Computing Reviews

Message-passing programming for embedded multicore signal-processing platforms
Hung S., Chiu P., Tu C., Chou W., Yang W. Journal of Signal Processing Systems75(2):123-139,2014.Type:Article
Date Reviewed: 03/23/15

The present-day tendency to use heterogeneous multicore profiles in digital signal processing applications is addressed in this paper, and the authors propose an optimized architecture to overcome difficulties in software programming, integration, and execution of such applications. Their solution especially addresses the challenging issues for embedded multiform platforms, where resource memory management is critical.

The paper addresses some existing standard libraries, such as the message passing interface (MPI) and the multicore communications application programming interface (MCAPI), and investigates their weaknesses in the context of some the more popular video coding platforms (PAC Duo, x86, and CELL). The authors propose MSG, a new, more flexible, and portable message passing library, consisting of three layers with different degrees of generality and different functionalities, to assume an essential subset of the previously mentioned packages.

The first level groups the most important communication functions of MPI, the middle level gathers control protocols and buffer management functions, and the bottom layer addresses the specificity of the implementing platform. Most of the functions are imported from the MPI library. The inter-core communication is designed to reduce latency of message transmission and relies on the user space copy memory scheme based on relay buffer pools.

The paper evaluates these design issues on several implementations of the MSG library. It is based on a rigorous knowledge of the actual context and of the advances in the field, a thorough examination of the implementing platforms, and sound justification of their options. This paper is a valuable reference for software designers and developers of embedded multicore architectures.

Reviewer:  Svetlana Segarceanu Review #: CR143256 (1506-0491)

Reproduction in whole or in part without permission is prohibited.   Copyright 2024 ComputingReviews.com™
Terms of Use
| Privacy Policy