Paper Authors

Abstract

NOTE: The first page of text has been automatically extracted and included below in lieu of an abstract

Educational Goals for Embedded Systems in the Multicore Era

Abstract

Embedded systems are becoming increasingly sophisticated. For example, today's automobiles have requirements for adaptive engine control to meet emissions and fuel-economy standards, advanced diagnostics for repair, reduction of wiring, new safety features, and new comfort and convenience features. The software required to support this feature set is quite complex and has strict performance requirements, and the hardware must operate well in extreme climate conditions with limited power resources. Yet, in order to keep materials costs low automobile manufacturers must deploy these capabilities using as few microprocessors as possible. For these reasons, the microprocessors used in automotive applications must be both high- performance and energy-efficient. These demands are at cross purposes with traditional high- performance microprocessor design, and the industry has responded with innovative embedded multicore architectures. Instead of throttling up the performance of a single processor core (which is very power intensive), a new breed of microprocessors incorporates multiple low power processor cores onto a single chip. This approach results in a higher throughput system capable of running many concurrent threads of computation in an energy-efficient fashion. These changes in the microprocessor landscape have satisfied the need for high-performance, energy-efficient processors. However, they have left engineers who are experienced only with single-core embedded systems and sequential programming languages at a disadvantage, lacking exposure to critical concepts required to design, program, debug, and optimize this new breed of embedded systems. The problem will surely be exacerbated in the not-too-distant future when (as many experts predict) the multicore era transitions to the manycore era. For these reasons, it is time to consider some new educational goals which will prepare future engineers for the multicore era and beyond. This paper discusses the essential concepts that should be included in an undergraduate computer engineering curriculum that takes embedded multicore systems into account.

Introduction

Many people believe that the incorporation of multiple cores on a single chip began in the high- end workstation and desktop market to mitigate the ever-increasing power demands of traditional performance enhancement techniques such as clock frequency scaling, increased pipeline lengths, and super-scalar instruction issue1. But the truth is that many embedded applications were employing multiple cores on a chip long before these high profile chip multiprocessors caught the attention of the public. For example, cell phone chips of the late 1990s commonly incorporated a general purpose applications processor core alongside a digital signal processor core.

A brief examination of the product portfolios of major semiconductor companies will show that multicore chips are becoming part of the mainstream offerings in many embedded application domains including consumer, industrial, networking, medical, military, gaming, and automotive. This change is happening very rapidly, and has caused consternation within the embedded programming community. Many in the industry believe that the concerns stem from two main

• Citation

• Format
  • APA
  • APA - LaTeX bibitem
  • MLA
  • MLA - LaTeX bibitem
  • Bibtex
  • EndNote - RIS

Holt, J., & Shi, H., & Stern, H. (2009, June), Educational Goals For Embedded Systems In The Multicore Era Paper presented at 2009 Annual Conference & Exposition, Austin, Texas. 10.18260/1-2--4739