Dataflow Execution of Hierarchically Tiled Arrays

TítuloDataflow Execution of Hierarchically Tiled Arrays
AutoresChih-Chieh Yang, Juan C. Pichel, and David A. Padua
TipoComunicación para congreso
Fonte European Conference on Parallel and Distributed Computing, Göttingen (Alemania), 2019.
RankRanked A in CORE
AbstractAs the parallelism in high-performance supercomputers con- tinues to grow, new programming models become necessary to maintain programmer productivity at today’s levels. Dataflow is a promising exe- cution model because it can represent parallelism at different granularity levels and to dynamically adapt for efficient execution. The downside is the low-level programming interface inherent to dataflow. We present a strategy to translate programs written in Hierarchically Titled Arrays (HTA) to the dataflow API of Open Community Runtime (OCR) sys- tem. The goal is to enable program development in a convenient notation and at the same time take advantage of the benefits of a dataflow run- time system. Using HTA produces more comprehensive codes than those written using the dataflow runtime programming interface. Moreover, the experiments show that, for applications with high asynchrony and sparse data dependences, our implementation delivers superior perfor- mance than OpenMP using parallel for loops.
Palabras chaveParallel programming, Dataflow, High-level programming abstraction, Parallel algorithm

Programas científicos