Publicação

Design and implementation of a library for parallel independent computations in Octave

Ver documento

Detalhes bibliográficos
Resumo:This essay aims at developing, implementing and assessing a framework allowing the writing of parallel applications in Octave. The framework will be targeted at the simul- taneous evaluation of multiple instances of the same function with different parameters. The functionality of the frameworks is inspired in MATLAB’s Parallel Computing Tool- box, namely the parfeval function that uses an asynchronous model of computation. The system is developed using Oct-Files that allow the extension of the functionality of Octave; these extensions allow the communication between the Octave interpreter and an infrastructure built in Python that supports parallel executions. The parallel in- frastructure allows the exploitation of both shared-memory multiprocessors and clusters. The system has been evaluated using traditional embarrassing-parallel applications and its performance is good, if we’re able to ensure that the computations last for a time significantly greater than the overhead startup.
Autores principais:Candeias, Miguel Balão
Assunto:Octave parallel applications in Octave asynchronous computation Oct-Files Python
Ano:2022
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
Descrição
Resumo:This essay aims at developing, implementing and assessing a framework allowing the writing of parallel applications in Octave. The framework will be targeted at the simul- taneous evaluation of multiple instances of the same function with different parameters. The functionality of the frameworks is inspired in MATLAB’s Parallel Computing Tool- box, namely the parfeval function that uses an asynchronous model of computation. The system is developed using Oct-Files that allow the extension of the functionality of Octave; these extensions allow the communication between the Octave interpreter and an infrastructure built in Python that supports parallel executions. The parallel in- frastructure allows the exploitation of both shared-memory multiprocessors and clusters. The system has been evaluated using traditional embarrassing-parallel applications and its performance is good, if we’re able to ensure that the computations last for a time significantly greater than the overhead startup.