The use of domain-specific languages and actor model for the automation of high performance computing

The article describes the use of actor-oriented domain-specific languages for the automation of high-performance scientific computing in the Templet Web cloud service. The automatic programming in Templet Web is based on the concepts of skeleton programming, domain-specific languages, and supporting visual editors. The sample program for checking the basic trigonometric identity was developed in the actor-based domain-specific language. While the program is simple enough, it implements all features of the domain-specific language. High-level task description is transformed into the code that uses actor-based run-time system. The article describes a measurement of the level of automation for the domain-specific language and actor programming method. The measurement is based on evaluating proportions of library code, generated code, and manually written code in the given sample. The criterion of complexity and degree of automation is based on a comparative analysis of code sizes from the different parts of the program. The sample is located at the GitHub service: https://github.com/templet-language/newtemplet. A high degree of automation is reached by simplicity of the actor model and by automatically constructed (from the domain-specific description) interface code. The current version of Templet Web cloud service is available at the following web address: http://templet.ssau.ru/app. It is deployed on supercomputer “Sergey Korolev” of Samara University. The service supports actor-based domain-specific language named Templet, and includes several ready-to-use skeleton programs: “TaskBag”, “Pipeline”, and others. The technology shows its effectiveness in the field of multivariate dynamic systems and processes, in researching gyroscope attitude control systems of spacecraft, and for teaching students the principles of work on a supercomputer.