| 摘要 |
A method is disclosed for selecting from among a plurality of computer processors connected together to distribute portions of an overall numerical problem to be solved among the processors to optimize the speed of communication between the processors in solving the overall problem. The method utilizes various factors in determining the selection, including whether any of the processors are located within the same computer or workstation, whether any of the processors share a network switch, whether any of the processors are located on the same sub-network, the speed of the individual network connections, and any user-configurable groupings of the processors. Also considered are the known communication requirements of the problem to be solved. A simulated annealing algorithm is then used to determine the optimal arrangement of the processors in solving the problem.
|
