.
/plots:- Description: Includes the plots used in the QOS paper presented at the conference.
- Description: The source code of the ASPLOS '23 paper "FrozenQubits: Boosting Fidelity of QAOA by Skipping Hotspot Nodes" by Ayanzadeh et al. There are minor adapations to make it work with QOS.
- Description: The main source code for the QOS system. Please check the internal READMEs.
- Description: The main source code for the Quantum Virtual Machine (QVM) project which implements circuit cutting and qubit reuse, which are leveraged by QOS's error mitigator. Please check the internal READMEs.