XIOS: Handle `CGField` #951

joewallwork · 2025-10-14T13:37:38Z

XIOS: Handle `CGField`

Fixes #911
Merges into #952

Task List

Defined the tests that specify a complete and functioning change (It may help to create a design specification & test specification)
Implemented the source code change that satisfies the tests
Documented the feature by providing worked example
Updated the README or other documentation
Completed the pre-Request checklist below

Change Description

Currently the XIOS implementation accounts for HField, VertexField, and DGField. #952 extends to cover DGSField. This PR extends further to cover CGField.

This involved checking the global and local dimensions and start index are set up correctly, as well as the XIOS domain, grid, and field.

Several error messages were also made clearer as part of this change.

Test Description

The XIOS read and write tests are updated to account for the CG field type.

A minor looping issue is also fixed in the value checking - it was missing the final value for VertexField. As such, the value checking code in XiosRead_test was reworked to loop over fields first and then dimensions. To make this easier to check, I changed the coordinate field to align with the coordinates of the grid (assuming unit spacing in each direction).

Note that the file reading code accounts for halos out-of-the-box! 👼

Documentation Impact

CGField is mentioned in the XIOS docs page.

Pre-Request Checklist

The requirements of this pull request are fully captured in an issue or design specification and are linked and summarised in the description of this PR
No new warnings are generated
The documentation has been updated (or an issue has been created to track the corresponding change)
Methods and Tests are commented such that they can be understood without having to obtain additional context
This PR/Issue is labelled as a bug/feature/enhancement/breaking change
This change conforms to the conventions described in the README

joewallwork · 2025-10-21T14:43:24Z

[Rebased on top of issue911_xios-dgs]

timspainNERSC

One small, non-essential suggestion: rather than using cice as the name of the CG variable, might I suggest uice, as ice velocity is written by the real model as a CG field, whereas ice concentration never will be.

joewallwork · 2025-10-24T12:12:27Z

One small, non-essential suggestion: rather than using cice as the name of the CG variable, might I suggest uice, as ice velocity is written by the real model as a CG field, whereas ice concentration never will be.

Thanks @timspainNERSC. I made this change in 8f40681. Given that that variable was being used for diagnostics, I switched to output hsnow as a diagnostic. Let me know if you'd prefer to use something different.

TomMelt

Thanks @jwallwork23 looks good to me. I don't think you need to change anything but I do want to make sure I understand the plan for CG Vectors before we merge this (and my changes #969). I will mark as request changes for now, but after we discuss next meeting I can probs just approve this.

TomMelt · 2025-10-27T16:19:17Z

core/src/Xios.cpp

+                } else if (dimType == ModelArray::Dimension::XCG) {
+                    localLength = CGDEGREE * metadata.getLocalExtentX() + 1;
+                    start = CGDEGREE * metadata.getLocalCornerX();
+                } else if (dimType == ModelArray::Dimension::YCG) {
+                    localLength = CGDEGREE * metadata.getLocalExtentY() + 1;
+                    start = CGDEGREE * metadata.getLocalCornerY();


FWIW this is also how I am currently determining the CG dims.... but I would like to hear Tim's comments on PR #969 first. I am worried we might end up in a situation where CG dims is set in several different places. I don't think you need to change this, just raising it as an FYI.

TomMelt · 2025-10-27T16:22:46Z

core/src/Xios.cpp

+                        throw std::runtime_error("Xios: inconsistent global dimensions for "
+                            + dimName + ": " + std::to_string(dim.getSize()) + " vs. "
+                            + std::to_string(
+                                ModelArray::definedDimensions.at(dimType).globalLength));
                    }
-                    if (localLength != ModelArray::size(dimType)) {
-                        throw std::runtime_error(
-                            "Xios: inconsistent local dimensions for " + dimName);
+                    if (localLength != ModelArray::definedDimensions.at(dimType).localLength) {
+                        throw std::runtime_error("Xios: inconsistent local dimensions for "
+                            + dimName + ": " + std::to_string(localLength) + " vs. "
+                            + std::to_string(
+                                ModelArray::definedDimensions.at(dimType).localLength));
                    }
                    if (start != ModelArray::definedDimensions.at(dimType).start) {
-                        throw std::runtime_error("Xios: inconsistent start index for " + dimName);
+                        throw std::runtime_error("Xios: inconsistent start index for " + dimName
+                            + ": " + std::to_string(start) + " vs. "
+                            + std::to_string(ModelArray::definedDimensions.at(dimType).start));


I think this alleviates my prev. concern about different sizes for CG dims in the code. However, I might need to understand how this works a bit better. Where will this be called in NextSim? Is it only called once?

