


Since 1978 Published in Sarov (Arzamas16), Nizhegorodskaya oblast 
RUSSIAN FEDERAL NUCLEAR CENTER 
ALLRUSSIAN RESEARCH INSTITUTE OF EXPERIMENTAL PHYSICS 

Русский  English

ON EXPERIENCE OF PARALLELING GAS DYNAMIC PROBLEM COMPUTATION USING DMK PROGRAMS ON MULTIPROCESSOR SYSTEMS
V.N. Motlokhov, V.V. Rasskazova, I.D. Sofronov VANT. Ser.: Mat. Mod. Fiz. Proc 1996. Вып.3. С. 38.
Methods and techniques of computation paralleling are demonstrated which authors experienced when parallel computation arrangement was of narrow directed character persued specific objectives. Algorithm of paralleling DMK (2D gas dynamic problem computation on nonregular polygonal and regular quadrangular grids) program is proposed.
 LOGISTICS OF SMOOTH GRID GENERATION IN COMPOSITE DOMAINS
G.P. Prokopov, O.I. Sofronova VANT. Ser.: Mat. Mod. Fiz. Proc 1996. Вып.3. С. 917.
Problems of description of grid and domain structure allowing to automatize to great extent the process of cutting composite domain into rectangular structure blocks in plane of grid node numbers. It is possible to solve grid generating equations obtained using variational functionals or direct approximation of differentional equations by means of local algorithms which can be easily parallelized for multiprocessor computer systems. The logistics described allows to provide grid smoothness for the entire composite domain.
 PARALLEL ALGORITHMS FOR SOLVING 3D HEAT CONDUCTION EQUATION AND RESULTS OF DEMONSTRATION PROBLEM COMPUTATIONS ON EIGHTPROCESSOR COMPUTER SYSTEM MP3 WITH DISTRIBUTED MEMORY
O.I. Butnev, A.N. Bykov, B.L. Voronin, S.I. Skrypnic, I.D. Sofronov VANT. Ser.: Mat. Mod. Fiz. Proc 1996. Вып.3. С. 7784.
The paper presents the description of parallel algorithms for solving 3D heat conduction equation and computation results of test problem on cube cooling obtained using eightprocessor computer system MP3 with distributed memory. Differentia] and finitedifference equations, scheme of spliting by directions, finitedifference equation solving technique are given. Two approaches to massively parallel computation arrangement are implemented and studied. Parallelization efficiency dependence on technique of problem geometrical decomposition for two modes of MP3 processor loading is investigated. Computations were carried out using orthogonal and nonorthogonal grids.
 [ Back ] 






