idefix-code · glesur · Apr 16, 2025 · Dec 3, 2024 · Dec 12, 2024 · Dec 14, 2024
diff --git a/.github/workflows/idefix-ci-jobs.yml b/.github/workflows/idefix-ci-jobs.yml
@@ -65,6 +65,8 @@ jobs:
         run: scripts/ci/run-tests $IDEFIX_DIR/test/MHD/OrszagTang -all $TESTME_OPTIONS
       - name: Orszag Tang 3D+restart tests
         run: scripts/ci/run-tests $IDEFIX_DIR/test/MHD/OrszagTang3D -all $TESTME_OPTIONS
+      - name: Axis Flux tube
+        run: scripts/ci/run-tests $IDEFIX_DIR/test/MHD/AxisFluxTube -all $TESTME_OPTIONS
 
   ParabolicMHD:
     runs-on: self-hosted
@@ -177,6 +179,8 @@ jobs:
         run: scripts/ci/run-tests $IDEFIX_DIR/test/MHD/sphBragTDiffusion -all $TESTME_OPTIONS
       - name: Spherical anisotropic viscosity
         run: scripts/ci/run-tests $IDEFIX_DIR/test/MHD/sphBragViscosity -all $TESTME_OPTIONS
+      - name: Collisionless thermal conduction
+        run: scripts/ci/run-tests $IDEFIX_DIR/test/MHD/clessTDiffusion -all $TESTME_OPTIONS
 
   Examples:
     needs: [Fargo, Dust, Planet, ShearingBox, SelfGravity]
@@ -201,3 +205,5 @@ jobs:
         run: scripts/ci/run-tests $IDEFIX_DIR/test/utils/lookupTable -all $TESTME_OPTIONS
       - name: Dump Image
         run: scripts/ci/run-tests $IDEFIX_DIR/test/utils/dumpImage -all $TESTME_OPTIONS
+      - name: Column density
+        run: scripts/ci/run-tests $IDEFIX_DIR/test/utils/columnDensity -all $TESTME_OPTIONS
diff --git a/.gitignore b/.gitignore
@@ -4,6 +4,7 @@ doc/source/_static/*
 doc/source/_public/*
 doc/source/api/*
 doc/source/xml/*
+doc/env/*
 
 # compiled files
 **/__pycache__

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -4,6 +4,24 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.2.01] 2025-04-16
+### Changed
+
+- Fix a bug that led to instabilities in the RKL scheme with very small grid spacings (#323)
+- Fix a bug that prevented Idefix from running with Sycl backend (required on Intel GPUs) (#331)
+- Fix an error that led to incorrect electrical current regularisation around the polar axis in non-Ideal MHD  (#333)
+- Improve div(B) checks with a dimensionless implementation, avoiding too large divB errors in grids with large stretch factors (#334)
+
+### Added
+
+- Time-Implicit drag for multiple dust species, preventing small time steps for strongly coupled dust grains (#321)
+- Collisionless heat flux added to the Braginskii module (#317)
+- New global `idfx::DumpArray` debugging function to dump any Idefix array into a numpy array that can read from python (#318)
+- Automatic benchmark plots in the documentation (#319)
+- More CI tests of grid coarsening (#329)
+- Dump outputs based on wallclock time (#335)
+
+
 ## [2.2.00] 2025-01-17
 ### Changed
 

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -6,9 +6,9 @@ set (CMAKE_CXX_STANDARD 17)
 
 set(Idefix_VERSION_MAJOR 2)
 set(Idefix_VERSION_MINOR 2)
-set(Idefix_VERSION_PATCH 00)
+set(Idefix_VERSION_PATCH 01)
 
-project (idefix VERSION 2.2.00)
+project (idefix VERSION 2.2.01)
 option(Idefix_MHD "enable MHD" OFF)
 option(Idefix_MPI "enable Message Passing Interface parallelisation" OFF)
 option(Idefix_HIGH_ORDER_FARGO "Force Fargo to use a PPM reconstruction scheme" OFF)