My subsequent PR #968 will make it a bit clearer, but it's only called once yeah. In that follow-up PR I have it happening when the Model gets configured, which in turn calls ModelMetadata.setTime.

TomMelt

Thanks @jwallwork23. It's all good for now

# Automation and pruning for the XIOS handler class Towards #850 Merges into #951 ### Task List - [x] Defined the tests that specify a complete and functioning change (*It may help to create a [design specification & test specification](../../../wiki/Specification-Template)*) - [x] Implemented the source code change that satisfies the tests - [x] Documented the feature by providing worked example - [x] Updated the README or other documentation - [x] Completed the pre-Request checklist below --- # Change Description This PR contains refactoring work to (a) automate the XIOS I/O implementation to streamline the user experience and (b) remove member functions that it turns out aren't needed. #### Automation * Separate `affixModelMetadata` into `parseInputFiles`, `setupDomains`, `setupAxes`, and `setupGrids`. Call `parseInputFiles` in the `Xios` constructor and the rest in the `ParaGridIO` constructor. * Automate setting `nCoords` for `VERTEX` in `setupDomains`. * Automate setting `nComponents` for `DG` in `setupDomains`. * Automate setting `nComponents` for `DGSTRESS` in `setupDomains`. * Automate `enableXios` by inlining it at the start of `Model.configureTime`. #### Pruning * Inline `enableXios` in `Model.configureTime` as mentioned above. * Remove `getClientMPISize` and `getClientMPIRank`. * Remove `setAxisValues` and `getAxisValues`. * No need to call `ModelMetadata.setTime` because it already gets called in `Model.configureTime`. * Inline `isInitialized` at the end of the `Xios` constructor, throwing an error if not. * Drop `getCalendarType` since it's hard-coded to `"Gregorian"`. * Drop `getCalendarTimestep` because this is given by `ModelMetadata.stepLength`. --- # Test Description The XIOS handler tests are reworked to account for the automation. This involved some reordering of how things are done. --- # Documentation Impact The XIOS docs page has been updated. --- # Other information I changed the default `DGCOMP` value in `discontinuousgalerkin/ModelArrayDetails.cpp` to be consistent with the default DG degree that gets set in the root `CMakeLists.txt`. --- ### Pre-Request Checklist - [x] The requirements of this pull request are fully captured in an issue or design specification and are linked and summarised in the description of this PR - [x] No new warnings are generated - [x] The documentation has been updated (or an issue has been created to track the corresponding change) - [x] Methods and Tests are commented such that they can be understood without having to obtain additional context - [x] This PR/Issue is labelled as a bug/feature/enhancement/breaking change - [x] This change conforms to the conventions described in the README

joewallwork self-assigned this Oct 14, 2025

joewallwork added enhancement New feature or request ICCS Tasks or reviews for the ICCS team labels Oct 14, 2025

joewallwork linked an issue Oct 14, 2025 that may be closed by this pull request

XIOS: Handle CGField and DGSField #911

Closed

joewallwork force-pushed the issue911_xios-cg-dgs branch from 5124da0 to 47e77cc Compare October 14, 2025 19:02

joewallwork changed the base branch from issue917_xios-write-diag-time to issue911_xios-dgs October 14, 2025 19:04

joewallwork changed the title ~~XIOS: Handle CGField and DGSField~~ XIOS: Handle CGField Oct 14, 2025

joewallwork force-pushed the issue911_xios-cg-dgs branch from 47e77cc to 5bcc1ce Compare October 14, 2025 19:18

