PCL - A LANGUAGE FOR PARALLEL OPTIMIZATION ON DISTRIBUTED WORKSTATIONS

In this paper a software prototype is presented for the parallel solution of nonlinear optimization problems on distributed UNIX-workstations by applying iterative algorithms. The Parallel Algorithm Control Language (PCL) and a distributed blackboard model axe the major parts of the developed system. The system uses remote procedure calls (RPC) as a powerful primitive for synchronization between the distributed processes. The PCL software describes the configuration (architecture) of the general system and the parallel use of the different algorithms. A distributed blackboard model, however, has been designed to solve the control and communication problems. Some parallel optimization strategies are demonstrated by several programming examples.