CCPP Framework Meeting Minutes 2019 02 07

New metadata format (in separate file):

• converter to new standard already working
• parser not yet working, Steve is on it
• new PR as soon as this is done
• htmlinclude question: GMTB will create the metadata to html converter
• module_name = attribute for schemes (subroutines): not required, can be obtained from scheme names:
  • routine name minus phase -> module name
  • filename can, but doesn't have to be identical to module name
  • metadata filename is Fortran filename minus extension + ".md"
• metadata now much like a Python config file (except multiple identical sections)
• columns as dimensions valid in new metadata standard
• warnings during the conversion process for guessing dimensions?
  • put each warning in one line and pipe to a stderr file?
  • also print a log statement for automatic substitutions
  • betterL use Fortran parser during conversion to figure out which dimension to use
    • will check for inconsistencies during the conversion process
    • this will be of great help for GMTB
• Can Fortran parser be packaged up as a tool for users to generate initial metadata from their code?
  • great idea from Cheryl, create an issue on NCAR GitHub for this
  • should be possible with a few changes

Discussion about host model registry used by CESM etc.

• FV3 doesn't have/use a registry, the host model metadata contains all the information we need right now
  • but relies on a scalar instance of cdata being passed to the CCPP interface (for both dynamic and static build) that contains block number and thread number
  • need to figure out how to handle block number when using cap_gen in the future
  • thread number won't be a problem, since it can be retrieved using OMP_GET_THREAD_NUM()
  • this is not something that we have to address right now, only need to be able to parse new metadata with current ccpp_prebuild system by the end of the month

Visit from NRL colleagues during the week of Feb 25

• will be at CCPP framework meeting during that week, put GFS physics into Neptune via CCPP
• Steve has a call tomorrow (Fri Feb 8) with Edward, Cecilia, Gerhard
  • these folks want to use NRL coupled aerosol component with CCPP going through ESMF interface
  • Ligia and Dom should attend the meeting if possible
  • one of the questions will be where to draw the line between CCPP (generated code) and NUOPC/ESMF
  • can our generator also create an ESMF/NUOPC cap?