qfusion-opt: Profile-Informed Gate Scheduling and Fusion Optimization for Accelerating Quantum Circuit Simulation

Quantum circuit simulation is essential for testing and developing quantum algorithms, and recent advances have produced a range of simulators with diverse characteristics. Quantum gate fusion is a powerful optimization technique that improves simulation performance by reducing memory access and computational load. However, existing gate fusion methods usually overlook the performance characteristics of underlying simulation environments, resulting in suboptimal results. In this work, we propose a novel gate fusion compiler framework, qfusion-opt, to address this challenge. By using microbenchmarks to characterize performance across different simulation environments, qfusion-opt adaptively tailors circuit optimizations to the target platform. This approach enables higher simulation performance without requiring detailed knowledge of the underlying simulator or hardware. Experimental results show that qfusion-opt consistently improves performance across different simulation environments on quantum benchmark circuits, achieving up to 9.6x speedup on Qiskit-Aer and 3.0x speedup on Queen compared to the original circuits without gate fusion. Overall, qfusion-opt provides a practical step toward high performance quantum circuit simulation across different platforms.