Performance Meets Programmabilty: Enabling Native Python MPI Tasks In PyCOMPSs

The increasing complexity of modern and future computing systems makes it challenging to develop applications that aim for maximum performance. Hybrid parallel programming models offer new ways to exploit the capabilities of the underlying infrastructure. However, the performance gain is sometimes a...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Proceedings - Euromicro Workshop on Parallel and Distributed Processing s. 63 - 66
Hlavní autoři: Elshazly, Hatem, Lordan, Francesc, Ejarque, Jorge, Badia, Rosa M.
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 01.03.2020
Témata:
Computational modeling
Distributed Computing
High Performance Computing
Hybrid Programming Models
MPI
Parallel processing
Parallel programming
Productivity
Python
Runtime
Task analysis
Task-based Parallel Programming Models
ISSN:2377-5750
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The increasing complexity of modern and future computing systems makes it challenging to develop applications that aim for maximum performance. Hybrid parallel programming models offer new ways to exploit the capabilities of the underlying infrastructure. However, the performance gain is sometimes accompanied by increased programming complexity. We introduce an extension to PyCOMPSs, a high-level task-based parallel programming model for Python applications, to support tasks that use MPI natively as part of the task model. Without compromising application's programmability, using Native MPI tasks in PyCOMPSs offers up to 3x improvement in total performance for compute intensive applications and up to 1.9x improvement in total performance for I/O intensive applications over sequential implementation of the tasks.
ISSN:2377-5750
DOI:10.1109/PDP50117.2020.00016