Parallel Computer Models 2013

traditional sequential computers are based on the model introducivo the forties by "JohnVon Neumann", which consists of an central processing unit (CPU) and a memory. The computational model uses a sequence of instructions and operates on a data stream.computers of this type are called computers SISD (Single Introduction Single Data).The speed of a computer SISD is limited by two factors: the speed of execution of instructions and the speed at which information is exchanged between memory and CPU. The latter can be based incrementadda increase the number of channels through which data can be accessed independently to each of foma indepependiente (memory interleaving).Another way is mejroar that speed memory using fast link (caches), which act as an intermediary between the main memory, much slower, and the CPU, substantially increasing system performance.On the other hand, the instruction execution rate can be increased by overlapping execution of different stages of various instructions (pipelining).These three techniques of improving the speed of a SISD computers are commonly used in high performance computaordas SSID, however has its limitations.An alternative way pra incrementear the instruction execution rate is to use multiple processing units and multiple memory units interconnected in some way (parallelism). procesameinto speed of the system grows as the number of CPUs and memory undidades increases.If we think of the number of processors used in the construction of computers, we are referring to the granularity.Thus, a parallel computer can be comprised of a small number of high power procedores cn, so as comunciacion calculation, or by a large number of coarse-grained processors, while the second, fine-grained are computers.also in my opinion I have ever commented, computers differ in other characteristics such as the multiplicity of the flow of instructions and data, the organization of the address space or the interconnection network.