No description available.

This thesis describes an environment to evaluate and compare static schedulers for real pipelined streaming applications on massively parallel architectures, such as Intel Single chip Cloud Computer (SCC), Adapteva Epiphany, and Tilera TILE-Gx series. The framework allows performance comparison of schedulers in their execution time, or the energy usage of static schedules with energy models and measurements on real platform. This thesis focuses on the implementation of a framework evaluating the energy consumption of such streaming applications on the SCC. The framework can run streaming applications, built as task collections, with static schedules including dynamic frequency scaling. Streams are handled by the framework with FIFO buffers, connected between tasks. We evaluate the framework by considering a pipelined mergesort implementation with different static schedules. The runtime is compared with the runtime of a previously published task based optimized mergesort implementation. The results show how much overhead the framework adds on to the streaming application. As a demonstration of the energy measuring capabilities, we schedule and analyze a Fast Fourier Transform application, and discuss the results. Future work may include quantitative comparative studies of a range of different static schedulers. This has, to our knowledge, not been done previously.

No description available.

No description available.

Em certas aplicações, a redução no tempo de processamento de programas é fundamental. Em sistemas multiprocessados, busca-se minimizar esse tempo através de algoritmos que proveem escalamento estático das tarefas nos processadores. Esses algoritmos podem ser exatos, significando que o tempo de processamento resultante é o mínimo para uma dada arquitetura de hardware, ou aproximados, indicando que a solução obtida e sub-ótima. Este trabalho apresenta uma biblioteca de algoritmos heurísticos que fornecem escalamentos sub-ótimos e um modelo que simula a operação de redes de transputers. Cada escalamento gerado é testado na rede simulada e seu tempo de processamento é avaliado. Após ter verificado o desempenho de todos os algoritmos disponíveis, é apresentado como resposta o melhor escalamento testado e seu tempo simulado de processamento. O usuário pode então alocar as tarefas nos transputers, de acordo com a ordem e o regime de operação (seq ou par) fornecidos. O programa escalador foi testado com diversos arranjos de tarefas parcialmente ordenadas em diferentes arquiteturas de transputers, para verificar a conformidade do modelo da rede de processadores com a realidade. Os resultados dos testes atenderam plenamente aos requisitos de desempenho, apresentando desvios máximos em torno de 1%. / In certain scientific applications the reduction in the processing time is fundamental. In the processing area and, particularly, in multiprocessing, the minimization of this time through algorithms that provide static scheduling of tasks in processors is aimed. These algorithms can be exact, meaning that the resulting processing time is minimum for a certain hardware architecture or approximate, indicating that the solution obtained is sub-optimal. This work presents a library of heuristic algorithms that provide sub-optimal scheduling and a model that simulates the operation of any Transputers network. Together they constitute the computational tool here nominated Scheduling Program. Each schedule created is tested in the simulated network and its processing time is evaluated. After verifying the performance of every available heuristic algorithms, the Scheduling Program provides as response the best tested schedule and its simulated processing time. Then the user can allocate the tasks on the Transputers, following the order and the construction (SEQ or PAR) provided. The Scheduling Program was tested on several partially ordered tasks arrangements in different Transputers architectures to check the conformity of the processors networks model with reality. The results of the tests complied completely to the performance requirements, presenting a maximum deviation of about 1%.

Em certas aplicações, a redução no tempo de processamento de programas é fundamental. Em sistemas multiprocessados, busca-se minimizar esse tempo através de algoritmos que proveem escalamento estático das tarefas nos processadores. Esses algoritmos podem ser exatos, significando que o tempo de processamento resultante é o mínimo para uma dada arquitetura de hardware, ou aproximados, indicando que a solução obtida e sub-ótima. Este trabalho apresenta uma biblioteca de algoritmos heurísticos que fornecem escalamentos sub-ótimos e um modelo que simula a operação de redes de transputers. Cada escalamento gerado é testado na rede simulada e seu tempo de processamento é avaliado. Após ter verificado o desempenho de todos os algoritmos disponíveis, é apresentado como resposta o melhor escalamento testado e seu tempo simulado de processamento. O usuário pode então alocar as tarefas nos transputers, de acordo com a ordem e o regime de operação (seq ou par) fornecidos. O programa escalador foi testado com diversos arranjos de tarefas parcialmente ordenadas em diferentes arquiteturas de transputers, para verificar a conformidade do modelo da rede de processadores com a realidade. Os resultados dos testes atenderam plenamente aos requisitos de desempenho, apresentando desvios máximos em torno de 1%. / In certain scientific applications the reduction in the processing time is fundamental. In the processing area and, particularly, in multiprocessing, the minimization of this time through algorithms that provide static scheduling of tasks in processors is aimed. These algorithms can be exact, meaning that the resulting processing time is minimum for a certain hardware architecture or approximate, indicating that the solution obtained is sub-optimal. This work presents a library of heuristic algorithms that provide sub-optimal scheduling and a model that simulates the operation of any Transputers network. Together they constitute the computational tool here nominated Scheduling Program. Each schedule created is tested in the simulated network and its processing time is evaluated. After verifying the performance of every available heuristic algorithms, the Scheduling Program provides as response the best tested schedule and its simulated processing time. Then the user can allocate the tasks on the Transputers, following the order and the construction (SEQ or PAR) provided. The Scheduling Program was tested on several partially ordered tasks arrangements in different Transputers architectures to check the conformity of the processors networks model with reality. The results of the tests complied completely to the performance requirements, presenting a maximum deviation of about 1%.