A COMPUTING METHOD PARALLELIZATION FOR LASER FUSION PROBLEMS

G.V. Dolgoleva
VANT. Ser.: Mat. Mod. Fiz. Proc 1999. Вып.1. С. 42-44.

      SND code parallelization pursued two aims. The first aim is to have computational multiprocessor version of the code and the second one is pure methodical: to ascertain if the codes constructed on the principle of splitting by physical processes can be effectively parallelized. A typical class of the tasks which solution requires taking account of a large number of various physical processes — are the controlled fusion (laser or heavy ionic) tasks. A great number of physical processes are traditionally taken into account in the codes sequentially. Numerical algorithms depend on execution priority of physical processes. Difficulty of parallelization of algorithms, constructed according to the principle of splitting by physical processes consists in strict priority of their execution. It was desirable when parallelizing to retain the difference schemes used in a program and tested by many years operating experience.
      Physical processes computed with SND code are enumerated. A scheme and a method parallelizing mechanism are described, the examples of multiprocessor numerical calculations are given.

      The numerical study results for compression and combustion of a capsule, proposed for investigation in NIF laser project (USA) are presented. The relation of energy output from a capsule and the relation of ignition margin for deuterium-tritium gas to copper atom concentration αCu in beryllium shell and temperature on its surface are constructed using ALF code. The shell composition is determined to reach the maximum energy output from a target. For these studies the spectral and one-group ranges for beryllium and copper mixture were obtained using PERST-3 code within the scope of MIXER technology. To estimate the assurance of range values they are compared with data, resulting from measurements and computations using other programs.