The TOP500 website (https://www.top500.org) lists the current most powerful computer systems on the planet. The top system at the end of 2017 was capable of more than 125000 teraflops (trillions of floating point operations per second). That this achievement falls far short of the expectations and needs of today’s high-performance computing (HPC) practitioners is nothing short of amazing. They want more exascale systems, which are literally thousands of times more powerful than the current leaders. To achieve this goal, pushing the limits of processing, memory, and network physics, researchers are designing new components and architectures. Plus, they are focusing on the portability of their efforts--sharing scalability and performance concepts and solutions in order to solve the most difficult computational problems in astrophysics and cosmology, climate science, high-energy physics, chemistry, and seismology, to name but a few.
The editors have assembled a comprehensive volume of cutting-edge research “focusing on the programming practices to achieve scalability ... while at the same time maintaining architectural and performance portability for different computer technologies.” In two dozen chapters, more than 140 authors and contributors from government supercomputer centers, academia, and industry detail their efforts to meet the exascale requirements of their research. The included papers cover methods to enable “power portability” (sharing techniques for power saving and management across differing architectures), along with sharable programming and debugging tools and libraries. Most of the papers highlight such work, detailing various use cases in wide-ranging fields such as computational chemistry, turbulent combustion, material science, and plasma physics, among others. Discussions include investigations of “massively threaded and many-core exascale systems,” as well as innovative grid architectures.
It is clear from the research presented that there is indeed a compelling need for evermore-powerful computational tools and systems, while at the same time understanding and dealing with the difficult problems of extreme physical scalability and infrastructure power management. One very useful feature of the book is its index, something not always included in such collections of research papers. This helps in looking up terms and references common to multiple papers.
The editors--recognized experts from the Argonne, Lawrence Berkeley, and Oak Ridge laboratories--have produced a valuable contribution to the future of extreme scale computing, useful for other HPC researchers and students in this rapidly evolving field.
More reviews about this item: Amazon
