Deeply Coupled Computing: Parallel Processing Architecture for Mass Market Applications
Solving many of the most important scientific and engineering problems today requires the use of High Performance Computers. There are, however, several major issues currently facing the field. These include: the elusiveness of attaining high percentages of peak hardware performance; the difficulty of developing, debugging, and optimizing applications; and the high price and operating costs limiting access by many organizations with high computational needs.
A new type of computing called Deeply Coupled Computing, combined with system-on-a-chip technology, addresses these issues. In addition to increasing the performance of existing high performance programs, Deeply Coupled Computing enables the application of high-performance computing to a broad range of problems that cannot be solved using today's architectures. This paper explains how CPU Tech applied Deeply Coupled Computing to one such problem and achieved a substantial improvement.