GoDEL: A multidirectional dataflow execution model for large-scale computing

Hits: 4058
Type of Publication:
  • Kulkarni, Abhishek
  • Lang, Michael
  • Lumsdaine, Andrew
First Workshop on Data-Flow Execution Models for Extreme Scale Computing, 2011-10-10 (Galveston Island, Texas, United States)
As the emerging trends in hardware architecture guided by performance, power eciency and complexity drive us towards massive processor parallelism, there has been a renewed interest in dataflow models for large-scale computing. Dataflow execution models, being declarative in nature, lead to improved programmability at scale by implicitly managing the computation and communication for the application. In this paper, we present a multidirectional dataflow execution model called GoDEL based on propagation networks. Propagator networks allow concurrent, general-purpose computation on partial data. Implemented with eciency and programmer productivity as its goals, we describe the implementation of the GoDEL language and its runtime. We further discuss a representative example that benefits from the flexibility in the execution model.

© 2018 New Mexico Consortium