Open AccessBauman Moscow State Technical University is implementing a project to develop operating principles of computer system having radically new architecture. A developed working model of the system allowed us to evaluate an efficiency of developed hardware and software. The experimental results presented in previous studies, as well as the analysis of operating principles of new computer system permit to draw conclusions regarding its efficiency in solving discrete optimization problems related to processing of sets.
The new architecture is based on a direct hardware support of operations of discrete mathematics, which is reflected in using the special facilities for processing of sets and data structures. Within the framework of the project a special device was designed, i.e. a structure processor (SP), which improved the performance, without limiting the scope of applications of such a computer system.
The previous works presented the basic principles of the computational process organization in MISD (Multiple Instructions, Single Data) system, showed the structure and features of the structure processor and the general principles to solve discrete optimization problems on graphs.
This paper examines two search algorithms of the minimum spanning tree, namely Kruskal's and Prim's algorithms. It studies the implementations of algorithms for two SP operation modes: coprocessor mode and MISD one. The paper presents results of experimental comparison of MISD system performance in coprocessor mode with mainframes.