Attending

Tom, Thomas, Don, Hans, Beau, Xin, Walls, Luke

Minutes
Thomas
  • Working with and testing OpenAD on MPAS prototype
    • Fails on numerous structured, Fortran90 oriented capabilities
    • Working with Jean Utke of Argonne to resolve and understand incompatibilities
    • OpenAD team amenable to resolving issues together. The Argonne team does not yet view OpenAD as a turnkey solution; it still requires iterative tweaking
Don
  • Documentation of wrf4DVar stand alone project
Beau
  • Testing an porting to Macintosh
    • Cleaning up code issues that arose from Mac-specific g95, pgi, and ifort compilation
    • Working on mpich incompatibility
  • wrf+ documentation additions
    • namelist.input document improvements
    • Visual flowchart for test_solve_em()
  • Variable isolation
    • Have isolated relevant variables
    • Working with Thomas N on pressure and rain bugs/ issues
Walls
  • Summary report on optimization efforts
  • Working on bugs with Xin
Xin
  • Resolving bugs in extra cost function for lateral boundary case
  • Working with Walls on wrf+ bug issues
Tom
  • Compiling summaries for end-of-year ACAP status report, due Friday
  • Planning for multi incremental 4DVar
Action Items
  • Thomas:
    • Timing/benchmark data of mpas prototype for acaps report.
    • Check if Tapenade can handle allocated arrays.
    • Ask Argonne about coding standards for OpenAD use.
  • Don: report to Tom
  • Beau: Testing of just mu in wrf+ tests; resolve apparent pressure bug in variable isolation
  • Xin:
    • Work with Walls to capture tests; bugs; and optimization efforts.
    • Document differences between Walls' "original" branch and the reference branch.
    • Investigate 4DVar crash after 20 time-steps (observed by Walls) on Bluefire with 32 procs and more.
  • Walls: investigate if the MPAS AD code generated with Tapenade can be optimized.
  • No labels