quEStab: Towards Scalable Quantum Circuit Simulation on Multi-GPU using an Extended Stabilizer Formalism

Abstract

Efficient simulation of large-scale quantum circuits remains a critical bottleneck in quantum computing research, as the exponential memory and runtime growth of state vector-based approaches limits scalability beyond tens of qubits. This paper presents a scalable quantum circuit simulation that extends the conventional stabilizer tableau formalism to support mixed Clifford and non-Clifford operations on multi-GPU platforms. The proposed quantum simulation introduces a dynamically extensible tableau representation that eliminates destabilizer redundancy and distributes independent Pauli groups across GPUs for fine-grained parallelism. A two-stage CUDA kernel pipeline efficiently manages the branching and merging of tableau rows during non-Clifford evolution by separating counting required rows from row updates, ensuring deterministic updates and compact memory usage. Comprehensive evaluations on 128 circuits from QASMBench demonstrate that the proposed simulation achieves 78.1% overall coverage, executing many non-Clifford circuits that other simulators failed to complete. Compared with state vector-based frameworks, our simulation reduces peak memory consumption by up to 10126 × and maintains stable multi-GPU scalability, successfully executing circuits up to 30000 qubits.