It has fundamental importance to reliably find numerical solutions for large-scale nonlinear global optimization problems. In this paper, we report an SPMD parallel algorithm that solves global optimization problem with continuous nonlinear objective function and constraints. Interval branch-and-bound algorithms are the basic algorithms in this paper. Our coarse-grained SPMD algorithm has the advantages of less communication overhead, minor modification of sequential program, reduction of total number of computation, and balanced workload. Initial implementations of our parallel algorithm have shown significant reductions of total number of computation and superlinear speedup.
展开▼
