Add general-topology support for GOAT-generated geometry by jlore · Pull Request #38 · iterorganization/B2.5

jlore · 2026-08-11T14:16:58Z

This PR adds support for optional explicit topology metadata in b2fgmtry, for GOAT-generated geometries and other general-topology grids.

The main changes are:

Centralize b2fgmtry header reads across B2.5 entry points.
Add optional b2fgmtry version 03.002.002 topology records.
Classify meshes with explicit topology metadata as GEOMETRY_GENERAL.
Use explicit topology data for core boundaries, separatrices, divertor target counts, strike/tangent vertices, and midplane setup.
Guard flux-surface traversal against cycles.
Compute midplane distances from vertex flux surfaces.
Handle b2.transport.inputfile profiles on general topologies.

The file format history is updated in b2mod_version.F for the new optional topology records.

Dependency

This branch is stacked on the B2.5 bugfix PR #17 . Please review/merge that PR first, then this branch can be rebased or retargeted cleanly onto feature/wg-release.

jlore · 2026-08-11T14:25:40Z

@hrdoktorx , @WouterDekeyser , @SVandenKerkhof

This is my progress towards basing the changes I needed from goat-integration-clean, along with other features, on top of the iterorg branch. This goes along with TME-TFE/goat#10

hrdoktorx · 2026-09-08T11:45:50Z

I assume no progress is to be expected here until the Code Camp?

jlore · 2026-09-08T11:48:01Z

Not sure. Sander recently reviewed my GOAT commit and approved of the changes. But I am fine waiting on the Coad Camp

hrdoktorx · 2026-09-08T11:50:55Z

Well, if you are done with it, don't leave it in draft state. Then it can be looked at and merged much sooner.

SVandenKerkhof

Hi Jeremy, sorry for the long wait. Looks like all things are there from the goat-integration-clean-rebased branch, plus some additional useful things like the b2mod_mwti changes and core boundary selection. I didn't yet test it out for my cases on my side here, but OK for me to merge and fix potential bugs along the way.

hrdoktorx

It is not clear to me whether this new format for b2fgmtry is backward compatible and/or if the topology data must be provided mandatorily even for standard supported topologies. It looks like, by providing the topological data, the code then assumes a general topology at the exclusion of any of the standard ones. So format 3.2.2 would need to exist in parallel to 3.2.1. That's not how it works. The new format needs to support the old grid types and add new ones, not replace them. Replacing them would destroy the GGD saving of the data into IMAS, so, from the IO point of view, an absolute non-starter.
I also fail to find in this PR any handling of the IMAS write-up.

hrdoktorx · 2026-09-09T07:30:33Z

+              ic = ic + 1
+            end do
+          end if
+      end if


In the structured code, for linear cases where there was no topological separatrix, the user could still specify a "separatrix line" with the b2mwti_jsep switch. The same capability should be made available for general topologies where the user can designate a particular flux surface as the reference one (and have its default be the first flux surface, like below, that's be okay, provided it is documented ).

hrdoktorx · 2026-09-09T07:33:35Z

-          on_wall = m%fcLbl(iFc).ne.0
-          if (on_wall) then
-            if (iFs1.eq.0) wall_face(1) = .true.
-            if (iFs1.ne.0 .and. iFs2.eq.0) wall_face(2) = .true.


Not sure this change is intentional, or whether it does not break the ITER cases for which it was coded in the first place. Maybe @shtirx or @pshenovaa could try.

hrdoktorx · 2026-09-09T07:37:39Z

@@ -354,4 +354,5 @@ end module b2mod_version
 c                                 Added srw arrays to b2fplasma
 c                                 Added multi-species netCDF tally output
 c                                 Open-source release


If introducing a new version number, you need to change the newversion variable at the top of the b2mod_version and provide an accompanying PR at the SOLPS-ITER level with all rebuilt examples with the new version number.
I recommend you read and follow the CONTRIBUTING rules.

hrdoktorx · 2026-09-09T07:38:39Z

-        mpg%cvOnClosedSurface = mpg%periodic_bc.eq.1
-      else if (mpg%nnreg(0).eq.2 .or. mpg%nnreg(0).eq.4 .or.
-     .         mpg%nnreg(0).eq.5 .or. mpg%nnreg(0).eq.8) then
+      if (mpg%hasTopologicalData .and. mpg%topoID.eq.0) then


This message is gobbledygook. Please write as if the user sees this for the first time.

hrdoktorx · 2026-09-09T07:40:42Z

+          mpg%cvOnClosedSurface = mod(mpg%cvReg,4).eq.1
+        else if (mpg%nnreg(0).eq.7) then
+          mpg%cvOnClosedSurface = mpg%cvReg.eq.1
+        else


