Massively parallel computing applied to the one-dimensional bin packing problem

The author presents a parallel algorithm for packing N⩾n² pieces in a massively parallel computing environment of n² processors operating in single-instruction multiple-data (SIMD) mode. The heuristic approach that is developed utilizes a grid topology that structures the input to increase packing efficiency without sorting the data set. The results of implementing this algorithm on a Connection Machine 2 system are compared with the results obtained from parallel versions of two traditional bin packing algorithms. It is concluded that the use of data partitioning to initialize the packing appears to be a practical method of allocating the packing workload to a set of parallel processors while maintaining the integrity of the packing algorithm. The results support the contention that parallelism offers opportunity for improvement in the efficiency of packing algorithms and that the SIMD model of computation with its data parallel programming approach is appropriate for bin packing problems