diff --git a/doc/python_requirements.txt b/doc/python_requirements.txt
@@ -14,3 +14,4 @@ exhale==0.3.7
 m2r2==0.3.2
 sphinx-copybutton==0.5.2
 #sphinxcontrib-applehelp==1.0.7
+matplotlib==3.10.0
diff --git a/doc/source/bench.json b/doc/source/bench.json
@@ -0,0 +1,162 @@
+[
+  {
+    "date": "2025-03-04_12:57:12",
+    "gpumodel": "v100",
+    "idefix_commit": "2bc09a0d218459f278e2b28506a09e4591b103ae",
+    "bench_commit": "37161676db15115c38fed3f35c94fa447cbac7bd",
+    "results": [
+      {
+        "nbgpu": 1,
+        "cell_updates": 1.193720E+8
+      },
+      {
+        "nbgpu": 2,
+        "cell_updates": 1.178864E+8
+      },
+      {
+        "nbgpu": 4,
+        "cell_updates": 1.155336E+8
+      },
+      {
+        "nbgpu": 8,
+        "cell_updates": 1.014338E+8
+      },
+      {
+        "nbgpu": 16,
+        "cell_updates": 9.855007E+7
+      },
+      {
+        "nbgpu": 32,
+        "cell_updates": 9.012061E+7
+      },
+      {
+        "nbgpu": 64,
+        "cell_updates": 8.538461E+7
+      },
+      {
+        "nbgpu": 128,
+        "cell_updates": 8.531021E+7
+      }
+    ]
+  },
+  {
+    "date": "2025-03-04_13:07:10",
+    "gpumodel": "a100",
+    "idefix_commit": "2bc09a0d218459f278e2b28506a09e4591b103ae",
+    "bench_commit": "b536949200e50fac68d8a46d5db38fc8e3f02da5",
+    "results": [
+      {
+        "nbgpu": 1,
+        "cell_updates": 2.044728E+8
+      },
+      {
+        "nbgpu": 2,
+        "cell_updates": 2.003563E+8
+      },
+      {
+        "nbgpu": 4,
+        "cell_updates": 1.963512E+8
+      },
+      {
+        "nbgpu": 8,
+        "cell_updates": 1.933039E+8
+      },
+      {
+        "nbgpu": 16,
+        "cell_updates": 9.759154E+7
+      },
+      {
+        "nbgpu": 32,
+        "cell_updates": 6.369645E+7
+      },
+      {
+        "nbgpu": 64,
+        "cell_updates": 4.629474E+7
+      },
+      {
+        "nbgpu": 128,
+        "cell_updates": 4.580281E+7
+      }
+    ]
+  },
+  {
+    "date": "2025-03-04_13:16:01",
+    "gpumodel": "h100",
+    "idefix_commit": "2bc09a0d218459f278e2b28506a09e4591b103ae",
+    "bench_commit": "b536949200e50fac68d8a46d5db38fc8e3f02da5",
+    "results": [
+      {
+        "nbgpu": 1,
+        "cell_updates": 3.079643E+8
+      },
+      {
+        "nbgpu": 2,
+        "cell_updates": 3.012300E+8
+      },
+      {
+        "nbgpu": 4,
+        "cell_updates": 2.944091E+8
+      },
+      {
+        "nbgpu": 8,
+        "cell_updates": 2.837224E+8
+      },
+      {
+        "nbgpu": 16,
+        "cell_updates": 2.827778E+8
+      },
+      {
+        "nbgpu": 32,
+        "cell_updates": 2.822657E+8
+      },
+      {
+        "nbgpu": 64,
+        "cell_updates": 2.767820E+8
+      },
+      {
+        "nbgpu": 128,
+        "cell_updates": 2.767322E+8
+      }
+    ]
+  },
+  {
+    "date": "2025-03-06_11:21:56",
+    "gpumodel": "mi250x",
+    "idefix_commit": "2bc09a0d218459f278e2b28506a09e4591b103ae",
+    "bench_commit": "868be0a87c6fcda665cbb62db7020aeff70dc62d",
+    "results": [
+      {
+        "nbgpu": 1,
+        "cell_updates": 1.436580E+8
+      },
+      {
+        "nbgpu": 2,
+        "cell_updates": 1.372499E+8
+      },
+      {
+        "nbgpu": 4,
+        "cell_updates": 1.344528E+8
+      },
+      {
+        "nbgpu": 8,
+        "cell_updates": 1.293602E+8
+      },
+      {
+        "nbgpu": 16,
+        "cell_updates": 1.260359E+8
+      },
+      {
+        "nbgpu": 32,
+        "cell_updates": 1.204980E+8
+      },
+      {
+        "nbgpu": 64,
+        "cell_updates": 1.163099E+8
+      },
+      {
+        "nbgpu": 128,
+        "cell_updates": 1.192343E+8
+      }
+    ]
+  }
+]
diff --git a/doc/source/conf.py b/doc/source/conf.py
@@ -35,6 +35,7 @@
 extensions = [
     "sphinx_rtd_theme",
     'sphinx_git',
+    'matplotlib.sphinxext.plot_directive',
     "breathe",
     "exhale",
     "m2r2",

diff --git a/doc/source/faq.rst b/doc/source/faq.rst
@@ -73,3 +73,6 @@ I want to test a modification to *Idefix* source code specific to my problem wit
 
