Application Of Genetic Algorithm Based On Charm++ Parallel Computing Architecture In Quantum Superposition State

In the era of big data,high performance computing has gradually become a key tool for in-depth research in computer science and other engineering fields,and it can provide relatively reliable results for a variety of complex problems.With the further improvement of the computing power of high performance computer systems,as well as the activeness of quantum chemistry research,quantum chemists begin to develop new quantum chemical computing algorithms and models,while making effort to improve the parallel scale and computing efficiency of the existing work.For quantum chemistry programs,the adoption of suitable APIs can accelerate code developments as well as maximize computing resources.The current deterministic parallel solutions are all based on MPI,OpenMP,CUDA and OpenCL.In recent years,the density matrix renormalization group(DMRG)is the most popular method in quantum chemistry.With the continuous expansion of the active space,the number of configuration arrangements in the electronic configuration wave function(the superposition of quantum states)will increase exponentially in a given molecular orbital space,which leads to the great limitations of the current deterministic parallel method.When the parallel computing load increases with the scale of computing,portability,communication delay,and load balancing are the challenges that an efficient parallel program must face.Charm++ can encapsulate high performance computing related technologies such as data communication,fault tolerance,and load balancing at runtime,it can also shield the complexity of supercomputer architecture and parallel technologies for application domain experts.And it can solve the above challenges easily.Charm++can encapsulate high-performance computing technologies such as communication,fault tolerance,and load balancing at runtime,it also can shield supercomputer architecture and the complexity of parallel technology.The paper focuses on the construction of matrix product state(MPS)wave function to configuration interaction(CI)wave function in the Hilbert space,and studies with genetic algorithm.The main work is as follows:First,it introduces the current research status of high performance computing and quantum chemistry at home and abroad,explains the research significance of this paper,and analyzes the main problems of efficient parallel programs.Then,aiming at these problems,the new parallel programming model-oriented Charm++ is proposed to implement efficient parallel programs.Finally,some basic knowledge of Charm++ used in the research and the knowledge in quantum chemistry are introduced.Second,according to the characteristics of the genetic algorithm and the development concept of the Charm++,we restructured the effective process of MPS-to-CI,introduced the entire process implemented using Charm++,the genetic algorithm EDGA combined with quantum information technology and Monte Carlo algorithms.Finally,based on the exponential expansion of the Hibbert space,we evolved the sampling program to a population expansion version of PE-EDGA.Finally,we used typical chemiluminescent molecules to perform parallel efficiency tests on MPS-to-CI,namely 1,2-dioxetanone and firefly dioxetanone anion(FDO-).It is proved that the method can be flexibly applied to various multi-core architectures.After increasing the proportion of asynchronous execution,we find that the parallel efficiency can be continuously improved.The PE-EDGA can be constructed from the Hilbert space in a few hours.A CAS-type CI wave function can save a lot of time cost in the experiment.