Active Objects for Coordinating BSP Computations (Short Paper)

Abstract

Among the programming models for parallel and distributed computing, one can identify two important families. The programming models adapted to data-parallelism, where a set of coordinated processes perform a computation by splitting the input data; and coordination languages able to express complex coordination patterns and rich interactions between processing entities. This article takes two successful programming models belonging to the two categories and puts them together into an effective programming model. More precisely, we investigate the use of active objects to coordinate BSP processes. We choose two paradigms that both enforce the absence of data-races, one of the major sources of error in parallel programming. This article explains why we believe such a model is interesting and provides a formal semantics integrating the notions of the two programming paradigms in a coherent and effective manner.

A Survey on Parallelism and Determinism

“…Only concurrent writes on the same memory address, in the implementations of BSP that support it, can be a source of non-determinism. Recently, [HHLS18] integrated BSP and actors, leading to a model that features both task and data parallelism with limited nondeterminism. Additionally, it is possible to design a static analysis that checks deadlock-freedom for BSP programs [JDB + 17,Dab18].…”

Distributed futures for efficient data transfer between parallel processes

“…In particular, we rely on BSP Active Objects [9], a C++ library that allows the coordination of several data-parallel tasks implemented in BSP by using an actor-based task-parallel interaction with basic (non distributed) futures. It runs on top of MPI and uses the BSPonMPI v0.2 implementation of the BSPlib specification.…”

“…We will use this per BSP process thread to perform tasks dedicated to our distributed future management. The use of BSP active objects is illustrated in [9].…”