Português English
Contato

Proposta de Tese de Julio Toss


Detalhes do Evento


Aluno: Julio Toss
Orientador: Prof. Dr. João Luiz Dihl Comba

Título: Parallel Algorithms and Data Structures for Interactive Data Problems

Linha de Pesquisa: Computação Gráfica, Processamento de Imagens e Interação

Data: 30/05/2017
Horário: 10h
Local: Prédio 43412 – Sala 215 (sala de videoconferência), Instituto de Informática

Banca Examinadora:
Prof. Dr. Luis Gustavo Nonato (ICMC-USP – por videoconferência)
Profª. Drª. Carla Maria Dal Sasso Freitas (UFRGS)
Prof. Dr. Lucas Mello Schnorr (UFRGS – por videoconferência)

Presidente da Banca: Prof. Dr. João Luiz Dihl Comba

Abstract: The quest for performance has been a constant through the history of computing systems. It has been more that a decade now since the sequential processing model had shown its first signs of exhaustion to keep performance improvements. Walls to the sequential computation pushed a paradigm shit and now the parallel processing has been established as the standard in most modern computing systems. With the widespread adoption of parallel computers, many applications have been ported to fit those new architectures. However several challenges still remain to make efficient parallelizations of unconventional applications such as in interactive and real-time problems. These problems are characterized by the fact that the computation is heavily dependent on dynamic data interactions. This raises the challenge of making a suitable problem decomposition while improving memory access and data locality. In this thesis we address data dependent problems on different application areas such as in physics based simulation and on streaming data analysis. We propose algorithms and data structures targeting the performance constraints in current and future scenarios of high-performance computing systems. We present a GPU parallelization of a data dependent algorithm used in the real-time  physics-based simulation community. We also propose a parallelizable data structure for data locality maintenance on dynamic data with application to in-memory streaming data exploration.

Keywords: parallel processing, data locality, stream processing, real-time processing, physics-based simulation