GoDEL: A Multidirectional Dataflow Execution Model for Large-Scale Computing

As the emerging trends in hardware architectureguided by performance, power efficiency and complexity driveus towards massive processor parallelism, there has been arenewed interest in dataflow models for large-scale computing.Dataflow programming models, being declarative in nature,lead to improved programmability at scale by implicitly man-aging the computation and communication for the application.In this paper, we present GoDEL, a multidirectional dataflowexecution model based on propagation networks. Propaga-tor networks allow general-purpose parallel computation onpartial data. Implemented with efficiency and programmerproductivity as its goals, we describe the syntax and semantics of the GoDEL language and discuss its implementation and runtime. We further discuss representative examples from various programming paradigms that are encompassed by and benefit from the flexibility in the multidirectional execution model.