High Performance Computing and Communications Glossary 2.1

A significant part of the material of this glossary was adapted from material originally written by Gregory V. Wilson which appeared as "A Glossary of Parallel Computing Terminology" (IEEE Parallel & Distributed Technology, February 1993), and is being re-printed in the same author's "Practical Parallel Programming" (MIT Press, 1995). Several people have contributed additions to this glossary, especially Jack Dongarra, Geoffrey Fox and many of my colleagues at Edinburgh and Syracuse.

Original version is from NPAC at <URL:http://nhse.npac.syr.edu/hpccgloss/>

Original author: Ken Hawick, khawick@cs.adelaide.edu.au

See also the index of all letters and the full list of entries (very large)

Sections: A B C D E F G H I J K L M N O P Q R S T U V W X Y Z


n-1/2 (n.) (pronounced as n-one-half, usually written with a subscript.) The minimum vector length on which a pipelined architecture delivers one half of its theoretical peak performance. The larger n-1/2 is, the longer calculations must be to amortize the startup cost of the pipeline. This measure was coined by Hockney and Jesshope in 1988. See also r-inf.

NC (n.) The class of parallel algorithms that can run in polylogarithmic (polynomial in the logarithm of the problem size) time on a polynomial number of processors in the PRAM model.

necklace (n.) the nodes a data item travels through in response to a sequence of shuffles.

need-predictable (adj.) Used to describe a concurrent algorithm in which the need for but not the nature of point-to-point communication between nodes is known prior to program execution. Need predictable problems are loosely synchronous.

NeTBIOS (n.) Network Basic Input/Output System that provides a Session layer interface between network applications running on a PC and the underlying protocol software of the Transport and Network layers on the OSI model. Normally a LAN protocol.

network(n.) A physical communication medium. A network may consist of one or more buses, a switch, or the links joining processors in a multicomputer.

neural network (n.) Man made devices using interconnects and processing capabilities suggested by models of the cortex are termed neural networks. These systems are widely used for optimization problems including content addressable memories and pattern recognition.

NFS (n.) network file system ia a protocol developed to use IP and allow a set of computers to access each other's file systems as if they were on the local host.

NNTP (n.) network news transport protocol is used with uucp to transfer news across the Usenet.

node (n.) generic term used to refer to an entity that accesses a network.

non-blocking(adj.) An operation that does not block the execution of the process using it. Usually applied to communications operations, where it implies that the communicating process may perform other operations before the communication has completed. See also blocking.

non-deterministic model (n.) a task model in which the execution time of each task is represented by a random variable.

non-uniform memory access (adj.) See NUMA.

NP complete (adj.) A term used to describe a problem if it can only be solved in polynomial time by non-deterministic methods. In practice such problems are "hard" and are only solved by a number of heuristic methods that only give approximate answers near to the optimum solution.

NUMA (adj.) Nonuniform memory access; not supporting constant-time read and write operations. In most NUMA architectures, memory is organised hierarchically, so that some portions can be read and written more quickly by some processors than by others. See also UMA