A Generalized Streaming Model for Concurrent Computing

TL;DR

This paper introduces a generalized stream computing model to unify parallelization strategies, aiming to simplify multicore programming, improve language design, and enhance debugging tools for concurrent computing.

Contribution

It proposes a high-level, unified model inspired by stream computing that supports language, compiler, and OS design for concurrent programming.

Findings

Provides a foundational algebraic analysis of the model

Demonstrates a programming framework enabled by the model

Shows potential for powerful language, compiler, and OS design options

Abstract

Multicore parallel programming has some very difficult problems such as deadlocks during synchronizations and race conditions brought by concurrency. Added to the difficulty is the lack of a simple, well-accepted computing model for multicore architectures--because of that it is hard to develop powerful programming environments and debugging tools. To tackle the challenges, we promote a generalized stream computing model, inspired by previous researches on stream computing, that unifies parallelization strategies for programming language design, compiler design and operating system design. Our model provides a high-level abstraction in designing language constructs to convey concepts of concurrent operations, in organizing a program's runtime layout for parallel execution, and in scheduling concurrent instruction blocks through runtime and/or operating systems. In this paper, we give a high-level description of the proposed model: we define the foundation of the model, show its simplicity through algebraic/computational operation analysis, illustrate a programming framework enabled by the model, and demonstrate its potential through powerful design options for programming languages, compilers and operating systems.