Use integer flags to denote species units instead of strings #1796

yantosca · 2023-05-15T19:17:02Z

Name and Institution (Required)

Name: Bob Yantosca
Institution: Harvard + GCST

Confirm you have reviewed the following documentation

Contributing guidelines

Describe the update

This update will improve how unit conversions are done in GEOS-Chem:

Use integer unit flags instead of strings. This will improve runtime as integer comparisons are much faster than string comparisons.

Expected changes

This should be a zero-diff update, as we do not plan to introduce any differences. However, it may be that this ends up introducing differences at the level of numerical noise.

Reference(s)

N/A

Related Github Issue(s)

Issues to be resolved in the carbon simulation #1984

These updates are the first batch of modifications for converting species unit conversions to per-species. This will involve keeping track of the units of each species rather than assuming that all species have the same unit. This will allow us only to convert species that are absolutely needed (e.g. only convert wetdep species before wetdep, etc.) Headers/state_chm_mod.F90 - State_Chm%Spc_Units is now an integer instead of a character string GeosUtil/unitconv_mod.F90 - Add public integer unit parameter flags - Add private UNIT_STR array to convert integer unit flags to a string - inUnit, outUnit, and origUnit variables are now all integers - IS_Adjoint is renamed to isAdjoint - Now set State_Chm%Spc_Units with the appropriate integer flags - Trimmed trailing whitespace NOTE: This update will cause type mismatch errors until we fix the type of the outUnit and origUnit variables in routines that call Convert_Spc_Units. These will be fixed in upcoming commits. Signed-off-by: Bob Yantosca <[email protected]>

GeosUtil/unitconv_mod.F90 - UNIT_STR needs to be UNIT_STR(7) and PUBLIC - Corrected typos - Fixed incorrect unit conversion flag names - Fixed issue that prevented kg/kg -> mol/mol dry conversion from beinmg done (i.e. using the wrong flag in the CASE statement) - Fixed incorrect error message variable names - Fixed instances where origUnit was character, not integer - Use UNIT_STR to convert the integer value to a string Signed-off-by: Bob Yantosca <[email protected]>

In all routines where Convert_Spc_Units is called: - Pass an integer flag instead of a string via outUnit - Use an integer variable to store origUnit - Use the entire unitconv_mod module (i.e. "USE UnitConv_Mod) - Use UNIT_STR parameter array to convert integer to string values NOTE: These updates cause differences w/r/t the prior code. We will identify and fix these in subsequent updates. Signed-off-by: Bob Yantosca <[email protected]>

Interfaces/GCHP/gchp_chunk_mod.F90 - Bug fix: the #elif block should set State_Chm%Spc_Units to MOLES_SPECIES_PER_MOLES_DRY_AIR. Signed-off-by: Bob Yantosca <[email protected]>

GeosCore/aero_drydep_mod.F90 GeosCore/dust_mod.F90 GeosCore/tomas_mod.F90 GeosCore/wetscav_mod.F90 - Add "USE UnitConv_Mod" where it was missing - Take State_Chm%Spc_Units out of TRIM statements as it's an integer now - Use "UNIT_STR(State_Chm%Spc_Units)" in error messages, which will convert the integer unit flag to a string. NOTE: These modifications are just to preserve the status quo before refactoring (so that we can test for identicality). We will eventually be removing State_Chm%Spc_Units as we move to per-species unit variables. Signed-off-by: Bob Yantosca <[email protected]>

This brings the GEOS-Chem 14.2.1 code into the development branch feature/unitconv-per-species-category so that further development can take place. Signed-off-by: Bob Yantosca <[email protected]>

GeosCore/mercury_mod.F90 - Add missing origUnit variable GeosCore/set_boundary_conditions_mod.F90 - Add "USE UnitConv_Mod" statement - Changed LOC -> thisLoc - Changed origUnit variable from CHARACTER to INTEGER - Use KG_SPECIES_PER_KG_DRY_AIR instead of "kg/kg dry" - Use UNIT_STR(State_Chm%SPC_UNITS) for a string representation of units - Now save error msg in errMsg variable GeosCore/tracer_mod.F90 - Updated call to Convert_Spc_Units, use MOLES_SPECIES_PER_MOLES_DRY_AIR as the units parameter - Change origUnit from CHARACTER to INTEGER Signed-off-by: Bob Yantosca <[email protected]>

yantosca · 2023-10-12T17:57:49Z

Integration tests (14.2.2_r1) now running.

yantosca · 2023-10-12T18:52:02Z

All GCHP compilation tests failed. GEOS-Chem Classic tests are proceeding.

