1 / 10

Component-Based Parallel Meshing Support for Accelerator Applications

Component-Based Parallel Meshing Support for Accelerator Applications. Timothy J. Tautges Sandia National Labs Advanced Accelerator Science Project Meeting Fermilab, Aurora, IL Aug 11-12, 2004.

gur
Download Presentation

Component-Based Parallel Meshing Support for Accelerator Applications

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Component-Based Parallel Meshing Support for Accelerator Applications Timothy J. Tautges Sandia National Labs Advanced Accelerator Science Project Meeting Fermilab, Aurora, IL Aug 11-12, 2004 Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company,for the United States Department of Energy under contract DE-AC04-94AL85000.

  2. Outline • Parallel tetrahedral mesh generation • Shape optimization, mesh and geometry issues

  3. Parallel Tetrahedral Mesh Generation 1. Partition assembly model over processors: P3 P4 P1 P2 P3 P3 2. Mesh interface surfaces & communicate (concurrent, asynchronous) 3. Mesh non-interface surfaces & volume interior (concurrent) 4. Output mesh

  4. Parallel Tetrahedral Mesh GenerationParallel Performance

  5. Parallel mesh time of 'h60_quarter' with size 0.2 (280 million elements) 3250 3000 2750 2500 2250 time (sec) 2000 1750 1500 1250 1000 750 500 0 5 10 15 20 25 30 35 40 # of processors Parallel Tetrahedral Mesh GenerationMesh Size, Quality • Mesh quality: MetricAverageStd DevMinMax Sca Jac 0.67 0.12 0.02 1.0 Shape 0.85 0.09 0.02 1.0 Cond No. 1.19 0.14 1.00 45.3 *** More recent data: MetricAverageStd DevMinMax Shape 0.10

  6. Geom w/ BC’s, Decomp Mesh Mesh Mesh InitGeom Parallel Tetrahedral Mesh GenerationCode Components P0 P1 … (serial) CGM CGM CGM CUBIT CGM PTET PTET PTET MOAB MOAB MOAB Disk IO IPC (MPI) Function call

  7. Parallel Tetrahedral Mesh GenerationWhat’s Next • BC’s (material groups, dirichlet/neumann sets) • Done (since H. Kim returned from SLAC) • Higher-order tet elements • As of this morning, appears to work for smaller models, still testing on DDS • Mesh quality • Should be no worse than serial for DDS • Mesh assembly / single-file mesh IO • Developing parallel HDF5-based solution in MOAB

  8. Shape OptimizationGeneral Approach • Constant geometric & mesh topology over parametric variations • Smoother, less-stiff optimization • Amortization of direct solution matrix factoring • Just need surface mesh sensitivities wrt parameter variations, not volume sensitivities • Entire procedure must be: • In-situ on parallel machine • Automatic

  9. Omega3P Shape Optimization Lifecycle geometric model optimization Omega3P meshing Omega3P Sensitivity meshing sensitivity

  10. Shape OptimizationSciDAC Interactions, Leveraging • Interfaces to mesh & geometry data through TSTT interface specs • Allows mix ‘n match of tools to find best overall capabilities • Other TSTT tools plug in through mesh interface • Demonstrates rapid access to advanced components • Solid model-based geometry (SNL) • Adaptive refinement (RPI) • Adaptive, guaranteed-quality mesh smoothing (SNL/LLNL) • Real application • High-end computing absolutely necessary • Lots of other accelerator-related applications on deck

More Related