Programski jeziki za vzporedno programiranje

PINTAR, JANEZ (author), Slivnik, Boštjan (mentor)

Medtem ko so se večjedrni računalniki industrijsko že povsem uveljavili, se ni pomembno uveljavil še noben programski jezik, ki bi implicitno dovolil izkoriščanje vzporednega računanja. Trenutno je de facto standard za vzporedno programiranje programski jezik C++ z okoljem OpenMP za večnitno programiranje ter s knjižnico MPI za porazdeljeno programiranje. V delu smo spoznali nekatere razvijajoče se programske jezike, ki omogočajo implicitno vzporedno računanje in nato primerjali te med seboj in z že uveljavljenimi pristopi. Ugotovili smo, da hvala visokemu nivoju jezika je vzporedno programiranje v jeziku Chapel bolj produktivno kot z jezikom C++ in pri tem so hitrosti izvajanja primerljive. Žal je jezik Chapel zaradi majhnega ekosistema še vedno bol namenjen računalničarjem. Jezik Julia je posebno primeren za znanstvene programerje, ki nimajo nujno bogatega računalniškega znanja. Kljub interpretirani naravi je hitrost jezika presenetljiva, čeprav ne more tekmovati s hitrostjo prevedenih jezikov. Končno se jezik X10 ni posebno izkazal, ne v hitrosti izvajanja, ne v produktivnosti.

• Language: Slovenian
• Keywords: vzpredno programiranje, Chapel, Julia, X10
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2019
• PID: 20.500.12556/RUL-106378
• Publication date in RUL: 20.02.2019

Secondary language

• Language: English
• Title: Programming languages for parallel programming

Abstract

While multi-core computers have already become fully established, there is no such established programming language, that allows implicit exploitation of parallel computing. The current de facto standard for parallel programming is the C++ programming language, with the OpenMP framework for multi-threading programming and the MPI library for distributed programming. In this work we studied some evolving programming languages that allow implicit parallel programming. We compared them between each other and with the established methods. We found that thanks to the Chapel language's high-level, parallel programming in that is more productive than programming in the C++ language, even though the execution speeds are comparable. Unfortunately, because of the small ecosystem, Chapel language is still mostly meant for people with knowledge in computer science. Language Julia is especially suitable for scientific programmers who do not necessarily have computer science skills. Despite the interpreted nature, the speed of the language is surprising, although it can not compete with the speed of the compiled languages. Finally, the language X10 did not performed very well, not in the speed of execution, neither in productivity.

• Keywords: parallel programming, Chapel, Julia, X10