I do not think a region of closed surfaces should be assumed. For a linear device, this will always be wrong. For @nthont "coreless DDN" topology as well.

hrdoktorx · 2026-09-09T08:04:39Z

+     .      'boundary face has zero ID - inconsistency in ' //
+     .      'interconnectivity, check input')
+        end if
+


False, incorrect, and misleading.

hrdoktorx · 2026-09-09T08:06:41Z

@@ -78,6 +78,7 @@ subroutine b2agdr (ninp, nout, g, m)
      integer idum(0:9), redef_gmtry
      integer artificial_slab
      character lblgm*120


That's not how this works. You ALWAYS write the output files with "newversion. You can choose to use b2fgmtry_version` here, which is equivalent.

hrdoktorx · 2026-09-09T08:07:50Z

        idum(2) = -1
      endif
      call cfwuin (nout(1), 3, idum, 'nx,ny,nncut')
+


Looks to me that this is not backward compatible. Does this mean that version 03.002.002 can only be used with general topologies and not the standard ones?

hrdoktorx · 2026-09-09T08:08:42Z

-          else
-            call xerrab('Grid version not equal to 03.002.00')
-          endif  ! if of the grid_version
+           else


Again, lead the user to run b2us to convert any older b2fgmtry format.

hrdoktorx · 2026-09-09T08:09:59Z

@@ -529,10 +529,10 @@ subroutine b2agmt (g, m, isymm, artificial_slab)
       ! With this method, numerical error on the 5-point contribution on
       ! gradient calculation is significantly reduced for general grids



I still need a justification, with a concrete example, for the necessity to keep fcQgam positive.

hrdoktorx · 2026-09-09T08:18:03Z

I understand what you're trying to do, but there are a great many things wrong with your approach, and it is incomplete in IO-unacceptable ways. Please read and address my comments in detail.

jlore added 4 commits August 10, 2026 13:23

Fix recombination electron heat balance coefficients

60efeea

Use positive face-normal projection for fcQgam

63f0c4c

Allocate transport input profile buffers dynamically

afa3c52

Rewind b2mn.dat before reading physics overrides

76da083

jlore added 14 commits August 11, 2026 12:55

Fix fixed-form boundary-condition expression

d75cada

Set guard-cell flux tubes after connectivity update

714fd75

Centralize b2fgmtry header reads in preparation for topology metadata

5fc89d0

Add b2fgmtry 03.002.002 topology records

c5a2b9b

Identify explicit topology metadata as general geometry

50d8438

Handle explicit topology in geometry initialization

4254f33

Refine geometry classification for general topologies

285652d

Allow explicit core boundaries for general topologies

a5edfad

Use explicit divertor target count in mwti

dfadbd7

Guard flux-surface traversal against cycles

964fb5f

Use explicit topology separatrices for midplane setup

bbfb44e

Compute midplane distances from vertex flux surfaces

9829878

Handle input profiles on general topologies

25dd430

Fix fixed-form IMP warning output

6e21764

jlore force-pushed the feature/goat-general-topology branch from 3eae93d to 6e21764 Compare August 11, 2026 16:57

hrdoktorx requested review from SVandenKerkhof, WouterDekeyser, antonellozito and hrdoktorx and removed request for antonellozito August 12, 2026 09:36

hrdoktorx added 3 commits August 12, 2026 13:37

Merge branch 'feature/wg-release' into feature/goat-general-topology

f9d3fab

Merge branch 'feature/wg-release' into feature/goat-general-topology

b40b8fd

Merge branch 'feature/wg-release' into feature/goat-general-topology

5f9d2bc

Merge branch 'feature/wg-release' into feature/goat-general-topology

c83713b

Allow explicit topology strike faces with temp_ignore_geo

f45b73f

jlore marked this pull request as ready for review September 8, 2026 12:30

SVandenKerkhof approved these changes Sep 9, 2026

View reviewed changes

hrdoktorx requested changes Sep 9, 2026

View reviewed changes

Merge branch 'feature/wg-release' into feature/goat-general-topology

81dcb89

		@@ -529,10 +529,10 @@ subroutine b2agmt (g, m, isymm, artificial_slab)
		! With this method, numerical error on the 5-point contribution on
		! gradient calculation is significantly reduced for general grids

Uh oh!

Conversation

jlore commented Aug 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Dependency

Uh oh!

jlore commented Aug 11, 2026

Uh oh!

hrdoktorx commented Sep 8, 2026

Uh oh!

jlore commented Sep 8, 2026

Uh oh!

hrdoktorx commented Sep 8, 2026

Uh oh!

SVandenKerkhof left a comment

Choose a reason for hiding this comment

Uh oh!

hrdoktorx left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

hrdoktorx commented Sep 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

jlore commented Aug 11, 2026 •

edited

Loading