Update from Will, Christiane, Brian
latest tag of the FVcubed dycore, last check-in
fvcubed_cam3_6_34_tags/fvcubed21_cam3_6_34
issues in the current FVcubed dycore (parallelization, performance)
- Extensive problems on Jaguar (Art), still many on bluefire
- Latest GEOS FVcubed release may solve some problems.
current workaround/fix: no 4th order divergence damping with potential for noisy simulations
- Noise can be compensated for with higher 2nd order DD
- Upgrade to FVcubed may solve this
upgrade of the dycore to the newest NASA version, pathway for future upgrades
- Brian now can compile FVcubed on Bluefire and Dublin (PGI)
- Lahey still an issue
- Will has started modifying CAM-GFDL interface
FV_StateMod to accommodate new release. 1-2 days of work
status of baroclinic tests and built-in test suite
- Test cases 1-2 run, with tracers 1234. Results available
- Test case 4 seems to be running
- Test 5 develops a wave-4 instability, run dies after 0.5 days. So far tried different divergence damping configurations, is time step unstable? Test ran fine during 2008 colloquium. Further investigations needed
- Test cases 3, 6: problems with specification of the hybrid coefficients for the 60 and 20 levels, Will investigates
status of the ideal physics tests (Held-Suarez)
- 400 day simulation available. Results look good.
status of the aqua-planet tests
- Runs for 5 days, but cubed sphere artifacts appear.
- Dry air mass adjustment runs. Some scientific questions remain.
offline interpolation of the CAM history files with interpolation package GECoRe: status and current issues
- It is running now for both history and restart NetCDF files!
- Assumes equi-angular grid - not the case in FVCUBED; initialization needs to be tweaked (Paul and Christtiane)
- Document on WIKI page (Will)
initial condition netcdf files and use of the interpic_new tool
- This all works as advertized.
- Document on WIKI page (Will)
science issues with the dry air adjustment routine in the dycore
- Discuss with S.J.
- Mark Taylor has worked on this - contact
Open software engineering issues: (Will, Brian)
restart
- Bug found in mapping
- Now runs, but there is still a subtle bug
- Something in fv_dynamics is not being set directly
upgrade to new version: time line and complexity
- More information by the end of Friday, May 29.
analytic (online) initialization of the staggered edge-tangent velocities
- After discussion with
use of routine outfld
ESMF clock
will the namelist option INIC work? If yes, with staggered velocities?
Performance analyses (Art, Pat)
- selection of test case for performance studies: test 2 (baroclinic wave)
- FVcubed model configurations (resolution, number of tracers)
- compilation and run time issues on Bluefire and Jaguar
- comparisons to CAM FV lat-lon?
Documentation (all)
Plans for CCSM workshop (Christiane, all)
- poster: selection of test cases and simple performance plots
- round-table meeting at the CCSM workshop, common area, on Wednesday 4-5pm (MDT).
Plans for the next few weeks: (all, advise by HOMME team)
- aqua-planet runs
- analysis of aqua-planet results, standard NCL scripts?
- coupling to land, ice and ocean, are full CCSM runs possible? Brian: grid weight files needed. Contact Mark.
- AMIP-style runs with data ocean; Contact Mark.
- on the wishlist: settings that allow a single-level shallow water setup
- future maintenance
- Next Telecon: Friday, June 12, 2009.