Improving the Dynamic Creation of Processes in MPI-2

Márcia, C.; Pezzi, Guilherme Peretti; Mathias, Elton; Maillard, Nicolas; Navaux, Philippe O. A.

doi:10.1007/11846802_37

Cited by 15 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Por outro lado, a especificac ¸ão omite detalhes de implementac ¸ão sobre o gerenciamento e criac ¸ão das tarefas. Trabalhos recentes abordam o escalonamento [4] e roubo de tarefas hierárquico [17] em aplicac ¸ões dinâmicas com o propósito de distribuir a carga de trabalho entre os recursos disponíveis.…”

Section: Dinamismo E Aplicac ¸õEs Paralelas Irregularesunclassified

Controle de Granularidade com threads em Programas MPI Dinâmicos

Lima

Maillard

2008

Anais Do IX Simpósio Em Sistemas Computacionais De Alto Desempenho (SSCAD 2008)

Self Cite

View full text Add to dashboard Cite

O controle de granularidade é um fator importante no desempenho de programas paralelos. Problemas estáticos adaptam sua granularidade pela decomposição e atribuição de dados a cada tarefa, mas em irregulares não é possível prever a carga de trabalho antes da execução. Os irregulares que utilizam decomposição recursiva, como ordenação, necessitam de dinamismo com suporte a criação de tarefas sob demanda. Alguns ambientes de programação, como Cilk e KAAPI, oferecem dinamismo e trabalham com granularidade através do conceito abstrato de tarefa porém possuem limitações que dificultam seu uso em PAD. O MPI, padrão de fato em PAD, oferece dinamismo de processos e uso de threads mas atribui à implementação especificar o comportamento na criação de uma tarefa. Este trabalho propõe investigar as vantagens no controle de granularidade com threads em programas MPI dinâmicos, através da substituição da criação de processos por tarefas onde um mecanismo (libSpawn) decide entre lançar processo(s) ou thread(s). Os resultados obtidos com o programa de ordenação Cilksort, que segue o modelo Divisão-e-Conquista, demonstram ganhos de até 85% em criação de tarefas e comunicação.

show abstract

Section: Dinamismo E Aplicac ¸õEs Paralelas Irregularesunclassified

Controle de Granularidade com threads em Programas MPI Dinâmicos

Lima

Maillard

2008

Anais Do IX Simpósio Em Sistemas Computacionais De Alto Desempenho (SSCAD 2008)

Self Cite

View full text Add to dashboard Cite

show abstract

“…with standard Send/Receive), with their parent and their children processes. A previous work [3] has shown how to schedule, at run-time, such spawned processes, with a simple, centralized daemon. This section explains how a D&C program implemented with MPI-2 can be scheduled on-line by Work Stealing.…”

Section: Scheduling Dandc Mpi-2 Programs With Hierarchical Work Stealingmentioning

confidence: 99%

On-line Scheduling of MPI-2 Programs with Hierarchical Work Stealing

Pezzi¹,

Cera²,

Mathias³

et al. 2007

19th International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD'07)

Self Cite

View full text Add to dashboard Cite

MPI (MessagePassing Interface) is the de facto standard in High Performance Computing. By using some MPI-2 new features, such as the dynamic creation of processes, it is possible to implement highly efficient parallel programs that can run on dynamic and/or heterogeneous resources, provided a good schedule of the processes can be computed at run-time. A classical solution to schedule parallel programs on-line is Work Stealing. However, its use with MPI-2 is complicated by a restricted communication scheme between the processes: namely, spawned processes in MPI-2 can only communicate with their direct parents. This work presents an on-line scheduling algorithm, called Hierarchical Work Stealing, to obtain good load-balancing of MPI-2 programs that follow a Divide & Conquer strategy. Experimental results are provided, based on a synthetic application, the N-Queens computation. The results show that the Hierarchical Work Stealing algorithm enables the use of MPI with high efficiency, even in parallel dynamic HPC platforms that are not as homogeneous as clusters. 19th International Symposium on Computer Architecture and High Performance Computing 1550-6533/07 $25.00

show abstract

“…In order to maximize the parallel computing of a multicore system, load balance becomes an issue that needs to be considered when scheduling. Round robin is one of the simple methods to dispatch tasks in a multicore system [ 15 ], where tasks are dispatched to processor cores in a rotated order. Shortest queue first [ 16 ] is another method that is often used.…”

Section: Introductionmentioning

confidence: 99%

A High Performance Load Balance Strategy for Real-Time Multicore Systems

Cho

Tsai

Chiu

et al. 2014

The Scientific World Journal

View full text Add to dashboard Cite

Finding ways to distribute workloads to each processor core and efficiently reduce power consumption is of vital importance, especially for real-time systems. In this paper, a novel scheduling algorithm is proposed for real-time multicore systems to balance the computation loads and save power. The developed algorithm simultaneously considers multiple criteria, a novel factor, and task deadline, and is called power and deadline-aware multicore scheduling (PDAMS). Experiment results show that the proposed algorithm can greatly reduce energy consumption by up to 54.2% and the deadline times missed, as compared to the other scheduling algorithms outlined in this paper.

show abstract

Improving the Dynamic Creation of Processes in MPI-2

Cited by 15 publications

References 7 publications

Controle de Granularidade com threads em Programas MPI Dinâmicos

Controle de Granularidade com threads em Programas MPI Dinâmicos

On-line Scheduling of MPI-2 Programs with Hierarchical Work Stealing

A High Performance Load Balance Strategy for Real-Time Multicore Systems

Contact Info

Product

Resources

About