Evaluación del rendimiento de FPGAs para aceleración de kernels de cómputo: Un estudio comparativo
Loading...
Download
Official URL
Full text at PDC
Publication date
2024
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
El proyecto que aquí se presenta se centra en el uso de FPGAs como aceleradores de cómputo, explorando su potencial y eficiencia energética al ejecutar diversos kernels. Un kernel, en el contexto computacional, se refiere a la parte del código que se ejecuta repetidamente y consume la mayor cantidad de tiempo de procesamiento. La selección de kernels específicos para este estudio se basa tanto en su relevancia y aplicabilidad en diferentes campos, como en la inclusión de ciertas características cuyo estudio es relevante en el área de la optimización. Este trabajo no solo pretende medir el rendimiento y eficiencia energética de las FPGAs, sino también compararlos con otras tecnologías de procesamiento actuales. La metodología adoptada en este proyecto implica la implementación de cada kernel en una FPGA, seguida de una serie de pruebas y mediciones. Los resultados obtenidos se comparan con implementaciones equivalentes en procesadores de uso generalista, con el objetivo de determinar las ventajas y limitaciones de las FPGAs. A través de este estudio se espera contribuir al conocimiento existente sobre el uso de FPGAs en el ámbito de la aceleración y la eficiencia de cómputo, proporcionando datos empíricos y análisis detallados que puedan guiar futuras investigaciones y aplicaciones prácticas.
The project presented here focuses on the use of FPGAs as computational accelerators, exploring their potential and energy efficiency when running various kernels. A kernel, in the computational context, refers to the part of the code that is executed repeatedly and consumes the most processing time. The selection of specific kernels for this study is based both on their relevance and applicability in different fields, as well as the inclusion of certain characteristics whose study is relevant in the area of optimization. This work not only aims to measure the performance and energy efficiency of FPGAs, but also to compare them with other current processing technologies. The methodology adopted in this project involves the implementation of each kernel on an FPGA, followed by a series of tests and measurements. The results obtainedare compared with equivalent implementations in general-purpose processors, with the aim of determining the advantages and limitations of FPGAs. Through this study we hope to contribute to the existing knowledge on the use of FPGAs in the field of acceleration and computational efficiency, providing empirical data and detailed analysis that can guide future research and practical applications
The project presented here focuses on the use of FPGAs as computational accelerators, exploring their potential and energy efficiency when running various kernels. A kernel, in the computational context, refers to the part of the code that is executed repeatedly and consumes the most processing time. The selection of specific kernels for this study is based both on their relevance and applicability in different fields, as well as the inclusion of certain characteristics whose study is relevant in the area of optimization. This work not only aims to measure the performance and energy efficiency of FPGAs, but also to compare them with other current processing technologies. The methodology adopted in this project involves the implementation of each kernel on an FPGA, followed by a series of tests and measurements. The results obtainedare compared with equivalent implementations in general-purpose processors, with the aim of determining the advantages and limitations of FPGAs. Through this study we hope to contribute to the existing knowledge on the use of FPGAs in the field of acceleration and computational efficiency, providing empirical data and detailed analysis that can guide future research and practical applications
Description
Trabajo de Fin de Máster en Ingeniería Informática, Facultad de Informática UCM, Departamento de Arquitectura de Computadores y Automática, Curso 2024/2025.