Skip to content
Snippets Groups Projects
  1. Jan 17, 2024
    • Lisa Julia Nebel's avatar
      Add flag useFixedOrder to the coarsen method of AMGs ParallelIndicesCoarsener. · 4b4c92ca
      Lisa Julia Nebel authored and Lisa Julia Nebel's avatar Lisa Julia Nebel committed
      If set to true, the indices for the coarser grid will be created in a fixed order, making parallel runs reproducible but the runtime is possibly not ideal.
      If set to false (which is the default), the order depends on the order of messages recieved from the processes responsible for the respective parts of the finer grid.
      Then the indices on the coarser grid may differ from run to run.
      4b4c92ca
  2. Jan 08, 2023
  3. Sep 30, 2022
  4. Jul 29, 2021
  5. Jul 28, 2021
  6. Mar 16, 2021
  7. Mar 10, 2015
  8. Feb 04, 2014
  9. Mar 11, 2013
  10. Feb 25, 2013
  11. Dec 18, 2012
  12. Jan 20, 2011
  13. Dec 06, 2010
  14. Nov 24, 2010
  15. Sep 10, 2010
  16. Dec 04, 2009
  17. Oct 13, 2009
  18. Sep 04, 2009
  19. Jun 30, 2009
  20. Jun 22, 2009
  21. Feb 09, 2009
  22. Sep 10, 2008
  23. Jul 13, 2008
  24. Apr 20, 2007
  25. Mar 16, 2007
  26. May 01, 2006
  27. Mar 28, 2006
  28. Mar 27, 2006
    • Markus Blatt's avatar
      Bugfix. If not all indices are present in the index set, even if the · 9015ee87
      Markus Blatt authored
      index of the first vertex is not in the set other vertices of the same
      aggregate might be represented there. Therefore we now check all of
      them whether they present anf if one is present the coarse index set
      will get an according index for the aggregate.
      
      [[Imported from SVN: r545]]
      9015ee87
  29. Feb 15, 2006
  30. Feb 09, 2006
    • Markus Blatt's avatar
      Improved GlobalLookup allocation and deallocation. · 50e6adc9
      Markus Blatt authored
      [[Imported from SVN: r505]]
      50e6adc9
    • Markus Blatt's avatar
      AMG runs in parallel now. · 176a5715
      Markus Blatt authored
      There was huge bug in the numbering of the aggregates:
      
      Due to the coarsening of the index set the matrix on the coarse level are ordered like this. First come the row corresponding to vertices owned by the process
      (0, 1, ..., n) and the the rows owned by other process (n+1, n+2, ..., N).
      
      Unfortunately during the setup of the sparsity pattern of the matrix this fact was neglected :-(
      
      Fixed this now.
      
      [[Imported from SVN: r500]]
      176a5715
  31. Jan 27, 2006
  32. Jan 11, 2006
  33. Dec 16, 2005
  34. Dec 14, 2005
  35. Dec 08, 2005
Loading