No abstract

No abstract

No abstract

Critical to the scalability of parallel adaptive simulations are parallel control functions including load balancing, reduced inter-process communication and optimal data decomposition. In distributed meshes, many mesh-based applications frequently access neighborhood information for computational purposes which must be transmitted efficiently to avoid parallel performance degradation when the neighbors are on different processors. This article presents a parallel algorithm of creating and deleting data copies, referred to as ghost copies, which localize neighborhood data for computation purposes while minimizing inter-process communication. The key characteristics of the algorithm are: (1) It can create ghost copies of any permissible topological order in a 1D, 2D or 3D mesh based on selected adjacencies. (2) It exploits neighborhood communication patterns during the ghost creation process thus eliminating all-to-all communication. (3) For applications that need neighbors of neighbors, the algorithm can creatennumber of ghost layers up to a point where the whole partitioned mesh can be ghosted. Strong and weak scaling results are presented for the IBM BG/P and Cray XE6 architectures up to a core count of 32,768 processors. The algorithm also leads to scalable results when used in a parallel super-convergent patch recovery error estimator, an application that frequently accesses neighborhood data to carry out computation.

The scalable execution of parallel adaptive analyses requires the application of dynamic load balancing to repartition the mesh into a set of parts with balanced work load and minimal communication. As the adaptive meshes being generated reach billions of elements and the analyses are performed on massively parallel computers with 100,000's of computing cores, a number of complexities arise that need to be addressed. This paper presents procedures developed to deal with two of them. The first is a procedure to support multiple parts per processor which is used as the mesh increases in size and it is desirable to partition the mesh to a larger number of computing cores than are currently being used. The second is a predictive load balancing method that sets entity weights before dynamic load balancing steps so that the mesh is well balanced after the mesh adaptation step thus avoiding excessive memory spikes that would otherwise occur during mesh adaptation.