Conclusion: Multi-core Processor Architectures Are Here to Stay

Abstract

It is clear the that the Multi-core Era has already started some time ago. Multi-core processor architectures clearly solve a number of issues related to power dissipation, the frequency wall etc. However, the adoption of the such platforms poses a number of new challenges of which programmability is definitely one of the toughest ones to crack. Even more so because parallelising compilers and parallel programming languages were a hot research topic in the eighties/nineties and no dominant technology or solution has emanated from that effort. Evidently and maybe most importantly because each time the next generation processor, with an ever higher operating frequency and transistor density, provided the necessary performance upgrade needed. With the arrival of the multi-core architectures, we are witnessing a renaissance of the parallel computing and programming paradigm. Where many research initiatives explore the possibility of designing a new language that would provide the necessary constructs and hooks to efficiently exploit the available hardware resources, they usually represent a radical rupture with the past leading to incompatibility between the existing code base and the latest programming languages. And exactly this incompatibility may turn out to be the show-stopper for many of those emerging approaches. Especially in the Embedded Systems industry, applications tend to have life span covering decades during which period both software and hardware have to be maintained and upgraded when necessary.