Add tracer advection to standalone driver [cycle 34] by DropD · Pull Request #1039 · C2SM/icon4py

DropD · 2026-02-04T09:27:46Z

Ready the standalone driver for running experiments with tracer advection.

Added

add tracer advection granule to driver
add tracer fields to prognostic state
fieldop and program to calculate deepatmo_* fields used in tracer advection.
update initialization logic for existing experiments to include tracer advection
any_float type alias in common.type_alias to express type hints where the type could be either float type allowed for vpfloat.

Improved

include tracer advection related modules in type checking and fix related problems
updated type hints around how the backend is passed around to make sure the backend is "customized" as soon as possible and then passed onwards, rather than in many individual places

Tests

The tracer advection granule remains independently tested
Passing existing tests confirms that the presence of the granule in the driver does not impact experiments without tracer advection
No data test actually running tracer advection in the driver exists yet (this is intended to be added separately, because it is a nontrivial task to itself).

nfarabullini

overall you're on the right track

model/standalone_driver/src/icon4py/model/standalone_driver/driver_utils.py

model/common/src/icon4py/model/common/metrics/compute_advection_metrics.py

DropD · 2026-02-12T14:55:24Z

cscs-ci run default

DropD · 2026-02-13T09:24:51Z

cscs-ci run

DropD · 2026-02-13T09:25:49Z

cscs-ci run

DropD · 2026-02-13T09:26:23Z

cscs-ci run defaul

DropD · 2026-02-13T09:28:35Z

cscs-ci run default

nfarabullini

a few comments

model/atmosphere/advection/src/icon4py/model/atmosphere/advection/advection_states.py

model/common/src/icon4py/model/common/metrics/compute_advection_metrics.py

model/common/src/icon4py/model/common/metrics/metrics_factory.py

model/standalone_driver/src/icon4py/model/standalone_driver/testcases/initial_condition.py

DropD · 2026-02-16T12:54:30Z

cscs-ci run default

DropD · 2026-02-17T10:18:51Z

cscs-ci run default

DropD · 2026-02-23T13:56:43Z

should tracers=[] be constructed as a class?

I read this as something like tracers=NewTracersClass(ntracer=0), where NewTracersClass contains the fields. Is that what you mean?

DropD · 2026-02-25T14:21:01Z

cscs-ci run default

DropD · 2026-02-25T15:28:27Z

cscs-ci run default

DropD · 2026-02-25T15:28:56Z

cscs-ci run distributed

nfarabullini

some last few comments

nfarabullini · 2026-02-26T07:28:39Z

model/atmosphere/advection/src/icon4py/model/atmosphere/advection/advection.py

+            self._grid,
+            dims.CellDim,
+            dims.KDim,
+            allocator=model_backends.get_allocator(self._backend),


is it not enough to just specify self._backend without the get_allocator wrapper?

and same question for all of the other ones below

No, because self._backend.allocator can be None, which leads to either type mismatches or a lot of checks for is not None or calls to get_allocator down the line.
It's a bit like the customize_backend, if we do it early on, we have to deal with fewer cases later.

nfarabullini · 2026-02-26T07:32:45Z

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

    cell_geometry = grid_savepoint.construct_cell_geometry()
+    assert (
+        cell_geometry.area is not None
+    ), "Broken assumption: this test assumes it's running from a save point containing a 'cell_area' field."


Suggested change

), "Broken assumption: this test assumes it's running from a save point containing a 'cell_area' field."

), "Broken assumption: this test assumes it's running from a savepoint containing a 'cell_area' field."

nfarabullini · 2026-02-26T07:34:07Z

model/standalone_driver/src/icon4py/model/standalone_driver/driver_utils.py

+            ),
+        ),
+        least_squares_state=advection_states.AdvectionLeastSquaresState(
+            # TODO(ricoh): [c34] check integration with #1065


do you still need this TODO?

thanks, I will remove it.

edopao · 2026-02-26T08:49:15Z

Just a note. The driver tests are not run with dace in the default pipeline. Please run the dace CI pipeline when you are close to merging this PR.

DropD · 2026-02-26T13:19:16Z

cscs-ci run default

DropD · 2026-02-26T13:24:52Z

cscs-ci run default

DropD · 2026-02-26T13:25:10Z

cscs-ci run distributed

DropD · 2026-02-26T13:25:21Z

cscs-ci run dace

DropD · 2026-02-26T15:38:18Z

cscs-ci run default

DropD · 2026-02-26T15:38:31Z

cscs-ci run distributed

DropD · 2026-02-26T15:38:37Z

cscs-ci run dace

github-actions · 2026-02-27T09:20:30Z

Mandatory Tests

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

cscs-ci run default
cscs-ci run distributed

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.

