Streamware: programming general-purpose multicore processors using streams

  • Authors:

  • Jayanth Gummaraju;Joel Coburn;Yoshio Turner;Mendel Rosenblum

  • Affiliations:

  • Stanford University, Stanford, CA;Stanford University, Stanford, CA;Hewlett-Packard Labs., Palo Alto, CA;Stanford University, Stanford, CA

  • Venue:
  • Proceedings of the 13th international conference on Architectural support for programming languages and operating systems
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recently, the number of cores on general-purpose processors has been increasing rapidly. Using conventional programming models, it is challenging to effectively exploit these cores for maximal performance. An interesting alternative candidate for programming multiple cores is the stream programming model, which provides a framework for writing programs in a sequential-style while greatly simplifying the task of automatic parallelization. It has been shown that not only traditional media/image applications but also more general-purpose data-intensive applications can be expressed in the stream programming style. In this paper, we investigate the potential to use the stream programming model to efficiently utilize commodity multicore general-purpose processors (e.g., Intel/AMD). Although several stream languages and stream compilers have recently been developed, they typically target special-purpose stream processors. In contrast, we propose a flexible software system, Streamware, which automatically maps stream programs onto a wide variety of general-purpose multicore processor configurations. We leverage existing compilation framework for stream processors and design a runtime environment which takes as input the output of these stream compilers in the form of machine-independent stream virtual machine code. The runtime environment assigns work to processor cores considering processor/cache configurations and adapts to workload variations. We evaluate this approach for a few general-purpose scientific applications on real hardware and a cycle-level simulator set-up to showcase scaling and contention issues. The results show that the stream programming model is a good choice for efficiently exploiting modern and future multicore CPUs for an important class of applications.