MONICA PEREIRA

Followers

Following

Public Views

Matthew J. Brown

Southern Illinois University

Muqtedar Khan

University of Delaware

Esther MacCallum-Stewart

Staffordshire University

Andreas Umland

National University of "Kyiv-Mohyla Academy"

Brian Burton

Abilene Christian University

Heather Douglas

Michigan State University

Armando Marques-Guedes

UNL - New University of Lisbon

Martin O'Neill

University of York

Andrew N Liaropoulos

University of Piraeus

rudolph bauer

Washington Center for Consciousness Studies

Interests

Uploads

Papers by MONICA PEREIRA

A Routing based Genetic Algorithm for Task Mapping on MPSoC

2020 X Brazilian Symposium on Computing Systems Engineering (SBESC)

This work proposes an optimized task mapping solution called Routing Model-based Genetic Algorith... more This work proposes an optimized task mapping solution called Routing Model-based Genetic Algorithm (RMGA) that combines task mapping and routing problems using an Integer Linear Programming (ILP) model as a fitness function. We compared our proposed RMGA with other Genetic Algorithms (GA) that address the mapping problem using a classical flow x distance fitness function evaluation. Experimental results evaluating communication latency demonstrate that the proposed algorithm outperforms two GAs from literature. It presents up to 30% lower delay when simulating tasks communication pattern of six different applications and varying packet injection rate and NoC dimension. Additionally, RMGA presents the lowest delay of more than 70 % in all simulated scenarios compared to the other GAs.

Mapping Wired Links in a Hybrid Wired-Wireless Network-on-Chip

2020 X Brazilian Symposium on Computing Systems Engineering (SBESC), 2020

The use of hybrid Networks-on-Chip has increased in the scenario of communication in MP-SoCs, com... more The use of hybrid Networks-on-Chip has increased in the scenario of communication in MP-SoCs, combining the advantages of different approaches. While a wired-wireless network can bring improvements in latency and power consumption, in some scenarios it presents a greater packet loss rate than in wired NoCs during communication. The purpose of this paper is to map the links that interconnect the wired and wireless routers so that we can optimize the network and achieve a trade-off between the average latency of the network and the number of delivered packets. To assess the mapping, three different topologies are tested, along with three different traffic patterns and three different injection rates. Experimental results show that for most cases it was possible to optimize the hybrid networks, achieving up to 80% of delivered packets, while reducing latencyby rates up to 5 times.

AtTune: A Heuristic based Framework for Parallel Applications Autotuning

Several aspects limit the scalability of parallel applications, e.g., off-chip bus saturation and... more Several aspects limit the scalability of parallel applications, e.g., off-chip bus saturation and data synchronization. Moreover, the high cost of cooling HPC systems, which can outweigh the cost of developing the system itself, has pushed the parallel application’s execution to another level of requirements, in terms of performance and energy. In this work, we propose AtTune: a heuristic-based framework for tuning the number of processes/threads and CPU frequency to optimize the parallel applications’ execution. AtTune is transparent for the user, independent of the input size, and it optimizes for different parallel programming models. We evaluated our proposed solution considering five well-known kernels implemented in MPI and OpenMP. Experimental results on two real multi-core systems showed that AtTune improves up to 36%, 11%, and 32% the energy efficiency, performance, and Energy-Delay Product, respectively.

MONICA PEREIRA

Related Authors

Uploads

Papers by MONICA PEREIRA

Log In