【作者】 吴华；

【导师】 邓玉欣；

【作者基本信息】 华东师范大学 ， 软件工程（专业学位）， 2021， 硕士

【摘要】 由于量子效应,经典计算机元器件的发展已达到现代物理学的极限。得益于量子叠加态,量子计算机与经典计算机相比能显著加快计算速度。因此,世界各国掀起了研发解决特定问题的量子计算机的浪潮。在过去的几十年间,虽然量子计算随着一系列量子算法的提出不断进步,但距离研制出可扩展的通用量子计算机仍有很长的路要走。因此,在经典计算机上模拟量子计算也变得异常重要,就像VLSI设计中的电路模拟一样。Qsimulation作为一款量子模拟器软件,可让使用者编写量子程序、画量子线路图、运行程序并查看结果。和很多其它的量子模拟器一样,Qsimulation模拟量子演化的效率很大程度上取决于它处理矩阵运算的能力。在本文中我们将提出一种能加速矩阵向量乘计算的新算法,优化Qsimulation得到更优的模拟器软件版本Qsimulation V2.0.s。该算法基于张量积和矩阵分解,很适合模拟量子线路的执行。实验结果表明Qsimulation V2.0.s优于开源软件框架Qiskit和Project Q。根据矩阵乘法属于计算密集型任务的特点,本文进一步提出多线程并行计算的优化方法,得到软件版本Qsimulation V2.0。实验结果表明Qsimulation运行效率进一步提高。此外,根据一种更直观的想法,本文又用一种启发式算法实现了Qsimulation。该算法将量子叠加态以键值对的形式存在表中,状态演化时直接对表中状态进行更改。实验结果表明该算法能模拟更多的量子比特,且当量子态的叠加态数目较少时运行效率更高。更多还原

【Abstract】 Due to quantum effect,the development of classical computer components has reached the limit of modern physics.Thanks to quantum superposition states,quantum computers can accelerate computation remarkably compared with classical computers.Therefore,research and development of quantum computers to solve specific problems has started a worldwide craze among countries.In the past few decades,quantum computation has made continuous progress with a series of quantum algorithms proposed.However,the research of quantum computers is still in their infancy,and there is still a long way to go to develop a scalable quantum computer.Therefore,it is of great importance to simulate quantum computation on classical computers,just like circuit simulation in VLSI design.As a quantum simulator software,Qsimulation allows a user to write quantum programs,draw quantum circuts,execute the programs and view the results.Similar to many other quantum simulation tools,the performance of Qsimulation largely depends on its capacity of dealing with matrix operations.In this thesis we present Qsimulation V2.0.s,an optimized quantum simulator that implements a new algorithm for accelerating matrixvector multiplications.The algorithm is based on matrix decomposition using tensor products and suitable for simulating the execution of quantum circuits.Experimental results show that Qsimulation V2.0.s outperforms the open source frameworks Qiskit and Project Q.Since matrix multiplication is a computationally intensive task,we then propose an optimization method of multithreading parallel computing to get Qsimulation V2.0.Experimental results show that the efficiency of Qsimulation is further improved.By using a more intuitive idea,we design a heuristic algorithm to perform quantum simulation.The algorithm stores the quantum superposition states in key-value pairs,and changes the states directly when they evolve.Experimental results show that the algorithm can simulate more quantum bits,and the execution efficiency is higher when the number of superposition states is small.更多还原