Programming Multicore and Many-core Computing Systems

Book Preface

Multicore and many-core computing systems have emerged as an important paradigm in high-performance computing (HPC) and have significantly propelled development of advanced parallel and distributed applications as well as of embedded systems. Multicore processors are now ubiquitous, indeed, from processors with 2 or 4 cores in the 2000s, the trend of increasing the number of cores keeps the pace, and processors with hundreds or even thousands of (lightweight) cores are becoming commonplace to optimize not only performance but also energy. However, this disruptive technology (also referred to as ‘continuum computing paradigm’) presents several major challenges such as increased effort and system-specific skills for porting and optimizing application codes, managing and exploiting massive parallelism and system heterogeneity to achieve increased performance, innovative modeling strategies for low-power simulation, etc. Among these, we would distinguish the challenge of mastering the multicore and many-core and heterogeneous systems – this is precisely the focus of this book!

The emergence of multicore processors has helped in addressing several problems that are related to single-core processors – known as memory wall, power wall and instruction-level parallelism wall – but they pose several other ‘walls’ such as the programmability wall or the coherency wall. Among these, programmability wall is a long-standing challenge. Indeed, on the one hand, program development for multicore processors, especially for heterogeneous multicore processors, is significantly more complex than for single-core processors. On the other hand, programmers have been traditionally trained for the development of sequential programs, and only a small percentage of them have experience with parallel programming.

In fact, in the past only a relatively small group of programmers interested in HPC was concerned with the parallel programming issues; the situation has changed dramatically with the appearance of multicore processors in commonly used computing systems. Traditionally parallel programs in HPC community have been developed by heroic programmers using a simple text editor as programming environment, programming at a lowlevel of abstraction and doing manual performance optimization. It is expected that with the pervasiveness of multicore processors, parallel programming will become mainstream, but it cannot be expected that a mainstream programmer will prefer to use the traditional HPC methods and tools.

The main objective of this book is to present a comprehensive view of the state-of-the-art parallel programming methods, techniques and tools to aid the programmers in mastering the efficient programming of multicore and many-core systems. The book comprises a selection of twenty-two chapter contributions by experts in the field of multicore and many-core systems that cover fundamental techniques and algorithms, programming approaches, methodologies and frameworks, task/application scheduling and management, testing and evaluation methodologies and case studies for programming multicore and many-core systems. Lessons learned, challenges and road map ahead are also given and discussed along the chapters.