一句话：提出 Quasar，用双 e-graph 的等式饱和在限定重写步数内生成量子电路的最低代价优化。
量子电路优化需搜索由门消去、交换等局部重写产生的巨大等价电路空间，现有优化器只能探索极小部分且常得到次优结果。本文提出 Quasar：它为量子电路的图表示与序列表示构建两个互补 e-graph，推导可同时作用于二者的原子重写集，并用新技术保证等式饱和的可靠性与可扩展性，从而在给定重写迭代上界内返回可达的最低代价电路。标准基准上，2 量子比特门数、总门数和电路深度分别几何平均降低 20.2%、33.5%、24.2%，保真度提升 21.4%，在四类指标上分别有 75%、92%、62%、80% 的电路优于或持平最先进的基于重写的优化器。
Optimizing a quantum circuit is hard because it requires exploring a vast space of functionally equivalent circuits, produced by applying local circuit rewrites such as gate cancellation and commutation. Each additional rewrite can exponentially expand the space of equivalent circuits, so existing optimizers can only explore a tiny fraction of this space and often produce suboptimal results. We present Quasar, a new quantum circuit optimizer that can generate a step-limited optimal circuit: given a bound on rewrite iterations, it returns the lowest-cost circuit reachable within that bound. To achieve this, Quasar constructs two complementary e-graphs for the quantum circuit’s graph and sequence representations. Quasar then infers an atomic rewrite set for application on both e-graphs, which improves optimization performance by orders of magnitude. Finally, Quasar employs a series of new optimization techniques to ensure both soundness and scalability of equality saturation on quantum circuits. Across standard benchmarks, Quasar achieves geometric-mean reductions of 20.2% in 2-qubit gate count, 33.5% in total gate count, and 24.2% in circuit depth, and a 21.4% fidelity improvement over unoptimized circuits, outperforming or matching state-of-the-art rewrite-based optimizers on 75%, 92%, 62%, and 80% of circuits, respectively.
@article{YangRTG26,
title = {Equality Saturation for Quantum Circuit Optimization},
author = {Ganxiang Yang and Paige Raun and Runzhou Tao and Ronghui Gu},
journal = {Proceedings of the ACM on Programming Languages},
volume = {10},
number = {PLDI},
year = {2026},
doi = {10.1145/3808254}
}