Publication Details




Stability of Advancing Front Local Reconnection for Parallel Data Refinement


Kevin Garner, Thomas Kennedy, Christos Tsolakis and Nikos Chrisochoides.


Published in Virginia Space Grant Consortium (VSGC) 2018 Student Research Conference, Norfolk, VA, April, 2018




The status of a long-term project entailing the parallelization of an industrial-strength sequential mesh generator, called Advancing Front Local Reconnection (AFLR), is presented in this paper. AFLR has been under development for the last 25 years at the NSF/ERC center at Mississippi State University. AFLR is currently used at NASA (including NASA/LaRC) and other government agencies as well as in the aerospace industry such as Boeing. The parallel procedure that is presented is called Parallel Data Refinement (PDR) and consists of the following steps: (i) use an octree data-decomposition scheme to break the original geometry into subdomains (octree leaves), (ii) refine each subdomain with the proper adjustments of its neighbors using the given refinement code, and (iii) combine all subdomain data into a single, conforming mesh. AFLR is shown to maintain weak reproducibility (i.e., able to produce results of the same quality when executed recurrently using the same input) as required in parallel mesh generation. The data-decomposed AFLR implementation is shown to be stable (i.e., guarantees sufficient mesh quality that is comparable to that of the original AFLR software). By the completion of this project, PDR.AFLR will be robust (i.e., generate meshes for the same type of geometries that AFLR can) and will be the first fully functional unstructured mesh generation/refinement application that will be capable of maintaining good parallel efficiency at 10^6 concurrency levels in order to improve end-user productivity.




  [PDF]          [BibTex] 



[Return to Publication List]