Adaptive dynamic relaxation algorithm for non-linear hyperelastic structures Part III. Parallel implementation Original Research Article

An efficient parallel implementation of the Adaptive Dynamic Relaxation (ADR) algorithm is developed. It is designed to minimize the extent and frequency of interprocessor communication. This version is implemented on a 128-processor Intel hypercube and on a cluster of 16 engineering workstations using PVM. Relative speedups are presented for the non-linear static analysis of two- and three-dimensional hyperelastic systems involving frictionless contact. Impressive speedups are achieved on the hypercube and demonstrate the high scalability of the ADR algorithm. Good speedups are also achieved using PVM, suggesting that PVM is a viable parallel-processing resource for this application.