re-order elements and vertices by weiyuan-jiang · Pull Request #4234 · GEOS-ESM/MAPL

weiyuan-jiang · 2025-12-09T18:09:55Z

Types of change(s)

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)
Trivial change (affects only documentation or cleanup)
Refactor (no functional changes, no api changes)

Checklist

Tested this change with a run of GEOSgcm
Ran the Unit Tests (make tests)

Description

for discussion

Related Issue

github-actions · 2025-12-09T18:10:07Z

This PR is being prevented from merging because you have added one of our blocking labels: 🚫 Contingent - DNA. You'll need to remove it before this PR can be merged.

github-actions · 2025-12-09T18:10:08Z

This PR is being prevented from merging because you have not added one of our required labels: 0 diff, 0 diff trivial, 😲 Non 0 Diff, 🔧 Github Actions. Please add one so that the PR can be merged.

github-actions · 2025-12-09T18:10:08Z

This PR is being prevented from merging because you have not added one of our required labels: 0 diff, 0 diff trivial, 😲 Non 0 Diff, 🔧 Github Actions. Please add one so that the PR can be merged.

mathomp4 · 2025-12-09T19:47:40Z

Huh. Lots of CI failures. ninja seems unhappy with:

ninja: build stopped: multiple rules generate mesh2/sf_direction.mod.

As for ifx, not sure. Seems like our spurious random failure.

Also from GNU:

Error copying Fortran module "mesh2/sf_direction.mod".  Tried "mesh2/SF_DIRECTION.mod" and "mesh2/sf_direction.mod".
make[2]: *** [mesh2/CMakeFiles/to_mesh2.x.dir/depend.make:7: mesh2/CMakeFiles/to_mesh2.x.dir/sf_direction.mod.stamp] Error 1
make[1]: *** [CMakeFiles/Makefile2:9260: mesh2/CMakeFiles/to_mesh2.x.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 89%] Building Fortran object mesh2/CMakeFiles/xgrid.x.dir/Direction.F90.o
Error copying Fortran module "mesh2/sf_direction.mod".  Tried "mesh2/SF_DIRECTION.mod" and "mesh2/sf_direction.mod".
make[2]: *** [mesh2/CMakeFiles/xgrid2.x.dir/depend.make:7: mesh2/CMakeFiles/xgrid2.x.dir/sf_direction.mod.stamp] Error 1
make[1]: *** [CMakeFiles/Makefile2:9356: mesh2/CMakeFiles/xgrid2.x.dir/all] Error 2

weiyuan-jiang · 2025-12-16T17:44:44Z

There are a few make_esmf_mesh subroutines. My tests show only the one works. @tclune, We should clean those up

tclune · 2025-12-16T18:20:05Z

There are a few make_esmf_mesh subroutines. My tests show only the one works. @tclune, We should clean those up

Please do. The extra copies were all just attempts to work around an unknown issue.

weiyuan-jiang · 2026-01-12T16:47:48Z

Notes:
With the latest update, time mpirun -np 6 ./xgrid2.x
multi_connections: 49 s
ordered multi_connections: 38 s

tri_conn : 83 S
ordered tri_conn 75 s

quad_conn gets error when creating handler from xgrid to latlon grid:

pe=00002 FAIL at line=00085 xgrid2.F90 <status=513>

mathomp4 · 2026-01-22T13:37:51Z

Notes: With the latest update, time mpirun -np 6 ./xgrid2.x multi_connections: 49 s ordered multi_connections: 38 s

tri_conn : 83 S ordered tri_conn 75 s

quad_conn gets error when creating handler from xgrid to latlon grid:

pe=00002 FAIL at line=00085 xgrid2.F90 <status=513>

@weiyuan-jiang Just for my own benefit, are these good? Bad? That is, should "ordered tri_conn" be the best? quad_conn?

weiyuan-jiang · 2026-01-22T15:38:58Z

Notes: With the latest update, time mpirun -np 6 ./xgrid2.x multi_connections: 49 s ordered multi_connections: 38 s
tri_conn : 83 S ordered tri_conn 75 s
quad_conn gets error when creating handler from xgrid to latlon grid:
pe=00002 FAIL at line=00085 xgrid2.F90 <status=513>

@weiyuan-jiang Just for my own benefit, are these good? Bad? That is, should "ordered tri_conn" be the best? quad_conn?

After Bob made the change, the tri and quad are not necessary any more. Every thing will be the original version except we order the elements and vertices.

weiyuan-jiang · 2026-01-23T14:28:22Z

For the record, here is the summary of xgrid.F90 print out ( mesh min resolution C1440, destination C720)

	random (s) 96 pe	Ordered (s) 96 pe	ordered 216 pe	ordered 864 pe
Creating xgrid (mesh, agrid)	6.3	4.7	2.15	0.59
Creating handle(Mesh 2 xgrid	6.0	2.1	1.4	1.14
Mesh 2 xgrid( regrid)	0.001	0.008	0.0006	0.0006
Creating handle(xgrid 2 agrid)	6.6	2.4	1.67	0.78
xgrid 2 mesh(regrid)	0.002	0.0009	0.0005	0.00016
Creating Handle(mesh 2 agrid)	7.3	3.67	1.97	1.4
Mesh 2 agrid(regrid)	0.002	0.0016	0.0007	0.002

tclune · 2026-01-23T14:51:05Z

@weiyuan-jiang Please use triple back-ticks and spacing to make the table readable here.

mathomp4 · 2026-01-26T18:20:42Z

@weiyuan-jiang Please use triple back-ticks and spacing to make the table readable here.

I converted it into a Github Markdown Table. I forget those are available!

weiyuan-jiang · 2026-01-26T18:23:41Z

@weiyuan-jiang Please use triple back-ticks and spacing to make the table readable here.

I converted it into a Github Markdown Table. I forget those are available!

That is nicer !

weiyuan-jiang · 2026-01-29T01:36:06Z

Test for mesh minresolution c2880, destination c1440)

	96(r)	96	216	864	2400	5400
Creating xgrid (mesh, agrid)	31.3	22.5	10.	2.7	1.47	1.55
Creating handle(Mesh 2 xgrid)	34.4	10.7	7.7	7.0	7.1	12.9
Mesh 2 xgrid(regrid)	0.006	0.0037	0.0015	0.0008	0.003	0.0037
Creating handle(xgrid 2 agrid)	11.6	11.6	8.6	7.3	3.16	2.4
xgrid 2 agrid(regrid)	0.0037	0.0037	0.0018	0.0006	0.00037	0.00025
Creating Handle(mesh 2 agrid)	44.0	19.1	11.2	7.6	8.26	15.4
Mesh 2 agrid(regrid)	0.004	0.0024	0.0014	0.002	0.0021	0.0027

This commit implements MPI-parallel domain decomposition for LocStream geometry types in MAPL, enabling efficient distributed processing of location-based data (e.g., station observations, satellite swaths). Key changes: - Added LocStreamDecomposition class to manage 1D point distribution across MPI processes using mapl_GetPartition for even load balancing - Integrated decomposition into LocStreamGeomSpec to store decomposition metadata alongside geometry specifications - Updated LocStreamGeomFactory to apply decompositions when creating distributed LocStreams, extracting local coordinate subsets (i_0:i_1) based on each PE's assigned point range - Comprehensive test suite with 6 tests covering: * Decomposition constructors and equality * Local index calculation across multiple PE counts * Actual LocStream creation with proper distribution on 1, 2, and 4 PETs * Verification of coordinate value cor * Verification of coordinate value cor * Verification of coordinate value cor * Verificosi * Verification of coordinate value cor * Verification of coordinaubsystem. Tested with NAG 7.2.41 compiler - all tests passing.

tclune

At some point we will ned to find a proper home and really clean this up.

The base branch was changed.

…ition

…iang/task2_reorder

tclune and others added 3 commits September 8, 2025 08:57

recreated reproducer for ESMF problem.

50aa8dd

Oops - did not commit files.

ba0911a

re-order elements and vertices

c03a1d8

weiyuan-jiang requested a review from a team as a code owner December 9, 2025 18:09

weiyuan-jiang added the 🚫 Contingent - DNA Do Not Approve (DNA). These changes are contingent on other PRs label Dec 9, 2025

bug fixed and re-ordered

7d25f07

weiyuan-jiang added 3 commits January 7, 2026 08:57

remove wrong comments

ea862b7

clean up. remove some subroutines

1c80e30

added ordered multi connections outout and four connections

6817853

weiyuan-jiang added 2 commits January 14, 2026 15:43

error exposed

eedce4a

refactoring...

851792b

tclune previously approved these changes Feb 17, 2026

View reviewed changes

weiyuan-jiang mentioned this pull request Feb 18, 2026

more work for mesh generation code and benchmark #4419

Open

weiyuan-jiang added 2 commits February 23, 2026 19:45

add command line arguments

384be42

Merge branch 'release/MAPL-v3' into feature/wjiang/task2_reorder

7a6a9c3

mv mesh generating codes to geom/Mesh/utils

29fc94a

weiyuan-jiang changed the base branch from feature/mesh-take-2 to release/MAPL-v3 March 3, 2026 17:23

weiyuan-jiang added 3 commits March 3, 2026 12:42

Merge branch 'release/MAPL-v3' into feature/locstream-domain-decompos…

7261ddc

…ition

Merge branch 'feature/locstream-domain-decomposition' into feature/wj…

ad850b1

…iang/task2_reorder

initial check in Mesh ( copy from locstream)

e53299d

Conversation

weiyuan-jiang commented Dec 9, 2025

Types of change(s)

Checklist

Description

Related Issue

Uh oh!

github-actions bot commented Dec 9, 2025

Uh oh!

github-actions bot commented Dec 9, 2025

Uh oh!

github-actions bot commented Dec 9, 2025

Uh oh!

mathomp4 commented Dec 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

weiyuan-jiang commented Dec 16, 2025

Uh oh!

tclune commented Dec 16, 2025

Uh oh!

weiyuan-jiang commented Jan 12, 2026

Uh oh!

mathomp4 commented Jan 22, 2026

Uh oh!

weiyuan-jiang commented Jan 22, 2026

Uh oh!

weiyuan-jiang commented Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tclune commented Jan 23, 2026

Uh oh!

mathomp4 commented Jan 26, 2026

Uh oh!

weiyuan-jiang commented Jan 26, 2026

Uh oh!

weiyuan-jiang commented Jan 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tclune left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

mathomp4 commented Dec 9, 2025 •

edited

Loading

weiyuan-jiang commented Jan 23, 2026 •

edited

Loading

weiyuan-jiang commented Jan 29, 2026 •

edited

Loading