What are the advantages of super parallel computers?

When applied to parallel computers, appropriate parallel algorithms must be adopted. The parallelism of the algorithm directly affects the efficiency of the computer and the actual speed of solving problems. 10.4, Science and Technology Daily 199 1 65438+ was rated as "Top Ten Scientific and Technological Achievements in the World 1990", and the sixth item said: "On June 20th, the United States announced that it had successfully transformed a 1000 computer system into a supercomputer. Since 1990 British scientist Pollard and others invented a new parallel computing method, the computer only ran for a few months to complete the factorization of large numbers. If you use the old method on a computer with 1059 times per second, it will take 1059 years.

The development of science and technology and society requires parallel computers to have higher speed and larger capacity. For example, in order to make the weather forecast more accurate, it is necessary to add spatial points and new physical parameters, and just doubling the points in all directions will make the fastest computer unable to complete the work in one day. In addition, the introduction of images and sounds also makes computers require higher speed and capacity. With the continuous development, maturity and breakthrough of large-scale parallel processing technology, the super computer with TFLops speed limit is just around the corner, and it will succeed in remote collaboration, large-scale data sharing and multimedia support, so that the computer can truly "see in all directions and hear in all directions" in an instant.

(1) pipeline technology: like the pipeline in industrial production, it improves the performance of computer operation and control components.

(2) Functional structure: independent components are applied to perform different functions, and they are allowed to process all kinds of data at the same time. A computer with pipeline and functional structure technology is called a vector computer.

(3) Array structure: provide an array composed of many processing components with the same function, which perform the same operation at the same time under unified control, but the objects (data) of the operation are different. This kind of parallel computer is called single instruction stream and multiple data streams.

(4) Multi-processor structure: multiple processors are provided, which can simultaneously perform their respective operations and communicate with each other. This kind of parallel computer is called multi-instruction stream and multi-data stream.

(5) Large-scale parallel processing (MPP) technology: Through high-speed network communication, multiple computers can work at the same time and complete the same task under a special management mechanism.