CCPP Framework Meeting Minutes 2019 05 09

GMTB Statement of Work (SOW) more or less approved, from July 1st for one year

• talks about removing dynamic build in favor of multi-suite static build
• this involves some work in the host models (SCM, NEMSfv3gfs)
• wrt cap_gen
  • note that additional variables can be passed from host model to physics (host variables) with new cap_gen
  • no exact timeline, try to start transitioning to cap_gen second half of the summer
• generally program office reluctant to forward-looking development, more focused on consolidation efforts
• GMTB has 100k$ (half an FTE) for ccpp-framework development
• most of the funding for GMTB goes into transitioning work and establishing CCPP in all systems that need it (regional, global workflow)

Challenges with other models

• WRF has DDTs in the host models and uses only components (Fortran standard types) in the physics, this is the ideal case
• chemistry uses DDTs that get passed between the routines, but are not used externally; easy, too
• GFS way of passing DDTs for entire blocks not yet fully implemented
  • Steve is trying to get alternative funding (since May's proposal was not accepted), because more people at NCAR are interested in running GFS physics
  • needs to be a joint effort between GMTB and NCAR

Build system for host model/CCPP with cap_gen

• cmake boilerplate runs cap_gen, adds all the targets
  • this is further than what we are doing now, but in the same direction
  • there is a hello-world example that shows how this works
• boilerplate: given a target and three variables (host files/scheme files/suite files arguments)
• issue with host-model specific versus file-specific compiler flags
  • how to combine?
  • host-model specific cmake script (basically pulling out what is in current ccpp-physics' CMakeLists.txt)
  • will be easier once the dynamic build is gone (no more compiling of ccpp-framework source code)
• try to make transition to new_metadata and cap_gen as smooth as possible (expose as few changes as possible to the users/developers)

UFS SIP meeting presentation by Ligia 2019/05/15

• 30 mins on "CCPP update"
• high level, no technical details
• Steve to provide a slide or two from the NCAR perspective

Review of "Big Document" (due May 13)

• Steve will look at this on the weekend
• this big document will go under version control in ccpp-framework (rst/sphinx) and replace the Developer's Guide

Alex Reinecke's GFS_typedefs.F90 question

• wants the physics to inform the host model on what to allocate (size of tracer array) - but needs further information like which of those need to be advected etc.
• need to understand better what requirements we don't have already
• recap on what new cap_gen will do in terms of interstitial variables (handled by framework)
  • variables generated by one suite that may be needed by another suite, but not by the host
  • chemistry has a generator for species, needs the host model to advect it
  • how do you communicate this between physics and host model
    • need to augment the metadata to do this
    • host-model specific copying phase that copies the physics tracer arrays into the advection arrays
    • Steve and Francis are working on this, this has been discussed in the past
  • this should take care of everything Alex needs

Second NCAR-NOAA-NRL meeting

• currently planned for November 5-6, 2019
• in Boulder (last time at NCAR)?
• who should be attending?

Skip next week's meeting (May 16) since both Ligia and Dom are gone