DropD · 2026-02-27T09:26:36Z

cscs-ci run default

DropD · 2026-02-27T09:26:45Z

cscs-ci run distributed

DropD added 2 commits February 2, 2026 10:32

[wip] start initializing tracer advection granule

b3bba30

[wip] work on advection granule init

2108eda

DropD requested review from jcanton and nfarabullini February 4, 2026 09:27

nfarabullini reviewed Feb 4, 2026

View reviewed changes

model/standalone_driver/src/icon4py/model/standalone_driver/driver_utils.py Show resolved Hide resolved

DropD commented Feb 4, 2026

View reviewed changes

model/standalone_driver/src/icon4py/model/standalone_driver/driver_utils.py Show resolved Hide resolved

jcanton reviewed Feb 4, 2026

View reviewed changes

model/standalone_driver/src/icon4py/model/standalone_driver/driver_utils.py Show resolved Hide resolved

add metric fields required for tracer advection

062d49f

jcanton reviewed Feb 5, 2026

View reviewed changes

model/common/src/icon4py/model/common/metrics/compute_advection_metrics.py Show resolved Hide resolved

DropD added 2 commits February 12, 2026 14:58

[wip] add advection loop in driver and related changes

77ffcab

fix missing future annotations

c8809e4

Merge branch 'main' into ricoh-c34-tracer-advection

b0ea1f8

nfarabullini reviewed Feb 13, 2026

View reviewed changes

ensure test_standalone_driver fails due to embedded remap

a6ed178

DropD added 9 commits February 17, 2026 12:57

fix tracer advection init (pass test_standalone_driver with JW)

f509aba

Merge branch 'main' into ricoh-c34-tracer-advection

0438f89

typecheck standalone driver & fix revealed errors

414eaad

[wip] type check advection.py

7f91a96

type check advection_horizontal.py

2ec8f48

type check advection_vertical.py

18770ac

type check advection_lsq_coeffs.py

8ffc8b6

typecheck advection_states.py

9ab9098

type check advection stencils

44778a6

DropD added 5 commits February 24, 2026 13:46

stylistic fixes from review

23a6a43

push backend customiztation out of advection classes

52c04fd

rename vpfloat_t -> anyfloat

79bdaa6

turn compute_advection_deepatmo_fields into a ProgramFieldProvider

e9c9610

fix deepatmo field provider

b618752

DropD added 4 commits February 25, 2026 15:53

Merge branch 'main' into ricoh-c34-tracer-advection

6e1b3cb

reference known gt4py issues/prs for the related type ignores

d1dd77f

improve standalone driver config documentation

44c0575

remove need to explicitly initialize tracer if ntracer=0

dd59b00

nfarabullini approved these changes Feb 26, 2026

View reviewed changes

remove outdated todo comment and fix an error message

3280d19

Merge branch 'main' into ricoh-c34-tracer-advection

a6e2f39

DropD added 2 commits February 26, 2026 15:04

Merge branch 'main' into ricoh-c34-tracer-advection

1e9a914

Merge branch 'main' into ricoh-c34-tracer-advection

f5f106d

Merge branch 'main' into ricoh-c34-tracer-advection

2d7676b

	), "Broken assumption: this test assumes it's running from a save point containing a 'cell_area' field."
	), "Broken assumption: this test assumes it's running from a savepoint containing a 'cell_area' field."

Conversation

DropD commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Added

Improved

Tests

Uh oh!

nfarabullini left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

DropD commented Feb 12, 2026

Uh oh!

DropD commented Feb 13, 2026

Uh oh!

DropD commented Feb 13, 2026

Uh oh!

DropD commented Feb 13, 2026

Uh oh!

DropD commented Feb 13, 2026

Uh oh!

nfarabullini left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

DropD commented Feb 16, 2026

Uh oh!

DropD commented Feb 17, 2026

Uh oh!

DropD commented Feb 23, 2026

Uh oh!

DropD commented Feb 25, 2026

Uh oh!

DropD commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

DropD commented Feb 25, 2026

Uh oh!

nfarabullini left a comment

Choose a reason for hiding this comment

Uh oh!

nfarabullini Feb 26, 2026

Choose a reason for hiding this comment

Uh oh!

nfarabullini Feb 26, 2026

Choose a reason for hiding this comment

Uh oh!

DropD Feb 26, 2026

Choose a reason for hiding this comment

Uh oh!

nfarabullini Feb 26, 2026

Choose a reason for hiding this comment

Uh oh!

nfarabullini Feb 26, 2026

Choose a reason for hiding this comment

Uh oh!

DropD Feb 26, 2026

Choose a reason for hiding this comment

Uh oh!

edopao commented Feb 26, 2026

Uh oh!

DropD commented Feb 26, 2026

Uh oh!

DropD commented Feb 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

DropD commented Feb 26, 2026

Uh oh!

DropD commented Feb 26, 2026

Uh oh!

DropD commented Feb 26, 2026

Uh oh!

DropD commented Feb 26, 2026

Uh oh!

DropD commented Feb 4, 2026 •

edited

Loading

DropD commented Feb 25, 2026 •

edited

Loading

DropD commented Feb 26, 2026 •

edited

Loading