Saidov Behruz Badridinovich, Saidov Naimjon Bozorovich 

Annotation:

This study investigates the problem of rapidly increasing memory requirements in classical simulation of quantum computations. As the dimensionality of a quantum system grows, the volume of data increases exponentially, which limits the applicability of traditional state-vector simulators to the available amount of RAM. As a result, simulation of systems exceeding 28–30 qubits becomes difficult without the use of specialized computing clusters. A method for optimizing the simulator architecture is proposed, based on the use of memory-mapped files to map data into the process address space and the utilization of high-speed NVMe storage devices as an extended level of the memory hierarchy. The software implementation is developed in the Rust programming language using safe memory management mechanisms and multithreaded data processing. Additionally, a bit-level amplitude indexing scheme has been designed to ensure sequential access to data arrays and reduce the rate of CPU cache misses. Experimental results demonstrate performance reaching up to 362 million operations per second and significant acceleration compared with interpreted implementations. At the same time, RAM consumption remains below 150 MB even when processing data arrays exceeding 16 GB. The obtained results confirm the feasibility of simulating quantum systems up to 40 qubits on standard workstations without requiring expensive HPC resources.

Keywords: quantum computing, state-vector simulation, memory-mapped files, parallel computing, Rust, NVMe, memory optimization.