Description: Computations associated with PC3D elements and contact between SPH particles and finite element-based surfaces are now implemented in domain parallel, enabling better parallel scaling if multiple CPUs are used. An evolving domain decomposition is used for SPH analysis to avoid large spatial overlap among SPH domains (and, therefore, to maintain good parallel scaling) after large relative motions of SPH particles.

The speedup for multiple-CPU runs of the bottle drop test (see “Impact of a water-filled bottle,” Section 2.3.2 of the Abaqus Example Problems Guide) is plotted in Figure 5–2. It shows that the speedup for a two-CPU run is 1.79, and the performance efficiency is 0.89; for a four-CPU run, the speedup is 2.91, and the performance efficiency is 0.73.