joewallwork force-pushed the issue911_xios-dgs branch from 1006aa6 to 7ab746e Compare October 21, 2025 13:43

joewallwork force-pushed the issue911_xios-cg-dgs branch from 5bcc1ce to 4fc926e Compare October 21, 2025 14:43

joewallwork added 4 commits October 21, 2025 16:00

Align coords field with grid

d9cd786

Fix reporting of inconsistent dims and indices [skip ci]

c40666f

Handle CGFields

330e5c8

Consider CG in XIOS write test

4e9d7e6

joewallwork force-pushed the issue911_xios-cg-dgs branch from 4fc926e to adfb12a Compare October 21, 2025 17:11

joewallwork marked this pull request as ready for review October 21, 2025 17:11

joewallwork added 4 commits October 21, 2025 18:27

Consider CG in XIOS read test

38cdf7b

Mention CGField in XIOS docs

6e7df34

Correct start index setting for CG

e150bc0

Neater error message

536b02b

joewallwork force-pushed the issue911_xios-cg-dgs branch from adfb12a to 536b02b Compare October 21, 2025 17:28

joewallwork requested review from TomMelt, einola and timspainNERSC October 22, 2025 11:22

timspainNERSC approved these changes Oct 22, 2025

View reviewed changes

Switch variable names

8f40681

joewallwork mentioned this pull request Oct 24, 2025

Automation and pruning for the XIOS handler class #968

Merged

11 tasks

TomMelt requested changes Oct 27, 2025

View reviewed changes

joewallwork added 3 commits November 3, 2025 11:37

Merge branch 'issue911_xios-dgs' into issue911_xios-cg-dgs

ae7c4e0

Merge branch 'issue911_xios-dgs' into issue911_xios-cg-dgs

26a3c6f

Merge branch 'issue911_xios-dgs' into issue911_xios-cg-dgs

0aadc1b

joewallwork mentioned this pull request Nov 10, 2025

XIOS: Handle DGSField #952

Merged

11 tasks

Base automatically changed from issue911_xios-dgs to develop November 10, 2025 10:17

joewallwork requested a review from TomMelt November 10, 2025 10:21

TomMelt approved these changes Nov 10, 2025

View reviewed changes

joewallwork merged commit 5266a43 into develop Nov 10, 2025
9 checks passed

joewallwork deleted the issue911_xios-cg-dgs branch November 10, 2025 10:38

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

XIOS: Handle `CGField` #951

XIOS: Handle `CGField` #951

Uh oh!

joewallwork commented Oct 14, 2025 •

edited

Loading

Uh oh!

joewallwork commented Oct 21, 2025

Uh oh!

timspainNERSC left a comment

Uh oh!

joewallwork commented Oct 24, 2025

Uh oh!

TomMelt left a comment

Uh oh!

TomMelt Oct 27, 2025

Uh oh!

TomMelt Oct 27, 2025

Uh oh!

joewallwork Oct 27, 2025

Uh oh!

TomMelt left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

XIOS: Handle CGField #951

XIOS: Handle CGField #951

Uh oh!

Conversation

joewallwork commented Oct 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

XIOS: Handle CGField

Task List

Change Description

Test Description

Documentation Impact

Pre-Request Checklist

Uh oh!

joewallwork commented Oct 21, 2025

Uh oh!

timspainNERSC left a comment

Choose a reason for hiding this comment

Uh oh!

joewallwork commented Oct 24, 2025

Uh oh!

TomMelt left a comment

Choose a reason for hiding this comment

Uh oh!

TomMelt Oct 27, 2025

Choose a reason for hiding this comment

Uh oh!

TomMelt Oct 27, 2025

Choose a reason for hiding this comment

Uh oh!

joewallwork Oct 27, 2025

Choose a reason for hiding this comment

Uh oh!

TomMelt left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

XIOS: Handle `CGField` #951

XIOS: Handle `CGField` #951

joewallwork commented Oct 14, 2025 •

edited

Loading

XIOS: Handle `CGField`