Feedback from the scientific supervisor on the graduation thesis of Porsev Egor Valerevich from the Faculty of Mathematics and Mechanics:

"Development of a library for multithreaded and parallel computations based on the model of coroutines and channels"

Simplifying the use of parallel and concurrent programming is one of the main directions of modern programming languages' development. Concurrent and parallel computations are essential for reducing computation time and minimizing delays, but parallel programming is considered challenging for developers.

Porsev Egor Valerevich's work is devoted to developing a modified approach to multithreaded and parallel programming based on coroutines and channels in modern programming languages. This approach protects user-selected program segments from non-deterministic errors, such as races, event loss, and hang-ups, in cases where overhead costs to ensure determinism are insignificant for performance. It does not limit developers in other cases, allowing them to easily change the set of constraints during development, thereby providing more flexibility to find a compromise between performance and error protection, thus significantly increasing program reliability.

During the work, a review of current approaches to writing multithreaded and parallel programs used in popular modern languages was conducted. The proposed solution extends the widely used coroutine and channel approach by adding additional error protection. The developed approach was implemented as a library for the Java language. The implementation was fully covered by unit tests. Based on an analysis of typical examples of parallel and multithreaded programs in popular programming languages, a set of usage examples was developed, also serving as functional tests. An experimental performance evaluation was conducted vs low-level tools provided by the Java language, which did not reveal any critical losses. The source code of the library is uploaded to an open repository.

The following comments are made regarding the work:

1. No strict mathematical justifications are given for the guarantees provided by the proposed approach. Instead, analogies with existing models with proven properties are used for justification.
2. The integration of the proposed approach into the programming language at the syntax level is not considered. The integration at the library level is quite wordy.

Despite the above comments, I believe that Porsev Egor Valerevich's work is a completed study, achieves all set goals, and deserves an "excellent" A rating.

PhD in Mathematics and Mechanics, Senior Lecturer of the Department of Informatics, Faculty of Mathematics and Mechanics, St. Petersburg State University
June 4, 2023
_________________Salischev S.I.