THE POSSIBILITY OF INCREASING COMPUTATIONAL PERFORMANCE IN MULTIPROCESSOR COMPUTER SYSTEMS DATA PROCESSING

UDC 681.32

S.M.Tolstyh

The paper discusses ways of increasing performance of modern computer systems. This can be achieved through the use of a multiprocessor architecture, as well as due to the involvement of heterogeneous platforms on the basis of specialized computational modules. Due to the fact that there is a finite speed in the transmission of information among the elements of computing machines, we can come to limit productivity in calculations. There are two basic models of computations, which are used in the existing conditions model computer and the model of collective of calculators. From the point of view of the processor types is possible to divide multiprocessor systems into homogeneous and inhomogeneous, single-level and hierarchical. Classification one-level multiprocessor systems from the point of view of the communication type processor memory data is given. As an advantage of a multiprocessor system with shared data memory, it can be noted that shared memory is used only in order to implement interprocessor exchange is shown.

Keywords: CPU, performance, parallel computing, data processing.

Full text:
Tolstyh_4_16_1.pdf