b4b: Move the changes to initialize the task decomposition from mpi_scan to main development#3666

Draft

ekluzek wants to merge 63 commits intoESCOMP:b4b-devfrom

ekluzek:decomp_mpi_scan_move_to_b4b

Commits on Dec 15, 2025

Start adding timers to lnd_set_decomp_and_domain_from_readmesh
ekluzek
committed
Make an internal subroutine for deallocation inside of lnd_set_decomp_and_domain_from_readmesh
ekluzek
committed
Add release of the ESMF objects in the lnd_set_decomp_and_domain_from_readmesh subroutine
ekluzek
committed
ESMF tells me that some of these objects are used later and can not be destroyed so remove the destroy for the distgrid, and the two meshes, this runs but doesn't seem to lower memory
ekluzek
committed
Turn on removing all ESMF garbage for the things deleted, and add note about leaving the distgrid around, and also delete the meshes as it seems to work with this in place
ekluzek
committed
Deallocate lndfrac_loc_input
ekluzek
committed
Add timers
ekluzek
committed
Add more timers
ekluzek
committed
Add return's after endrun calls so that the self-tests can trap expected error messaging
ekluzek
committed
Change endrun calls to use new format for file and line, and remove the subname, create new internal subroutines in decompInit_lnd for allocate, clean, and check errors, move the check errors part …
ekluzek
committed
decompInit/decompMod: initialize pointers to null, add checking that array sizes are set before allocates, initialize some decompMod values to invalid for error checking, add error checking to get_…
ekluzek
committed
Allow get_proc_bounds to exit early if only b3gg/endg will be needed, add error handling of nsegspc, don't check endCohort in get_proc_bounds and get_clump_bounds as doesn't seem to be set
ekluzek
committed
Return up the chain in decompInit after endrun, for the testing. Do setup/clean for each DecompInit test, move the decomp_mod_clean to decompMod and use it for the decompInit tests
ekluzek
committed
Move comment over for clarity
ekluzek
committed
Add a MPI_SCAN as a temporary to start testing how it works and compare to the regular operation
ekluzek
committed
Start filling other procinfo settings with mpi_scan, this passes for mpi-serial
ekluzek
committed
Remove the mpiscan bit about clumpcnt
ekluzek
committed
Add global i,j indices for each gridcell on a processor to the processor_type structure, and start adding a couple methods to help get them set
ekluzek
committed
Add more testing, test new processor_type methods, set gindex_global for mpiscan and verify it, allocate the new procinfo gi and gj indices, make sure they are set, compiles but fails at run
ekluzek
committed
Can't do this test yest, as only gridcells are set at this point and the call expects all subgrid levels to be set
ekluzek
committed
Fix setting of gi and gj indices, the serial case runs to completion now
ekluzek
committed
Add ggidx global index to the processor_type, allocate, set and deallocate for the local task, this works for the serial case
ekluzek
committed
Correct the indexing for ggidx
ekluzek
committed
Remove the prints that weren't with an abort
ekluzek
committed
Get it working for a threaded case, by adding an additional loop over clump_pproc for the setting of ggidx
ekluzek
committed
Add more testing for correctness, add some notes about what can be removed for the final version
ekluzek
committed
Get it working for the threaded case
ekluzek
committed
Comment out the old code and just use the new MPI_SCAN way, this works for serial mode
ekluzek
committed
Just do the checking over the local processor clumps and not all the global clumps
ekluzek
committed
Remove the decompInit_lnd old code commented out, and extra writes to the log that aren't needed anymore
ekluzek
committed
Initialize the gi/gj arrays to unset
ekluzek
committed
Add in an if for hist_printflds if noio from mpi_scan branch
ekluzek
committed
Bring some of the missing things from the mpi_scan branch missed in the cherry-pick, mostly comments
ekluzek
committed
Remove some of the timers we don't really need on main
ekluzek
committed
A few changes on mpi_scan that don't need to come to main, remove some things referring to testing
ekluzek
committed
Add some comments and comment line breaks for new subroutines
ekluzek
committed

Commits on Dec 18, 2025

Only deallocate lndfrac_loc_input if it's associated, and initialize it to null
ekluzek
committed

Commits on Dec 19, 2025

Merge branch 'b4b-dev' into decomp_mpi_scan_move_to_b4b
ekluzek
authored

Commits on Dec 20, 2025

Add add_simple_patch optional argument in #3672, and use it for the one failing test, so that now the PF unit tests work
ekluzek
committed

Commits on Jan 7, 2026

Turn on more options to get more fields turned on
ekluzek
committed

Commits on Jan 8, 2026

Merge branch 'b4b-dev' into decomp_mpi_scan_move_to_b4b
ekluzek
authored

Commits on Jan 15, 2026

Commits on Jan 19, 2026

Move nglob_x/nglob_y from decompInit to decompMod as module data
ekluzek
committed

Commits on Jan 20, 2026

Merge branch 'decompMod_unittest' into decomp_mpi_scan_move_to_b4b
ekluzek
committed

Commits on Jan 22, 2026

Merge remote-tracking branch 'escomp/b4b-dev' into decomp_mpi_scan_move_to_b4b
ekluzek
committed

Commits on Jan 28, 2026

Commits on Jan 29, 2026

Add a test for calc_globalxy_indices, to get this working temporarily remove pure from the calc_ routines in decompMod, and write out information, the unit test does work now
ekluzek
committed

Commits on Feb 18, 2026

Move settings of the decomp from the decomp test to the unittest share set_decomp_info subroutine and make it public for unit testing, this doesn't work as it fails in clm_ptrs_compdown, and the en…
ekluzek
committed

Commits on Mar 12, 2026