Events
Efficient parallelization of OpenPhase for massively parallel computers
Date: 23.09.2014Place: Materials Science and Engineering MSE, Darmstadt, Germany
Marvin Tegeler
Oleg Shchyglo
Alexander Monas
Ingo Steinbach
Godehard Sutmann
Large scale simulations of multiple phase fields require large amounts of memory and computer time. In order to reduce both the CPU time and memory requirements, an efficient implementation of the multi-phase-field library OpenPhase was developed, which tracks the diffuse interface between different phases and stores them in sparse data patterns. Although efficient, large scale simulations of thousands of different phase fields together with the requirement for large computational grids still exceed the capabilities of single node computer architectures.
