Simulação de uma arquitetura de computação própria ao paradigma orientado a notificações
Traditional programming paradigms lack techniques for the productive and quality software’s design that makes efficient use of the parallel execution capabilities provided by the modern computing systems hardware. The Notification Oriented Paradigm (NOP) presents a new approach to develop more effic...
| Autor principal: | Pordeus, Leonardo Faix |
|---|---|
| Formato: | Dissertação |
| Idioma: | Português |
| Publicado em: |
Universidade Tecnológica Federal do Paraná
2017
|
| Assuntos: | |
| Acesso em linha: |
http://repositorio.utfpr.edu.br/jspui/handle/1/2738 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: |
Traditional programming paradigms lack techniques for the productive and quality software’s design that makes efficient use of the parallel execution capabilities provided by the modern computing systems hardware. The Notification Oriented Paradigm (NOP) presents a new approach to develop more efficient and decoupled software when compared to the traditional programming paradigms, such as the Imperative Paradigm and the Declarative Paradigm. The NOP allows the exploitation of parallelization and/or distribution in a simpler and more efficient way than more commonly used programming paradigms. Initially, the NOP was conceived as an alternative software paradigm, highlighting the recent NOP language and compiler. However, other research was focused on hardware development using reconfigurable logic, which enables an execution model closer to the NOP model. Indeed, in parallel processing, the NOP applications cannot always be executed efficiently by the traditional computer architectures hardware based on set of cores with sequential execution model. Aiming at more efficient execution of NOP applications, a new Notification Oriented Computer Architecture (NOCA) has been developed. NOCA is based on multiple processing units, where each unit corresponds to specific elements of the NOP model. This way, parallelism is adequately exploited. Experiments were performed on a prototype of this architecture. When compared with Imperative Paradigm, execution time was improved under certain conditions, in particular, the low scale of parallelization obtained as a function of stage limitation of experimentation. In this sense, this Master of Science dissertation presents the development of a simulator for the NOCA architecture, which is called NOCASim. The NOCA implementation over software approach aims to simulate NOCA behavior with amount of simulated processors in the same proportions as the amount of NOP elements of an application. This allows obtain data for validation of NOCA functional model. As a development strategy, firstly this work adapted a NOP language to generate NOCA assembly facilitating the development of applications for this architecture. In the sequence, the own NOCA's design artifacts were used for the NOCASim development. The NOCASim operation consists of simulating in software to update the NOCA internal states in a similar way to each change in the clock cycle. The experiments results using NOCASim with different configurations have demonstrated that the NOCA performance improve when the number of processing units increases. In addition, were observed new possibilities for conceptual improvements to NOCA, which tend to strengthen it as a parallel execution alternative. |
|---|