Multi-core processor

Tuesday, November 3, 2009

A multi-coreprocessor is a processing system, that from two exists or more independent cores. The cores are on individual integrated circuit die (well-known as splinter multiprocessor system or CMP) usually integrated, or they can be integrated on repeated cubes in individual splinter package. A many core processor is one, in which the number cores is enough large that traditional multiprocessor system techniques are no longer efficient - this threshold is somewhere in the distance of some 10 cores - and probably a net on splinter require. A double core processor contains two cores, and a Viererkabelkern processor contains four cores. A multi-core processor inserts parallel processing in individual physical package.

Cores in a multi-core device can be connected together firmly or loosely. E.G. cores can or know buffer memories possibly not divide, and them methods of report exceeding can would bring in or the divided memory Zwischen-kern communication. Topologies of the general net to the linkage cores cover: Ring, measure into one another 2 and lateral ripple seizes bus. All cores are identical in the homogeneous multi-core systems and them are identical not in the heterogeneous multi-coresystems. Straight ones as with particular processor systems, know cores in the multi-coresystems architecture as superscalar, VLIW, which, SIMD would bring processing vector in or multithreading.

Multi-core processors are used over many areas of application inclusively at most: universally, embedded, net, digital signal processing and diagrams. The quantity of the achievement won with use of a multi-core processor depends strongly on the software algorithms and - implementation. In particular the possible profits are limited by the break of the software, which can be made parallel, in order to run on repeated cores at the same time; this effect becomes by Amdahl' described; S-law. In the best case embarrassingly parallel problems so called can realise Speedupfaktoren close of the number cores. Many typical applications however realise not such large Speed up faktoren and therefore, are the Paralellisierung of software an important current topic of the research.

0 comments: