Better torus support #965

jcanton · 2025-12-01T09:48:26Z

Based upon #953 #954 and #957 by @msimberg
Adds torus support to geometry, RBF and interpolation.
Also adds torus grids to all datatests

* main: Update DaCe version (#962) blueline: move vertical grid to grid_wrapper, add flat_height (#956) CI testdata moved to `cwci02` project (#960) Bump actions/checkout from 5 to 6 (#955) Test fixes for metrics fields (#949) Remove pytest-deadfixtures from test dependency (#950) Revert "Add debug info to stencil tests" Add debug info to stencil tests

@msimberg

…nto better-torus-support-metrics * origin/better-torus-support-geometry: Apply suggestion from @msimberg

…his but it doesn't work)

(r,b) are not used anymore to check whether we're dealing with torus: we have GeometryType for that, so they can be whatever value for torus and we can remove a bunch of | Nones

model/common/src/icon4py/model/common/interpolation/interpolation_fields.py

jcanton · 2025-12-11T13:43:46Z

cscs-ci run default

model/common/src/icon4py/model/common/grid/geometry.py

model/common/src/icon4py/model/common/grid/geometry_stencils.py

yiluchen1066 · 2025-12-16T14:03:48Z

model/common/src/icon4py/model/common/grid/geometry_stencils.py

+    return (
+        tangent_x,
+        tangent_y,
+        tangent_z,
+        tangent_u,
+        tangent_v,
+        normal_x,
+        normal_y,
+        normal_z,
+        normal_u,
+        normal_v,
+    )


here why we cannot just return the components on x and y. I think u and v are just the copy of them

we could but it does not make a difference in terms of memory usage to do it here (and doing it in other places makes it less readable IMO)

model/common/src/icon4py/model/common/interpolation/interpolation_fields.py

msimberg · 2025-12-17T10:45:04Z

cscs-ci run default

msimberg

I'm a bit blind to my own changes, but @jcanton improvements look very good to me.

I only have a few minor requests for changes.

model/common/src/icon4py/model/common/grid/geometry.py

model/common/src/icon4py/model/common/interpolation/rbf_interpolation.py

model/common/tests/common/grid/unit_tests/test_icon.py

model/testing/src/icon4py/model/testing/fixtures/benchmark.py

model/testing/src/icon4py/model/testing/grid_utils.py

* main: Update dace version to tag 2025_12_11 (#983) update gt4py version in standalone driver (#977) [greenline] standalone driver for jw test case (#902) Fix second call to mo_intp_rbf_rbf_vec_interpol_vertex in diffusion (#974) Bump actions/cache from 4 to 5 (#975)

jcanton · 2025-12-23T14:27:07Z

cscs-ci run default

jcanton · 2025-12-24T10:53:52Z

cscs-ci run default

model/common/src/icon4py/model/common/grid/geometry_stencils.py

OngChia · 2025-12-29T08:53:33Z

model/common/src/icon4py/model/common/grid/geometry_stencils.py

+
+    """
+    return distance_on_edges_torus(
+        vertex_x(E2C2V[2]),


Suddenly, I am thinking if is it better to define far indices in icon.py as a enum class or global constants or something else instead of putting a hardcoded number here, such as
E2C2V_NEIGHBOR_INDEX1 = 0
E2C2V_NEIGHBOR_INDEX2 = 1
E2C2V_FAR_INDEX1 = 2
E2C2V_FAR_INDEX2 = 3
(I do not remember whether gt4py allows that or not, and I put this comment because it seems that the figure in the docstring of arc_distance_of_far_edges_in_diamond seems to be inconsistent with the actual indices, leading to confusion. If we have a universal definition, then it will be clearer)

I fixed the docstsring, nice catch!
I tried some ways (constants, enum class) to move the indexes to some named variable but gt4py doesn't like it, and I cannot find any other field_operator that does this so I am guessing this is not supported, right @havogt ?

model/common/src/icon4py/model/common/grid/icon.py

model/common/src/icon4py/model/common/interpolation/interpolation_factory.py

jcanton · 2025-12-29T13:44:53Z

cscs-ci run default

jcanton · 2025-12-29T13:46:19Z

cscs-ci run default

jcanton · 2025-12-30T09:14:34Z

cscs-ci run default

model/common/tests/common/metrics/unit_tests/test_metric_fields.py

msimberg

Looks good to me, thank you @jcanton!

Some of @OngChia's comments are still useful to address in my opinion, so leaving final approval to him.

github-actions · 2026-01-05T10:39:21Z

Mandatory Tests

Please make sure you run these tests via comment before you merge!

cscs-ci run default

Optional Tests

To run benchmarks you can use:

cscs-ci run benchmark-bencher

To run tests and benchmarks with the DaCe backend you can use:

cscs-ci run dace

To run test levels ignored by the default test suite (mostly simple datatest for static fields computations) you can use:

cscs-ci run extra

For more detailed information please look at CI in the EXCLAIM universe.

jcanton · 2026-01-05T10:39:38Z

cscs-ci run default

msimberg and others added 5 commits November 19, 2025 14:39

Add torus support to geometry

f21b246

Apply suggestion from @msimberg

5b14541

Add torus support for interpolation fields

bc7e2bc

Update metrics fields tests for torus grids

65bdfc3

jcanton closed this Dec 1, 2025

jcanton reopened this Dec 1, 2025

Merge remote-tracking branch 'origin/better-torus-support-geometry' i…

4c73db2

…nto better-torus-support-metrics * origin/better-torus-support-geometry: Apply suggestion from @msimberg

jcanton changed the title ~~Better torus support metrics~~ Better torus support Dec 1, 2025

jcanton added 17 commits December 1, 2025 15:12

merge back Icosahedron and Torus Geometry classes

afaf759

put these back

1d67822

clean up and add TODOs from github

6f98c70

clean GeometryType: either TORUS or ICOSAHEDRON, not None anymore

cedf32a

add links to updstream MRs

f4173f4

geometry_type() can return None

e3eee2a

update grid_manager test for torus indexes

b36d5fa

if damping_height is above model_top, reset it

21e1219

GeometryType can never be None (the type checker would make sure of t…

e0ced8f

…his but it doesn't work)

now that GeometryType cannot be None, assert_never(s) can be removed

0469778

actually better without the Nones and TORUS(r=0,b=0)

47e95ce

(r,b) are not used anymore to check whether we're dealing with torus: we have GeometryType for that, so they can be whatever value for torus and we can remove a bunch of | Nones

fix tests for GridSubdivision != None

3bf84dd

remove two more assert_none

76678a4

add type?

b2e86c5

I think this comment is useful

8bd2207

remove assert_never

4535bad

renaming

682db97

jcanton commented Dec 8, 2025

View reviewed changes

model/common/src/icon4py/model/common/interpolation/interpolation_fields.py Show resolved Hide resolved

jcanton added 3 commits December 9, 2025 09:43

small cleanup

ee51e8a

remove TODOs: all others work with _torus, these too ok

65eef62

remove one other case

7da8935

remove actually unused vn_only after discussion with @egparedes

60ebdf9

yiluchen1066 reviewed Dec 16, 2025

View reviewed changes

msimberg requested changes Dec 17, 2025

View reviewed changes

msimberg mentioned this pull request Dec 17, 2025

Initialize geometry fields for torus grids #853

Closed

jcanton added 4 commits December 23, 2025 11:52

apply @yiluchen1066 review

ed78a24

pre-commit

0218c56

apply @msimberg review

20c2b12

jcanton added 2 commits December 24, 2025 11:47

remove unused input and hopefully fix bug showing up with cupy

a5f4c8d

put this back for consistency with icosahedron

48b5806

OngChia reviewed Dec 29, 2025

View reviewed changes

jcanton added 3 commits December 29, 2025 12:10

fix wrong docstring (thanks @OngChia) and make returns consistent

188d235

apply @OngChia review

d490191

possible second bugfix for array_ns.full

0dbbe5a

pre-commit

c0420f0

jcanton requested review from OngChia and msimberg December 29, 2025 13:49

jcanton removed the request for review from halungge December 30, 2025 10:27

OngChia reviewed Jan 5, 2026

View reviewed changes

model/common/tests/common/metrics/unit_tests/test_metric_fields.py Outdated Show resolved Hide resolved

msimberg approved these changes Jan 5, 2026

View reviewed changes

one more commit

9642322

Better torus support #965

Are you sure you want to change the base?

Better torus support #965

Conversation

jcanton commented Dec 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

jcanton commented Dec 11, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

yiluchen1066 Dec 16, 2025

Choose a reason for hiding this comment

Uh oh!

jcanton Dec 23, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

msimberg commented Dec 17, 2025

Uh oh!

msimberg left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jcanton commented Dec 23, 2025

Uh oh!

jcanton commented Dec 24, 2025

Uh oh!

Uh oh!

Uh oh!

OngChia Dec 29, 2025

Choose a reason for hiding this comment

Uh oh!

jcanton Dec 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

jcanton commented Dec 29, 2025

Uh oh!

jcanton commented Dec 29, 2025

Uh oh!

jcanton commented Dec 30, 2025

Uh oh!

Uh oh!

msimberg left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 5, 2026

Uh oh!

jcanton commented Jan 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

jcanton commented Dec 1, 2025 •

edited

Loading

jcanton Dec 29, 2025 •

edited

Loading