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.