Efficient parallel multirate time stepping for accelerating explicit discontinuous Galerkin computations.
Bruno Seny, Jonathan Lambrechts, Vincent Legat et al. · Digital Access to Libraries (Université catholique de Louvain (UCL), l'Université de Namur (UNamur) and the Université Saint-Louis (USL-B)) · 2011
The development of suitable and fast time integration methods for ocean modeling con-\nstitutes an important challenge. No single time-discretisation works well for all physical\nprocesses in a complex marine model, as dierent subsystems have widely dierent charac-\nteristics in terms of time scales, dynamic behaviour, and accuracy requirements. We believe\nthat building appropriate time stepping strategies for multi-scale computations will enable\nus to gain an order of magnitude.\nIndeed, unstructured-mesh generation processes are complex and, even though it is possible\nto control average element sizes in specic regions of the domain, it is not the case for each\nelement size. The smallest element is usually much more smaller than the criterion that was\nprescribed a priori and it determines the stable time step for the entire model. Therefore, the\ncomputational eciency of explicit time-stepping methods may be drastically low.\nMultirate schemes represent a class of methods that use various time steps on dierent grid\ncells. The strategy consists in splitting the domain in a smart way. Grid cells are gathered in\ndierent groups that satisfy the local CFL stability conditions for a certain range of time steps.\nStandard explicit Runge- Kutta methods are applied on independent partitions while buer\ngroups have to be introduced between them, with adapted methods, in order to accommodate\nthe transitions between them. These methods are especially suited for the Discontinuous\nGalerkin spatial discretization. Nevertheless, development of such methods is still challenging.\nBoth, stability requirements and conservation properties should be satised. Two approaches\nare explored. Constantinescu introduced a conservative 2nd order scheme while Schlegel proposed a 3rd method that is, unfortunately, not conservative.\nLarge-scale applications like the Great Barrier Reef require the use of parallel computers.\nSome kind of load balancing strategy has to be supplied to accomodate multirate schemes:\nindeed, small elements have a higher cost than large elements in such a strategy. Moreover,\nsmall elements at inter-processor interfaces will require more frequent updates. The key\nidea consists in creating an optimized mesh partition in a way that the amount of grid\ncells of the dierent multirate groups is ideally the same on each computer core. However,\na compromise should also be found between the eective work on each processor and the\namount of communications between them.
Cited by 1 works.