A Semantic Quantum Circuit Cache for Scalable and Distributed Quantum-Classical Workflows
New semantic caching system speeds quantum-classical workflows 11.2x on real hardware
A team led by Mar Tejedor from the Barcelona Supercomputing Center has introduced a Semantic Quantum Circuit Cache designed to eliminate redundant computation in hybrid quantum-classical workflows. The system combines ZX-calculus reduction with Weisfeiler-Lehman graph hashing to generate deterministic identifiers for circuits, enabling constant-time lookup across distributed caches like LMDB and Redis. It remains backend-agnostic, working seamlessly across CPU, GPU, and QPU environments.
Evaluated on the MareNostrum 5 supercomputer, the cache demonstrated significant gains. In distributed wire cutting, it eliminated up to 91.98% of redundant subcircuit simulations, achieving 7x speedups on a single node and up to 1.6x with Redis-based caching under high parallelism. On a 35-qubit QPU, real hardware tests showed an 11.2x speedup. For QAOA optimization, equivalence-aware caching avoided up to 27.6% of circuit evaluations without altering the algorithm. The research highlights that circuit redundancy is a major systems bottleneck, with reuse scaling with concurrency and circuit complexity.
- Eliminates up to 91.98% of redundant subcircuit simulations in wire cutting workloads
- Achieves 11.2x speedup on a 35-qubit superconducting QPU in real hardware tests
- Uses ZX-calculus reduction and Weisfeiler-Lehman hashing for constant-time semantic equivalence detection
Why It Matters
This cache reduces quantum computing costs and latency, making hybrid workflows practical for real-world HPC applications.