Compiliation tests:
------------------------------------------------------------------------------
GCHP................................................Configure & Build.....FAIL
GCHP w/ carbon gases (as a KPP mechanism)...........Configure & Build.....FAIL
GCHP with RRTMG.....................................Configure & Build.....FAIL

Interfaces/GCHP/gchp_chunk_mod.F90 - Change origUnit variable from CHARACTER to INTEGER since we now use integer parameters for species units Signed-off-by: Bob Yantosca <[email protected]>

yantosca · 2023-10-12T18:58:09Z

Fixed an issue in Interfaces/GCHP/gchp_chunk_mod.F90, where the origUnit variable was still CHARACTER. This now needs to be integer. Rerunning integration tests.

yantosca · 2023-10-12T19:22:55Z

All GEOS-Chem Classic integration tests passed (except for TOMAS, which will be fixed soon):

==============================================================================
GEOS-Chem Classic: Execution Test Results

GCClassic #ad68c48 GEOS-Chem 14.2.1 and HEMCO 3.7.1 release
GEOS-Chem #eb803ca49 Fix several issues with integer unit parameters in newer code
HEMCO     #00aaf65 Update HEMCO 3.7.1 release date in CHANGELOG.md

Using 24 OpenMP threads
Number of execution tests: 27

Submitted as SLURM job: 4591914
=============================================================================
 
Summary of test results:
------------------------------------------------------------------------------
Execution tests passed: 25
Execution tests failed: 2
Execution tests not yet completed: 0

yantosca · 2023-10-12T19:25:16Z

Also all GEOS-Chem Classic integration tests were zero-diff w/r/t the prior integration test (for PR #1977), except for the usual suspects:

Checking gc_4x5_merra2_fullchem_APM
   -> 2 differences found in OutputDir
      * GCC_r30//rundirs/gc_4x5_merra2_fullchem_APM/OutputDir/GEOSChem.Metrics.20190701_0000z.nc4 
        GCC_r1/rundirs/gc_4x5_merra2_fullchem_APM/OutputDir/GEOSChem.Metrics.20190701_0000z.nc4 
      * GCC_r30//rundirs/gc_4x5_merra2_fullchem_APM/OutputDir/GEOSChem.SpeciesConc.20190701_0000z.nc4 
        GCC_r1/rundirs/gc_4x5_merra2_fullchem_APM/OutputDir/GEOSChem.SpeciesConc.20190701_0000z.nc4 
   -> 1 difference found in Restarts
      * GCC_r30//rundirs/gc_4x5_merra2_fullchem_APM/Restarts/GEOSChem.Restart.20190701_0100z.nc4 
        GCC_r1/rundirs/gc_4x5_merra2_fullchem_APM/Restarts/GEOSChem.Restart.20190701_0100z.nc4 


Checking gc_4x5_merra2_fullchem_RRTMG
   -> 1 difference found in OutputDir
      * GCC_r30//rundirs/gc_4x5_merra2_fullchem_RRTMG/OutputDir/GEOSChem.RRTMG.20190701_0000z.nc4 
        GCC_r1/rundirs/gc_4x5_merra2_fullchem_RRTMG/OutputDir/GEOSChem.RRTMG.20190701_0000z.nc4 
   -> No differences in Restarts

These simulations are known to have parallelization issues.

msulprizio

This PR should be good to merge after removing the obsolete MMR_Compute_Flux routine from emissions_mod.F90.

GeosCore/emissions_mod.F90

GeosCore/emissions_mod.F90 - Removed the routine MMR_Compute_Flux, which snuck back in with a Git merge when we brought the feature/unitconv-per-species-category up to 14.2.1. Signed-off-by: Bob Yantosca <[email protected]>

yantosca · 2023-10-13T14:14:17Z

Good catch @msulprizio. MMR_Compute_Flux is now removed again in commit 189cc1c.

lizziel

Looks good to me! In the future could you put formatting changes in a different PR? It was hard to hone in on the changes with so many indentation etc changes.

yantosca · 2023-10-30T21:15:07Z

Looks good to me! In the future could you put formatting changes in a different PR? It was hard to hone in on the changes with so many indentation etc changes.

Will do @lizziel.

yantosca · 2023-10-30T21:15:10Z

Looks good to me! In the future could you put formatting changes in a different PR? It was hard to hone in on the changes with so many indentation etc changes.

Will do @lizziel.

yantosca · 2023-10-31T13:51:18Z

After merging locally, all integration tests passed (except TOMAS) and were zero-diff w/r/t 14.2.2. (except APM and RRTMG diagnostic output).

