Features of working with dynamic-length data and strongly connected complex data on the example of the task of generating various types of unstructured grids
Abstract:
The discussion is about processing and transmitting of complex and strongly connected data in unstructured mesh generation and adaptive mesh refinement in 3-dimensional regions for distributed calculations. These tasks are examples for transmitting large amounts of such data, that a priori assessment of necessary buffer space is impossible, including the case, when the number of transmitted data is obviously larger buffer size, which is advisable to reserve. Second example consists in transmitting an subdomain in adaptive mesh refinement while dynamically balancing load between processes with saving correct connections between separate elements and recovery processor interconnection graph. Projectionbased tetrahedral mesh generation algorithm is used for generating unstructured tetrahedral grid. Adaptive mesh refinement implemented with using octree-tecnology is used for AMR part. All described software is implemented using C++ language with standard library, MPI library is used for implementing distributed calculations.
