Research of the application of GPGPU and TPU technologies for ensuring comment quality in version control systems

• Veröffentlicht in: PROBLEMS IN PROGRAMMING
• Datum: 2025
• Heft: 1
• Сторінки: 24-37
• ISSN: 1727-4907
• Автори та афіліації: B.O. Semonov — Kiev Taras Shevchenko National University — ORCID: 0009-0001-3692-9415; S.D. Pogorilyy — Kiev Taras Shevchenko National University — ORCID: 0000-0002-6497-5056
• Ключові слова: tpu, система контролю версій, опис внесених змін, rnn, gpgpu, репозиторій, rdp, gpu, rdpсервер, trpv1
• Hauptverfasser: Semonov, B.O., Pogorilyy, S.D.
• Format: Artikel
• Sprache: Ukrainisch
• Veröffentlicht: PROBLEMS IN PROGRAMMING 2025
• Schlagworte: commit message; GPGPU; repository; RNN; TPU; version control system; UDC 004.91
• Online Zugang: https://pp.isofts.kiev.ua/index.php/ojs1/article/view/762
• Назва журналу: Problems in programming

The study substantiates the relevance of solving the issue of ensuring the quality of descriptions for changes made in source code files within version control systems. Machine learning methods, particularly neural networks of various architectures, are employed for comment filtering. Neural networks are deemed appropriate due to the necessity of identifying descriptions that accurately reflect the purpose of the changes made. Recurrent neural networks were developed and trained on a dataset of change descriptions obtained through the GitHub REST API. To enhance training performance, various hardware and software platforms such as CPU, TPU, and GPGPU were utilized. The accuracy of the models was analyzed using metrics like Accuracy and the harmonic mean (F1-score).Prombles in programming 2025; 1: 24-37