This is the official release of GEOS-Chem (science codebase) 14.2.3. Updates include: - Use integer flags to denote species units instead of strings (PR #1796) - Prevent POAEMISS variable from being assigned if not allocated (PR #1987) - Add an --no-bootstrap option to integration tests (PR #1990) - Fixed incorrect comments on static H2O option (PR #2013) - Improve missing CH4 emissions error message (PR #2039) - Lower GCHP timestep threshold if > C180 rather than >= C180 (PR #2040) - Add fixes to HEMCO_Config.rc for CH4, carbon simulations (PR #2042) - Fix typos written to GCHP integration test log files (PR #2043) - Update GCClassic rundir scripts to read restat file paths from download_data.yml (PR #2050) Signed-off-by: Bob Yantosca <[email protected]>

yantosca self-assigned this May 15, 2023

yantosca added 4 commits May 15, 2023 17:40

Bug fix: Set State_Chm%Spc_Units to "mol/mol dry" in gchp_chunk_mod.F90

7c6a0f6

Interfaces/GCHP/gchp_chunk_mod.F90 - Bug fix: the #elif block should set State_Chm%Spc_Units to MOLES_SPECIES_PER_MOLES_DRY_AIR. Signed-off-by: Bob Yantosca <[email protected]>

msulprizio mentioned this pull request Jun 5, 2023

Update transport tracer simulation for consistency with GMAO's tracer gridded component #1816

Merged

1 task

yantosca added 2 commits October 11, 2023 15:40

Merge GEOS-Chem 14.2.1 into feature/unitconv-per-species-category

a73132a

This brings the GEOS-Chem 14.2.1 code into the development branch feature/unitconv-per-species-category so that further development can take place. Signed-off-by: Bob Yantosca <[email protected]>

yantosca changed the title ~~Optimize unit conversions and allow unit conversions for certain categories of species (instead of all species)~~ Use integer flags to denote species units instead of strings Oct 12, 2023

yantosca requested a review from msulprizio October 12, 2023 17:55

yantosca changed the base branch from main to dev/no-diff-to-benchmark October 12, 2023 17:55

yantosca marked this pull request as ready for review October 12, 2023 17:57

In gchp_chunk_mod.F90, origUnit must now be INTEGER

16e909f

Interfaces/GCHP/gchp_chunk_mod.F90 - Change origUnit variable from CHARACTER to INTEGER since we now use integer parameters for species units Signed-off-by: Bob Yantosca <[email protected]>

yantosca modified the milestones: 14.2.2, 14.2.1 Oct 12, 2023

msulprizio requested changes Oct 13, 2023

View reviewed changes

GeosCore/emissions_mod.F90 Outdated Show resolved Hide resolved

Remove MMR_Compute_Flux from emissions_mod (snuck in w/ merge)

189cc1c

GeosCore/emissions_mod.F90 - Removed the routine MMR_Compute_Flux, which snuck back in with a Git merge when we brought the feature/unitconv-per-species-category up to 14.2.1. Signed-off-by: Bob Yantosca <[email protected]>

yantosca requested a review from msulprizio October 13, 2023 14:14

lizziel self-requested a review October 13, 2023 14:50

lizziel approved these changes Oct 26, 2023

View reviewed changes

msulprizio approved these changes Oct 27, 2023

View reviewed changes

yantosca merged commit 5331282 into dev/no-diff-to-benchmark Oct 31, 2023

yantosca added this to the 14.2.3 milestone Nov 30, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use integer flags to denote species units instead of strings #1796

Use integer flags to denote species units instead of strings #1796

yantosca commented May 15, 2023 •

edited

Loading

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

msulprizio left a comment

yantosca commented Oct 13, 2023 •

edited

Loading

lizziel left a comment •

edited

Loading

yantosca commented Oct 30, 2023

yantosca commented Oct 30, 2023

yantosca commented Oct 31, 2023

Use integer flags to denote species units instead of strings #1796

Use integer flags to denote species units instead of strings #1796

Conversation

yantosca commented May 15, 2023 • edited Loading

Name and Institution (Required)

Confirm you have reviewed the following documentation

Describe the update

Expected changes

Reference(s)

Related Github Issue(s)

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

yantosca commented Oct 12, 2023

msulprizio left a comment

Choose a reason for hiding this comment

yantosca commented Oct 13, 2023 • edited Loading

lizziel left a comment • edited Loading

Choose a reason for hiding this comment

yantosca commented Oct 30, 2023

yantosca commented Oct 30, 2023

yantosca commented Oct 31, 2023

yantosca commented May 15, 2023 •

edited

Loading

yantosca commented Oct 13, 2023 •

edited

Loading

lizziel left a comment •

edited

Loading