Supercomputing on Conventional Architectures
There is no accepted definition of the term “supercomputing” apart from the rather glib statement that it is ‘the most powerful machine available’. For the purpose of this lecture we shall loosely extend the definition to include machines within a factor of about ten of the fastest available at any time so that the currently popular ‘mini—supers’ may also be discussed. It is also appropriate to reduce the discussion to computers of general applicability rather than include such special purpose machines as FFT engines for image and signal processing and computers that have been built for a single problem such as Lattice Gauge models.
KeywordsClock Cycle Main Memory Complementary Metal Oxide Semiconductor Memory Bandwidth Floating Point
Unable to display preview. Download preview PDF.
- Hockney R W, in “PDE Software: Modules, Interfaces and Systems”, ed Engquist and Smedsaas, North Holland, Amsterdam 1984Google Scholar