Closed GeorgeGayno-NOAA closed 4 years ago
All programs can now be built on Hera. Will test each component using 2044db3e166.
Some tests were also run on Theia for comparison. The Theia tests used 'develop' at 947145c.
First, the nemsio utilities were tested using the nemsio_utils.sh
script in /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/nemsio_utils
. The script invokes nemsio_read
, nemsio_chgdate
and nemsio_get
to process/read an FV3GFS nemsio surface file. The output from the test was correct and is stored in the log file (for interested users to look at).
Next, chgres_cube
was tested by running the regression tests in the feature/chgres_reg_test
branch (f5b2946) . The regression test driver script (driver.theia.sh
) was modified to run on Hera. All six tests ran to completion with no errors. Output was compared to the baseline regression test data on Theia (/scratch4/NCEPDEV/da/noscrub/George.Gayno/reg_tests/chgres_cube/baseline_data
). Files created on Hera were not bit identical to those created on Theia. However, differences were very small (floating point differences). So the test is declared a success. The log file from the test is stored in: /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/chgres_cube
Next, snow2mdl
was run to create a T1534 snow analysis (grib 1 format). The script and input data are here: /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/snow2mdl
The same test was also run on Theia. The grib 1 files created on Hera and Theia were bit identical and looked correct when viewed in Grads.
Next, ice_blend
was run to create a 5-minute blended sea ice analysis. The script and input data are here: /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/ice_blend
. The test was also run on Theia. The program ran to completion on both machines and the ice analyses (grib 1) were bit identical.
Next, global_cycle
was tested using the script and data in /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/cycle
. The script simulates the C768 surface update for the 2019073000 FV3GFS cycle. The test was run on Theia and Hera. The updated surface restart tile files were compared using the nccmp
utility and were identical.
Next, the grid generation procedure was tested using driver_scripts/driver_grid.hera.sh
. The following grids were created: C96 uniform, stretched, nest and regional. The output 'grid', 'oro' and 'sfc_climo' files were spot checked using 'ncview'. The C96 uniform files were also created on Theia. The Hera and Theia files had identical values according to the nccmp
utility.
Kate has successfully tested 63ee8bd within the global workflow. Her tests exercised global_cycle
, global_chgres
and the nemsio utilities.
The remaining components were tested earlier using the branch at 2044db3. Several updates were made to the branch since then. To be safe, rerun the tests on these components.
The set of chgres_cube
regression tests were run using 63ee8bd on Hera. All tests passed.
The grid generation process was successfully tested at 2044db3. It was retested using 63ee8bd for a uniform C96 grid. The output orography, grid and surface climo files were identical according to the nccmp
utility.
The snow2mdl
program was successfully tested at 2044db3 (on Hera) using the script and data in /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/snow2mdl
. It was rerun using 63ee8bd. The output T1534 grib 1 snow files were bit identical.
The ice_blend
program was successfully tested at 2044db3 (on Hera) using the script and data in /scratch1/NCEPDEV/da/George.Gayno/ufs_utils.git/hera_port/ice_blend
. It was rerun using 63ee8bd. The output grib 1 ice files were bit identical.
Ready to be merged.
Work complete. Closing issue.
Hera is the replacement for Theia. As such, this repo must be ported and tested on that platform.