 I want to use a lookup table from a CSV file in my idefix_loop. How could I proceed?
   Use the ``LookupTable`` class (see :ref:`LookupTableClass`)
+
+I want to compute a cumulative sum (e.g a column density) on the fly. How could I proceed?
+  Use the ``Column`` class (see :class:`::Column`)
diff --git a/doc/source/index.rst b/doc/source/index.rst
@@ -84,8 +84,11 @@ Jonah Mauxion
 Clément Robert
   gitlab integration, linter
 
-Jean Kempf & François Rincon
-    anisotropic diffusion
+Jean Kempf, Victor Reville, François Rincon
+  anisotropic diffusion and collisionless thermal conduction
+
+Marc Coiffier
+  Continuous integration, automatic benchmarking
 
 ========================
 About this documentation

diff --git a/doc/source/modules/braginskii.rst b/doc/source/modules/braginskii.rst
@@ -4,7 +4,7 @@ Braginskii module
 ===================
 
 Equations solved and methods
----------------------------
+----------------------------
 
 The ``Braginskii`` module implements the anisotropic heat and momentum fluxes specific
 to weakly collisional, magnetised plasma like the intracluster medium
@@ -37,7 +37,7 @@ though adapted to vector quantities.
     cell interface by a simple arithmetic average
     (Eq. (6)-(7) from Sharma & Hammett 2007).
     However in the same paper, the authors showed that this implementation can lead to
-    unphysical heat flux from high to low temperature regions.
+    unphysical heat flux from low to high temperature regions.
     So we implemented slope limiters for the computation of these transverse heat fluxes,
     as described in Eq. (17) from Sharma & Hammett (2007).
     Only the van Leer and the Monotonized Central (MC) limiters are available
@@ -72,18 +72,68 @@ of the Braginskii heat flux and viscosity.
 
 .. _braginskiiParameterSection:
 
+
+Saturation with collisionless heat flux
+---------------------------------------
+
+The ``Braginskii`` module can include a collisionless saturation of the Braginskii heat flux, typically due to supra-thermal electrons.
+The heat flux is then computed as follows:
+
+:math:`q = \alpha (q_B + q_\perp) + (1-\alpha)\beta*p*v`,
+
+where :math:`\alpha \in [0,1]` controls the transition between the Braginskii heat flux and the collisionless heat flux
+and :math:`\beta` controls the amplitude of the collisionless heat flux (typically :math:`\beta \in [1,4]`, see Hollweg 1976).
+
+.. note::
+    As a result, even with :math:`\kappa_\perp = 0`, the heat flux is no longer necessarilly strictly aligned with the magnetic field.
+.. note::
+    The collisionless heat flux is a hyperbolic term and the diffusion coefficient is set proportional to :math:`\alpha`.
+.. note::
+    If selected, slope limiters are also used in the collisionless flux, where an upwind scheme has been implemented for stability.
+.. note::
+    This saturation has been thought to be used mostly using the userdef function that takes four userdef arrays as input.
+
 Main parameters of the module
 -----------------------------
 
 The ``Braginskii`` module can be enabled adding one or two lines in the ``[Hydro]`` section
-starting with the keyword
-`bragTDiffusion` or/and *bragViscosity*. The following table summarises the different options
+starting with the keyword `bragTDiffusion` or/and *bragViscosity*. The following tables summarise the different options
 associated to the activation of the Braginskii module:
 
 +--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
 | Column |  Entry name           | Parameter type          | Comment                                                                               |
 +========+=======================+=========================+=======================================================================================+
-| 0      |  bragModule           | string                  | | Activates Braginskii diffusion. Can be ``bragTDiffusion`` or ``bragViscosity``.     |
+| 0      |  bragTDiffusion       | string                  | | Activates Braginskii thermal diffusion.                                             |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 1      | integration           | string                  | | Specifies the type of scheme to be used to integrate the parabolic term.            |
+|        |                       |                         | | Can be ``rkl`` or ``explicit``.                                                     |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 2      | slope limiter         | string                  | | Choose the type of limiter to be used to compute anisotropic transverse flux terms. |
+|        |                       |                         | | Can be ``mc``, ``vanleer`` or ``nolimiter``.                                        |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 3      | saturation mode       | string                  | | Include or not collisionless saturation. Can be ``nosat`` or ``wcless``.            |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 4      | diffusivity type      | string                  | | Specifies the type of diffusivity wanted. Can be ``constant`` or ``userdef``.       |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 5      | parallel diffusivity  | real                    | | Mandatory if the diffusivity type is ``constant``. Not needed otherwise.            |
+|        |                       |                         | | Value of the parallel diffusivity. Should be a real number.                         |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 6      | normal diffusivity    | real                    | | When bragModule ``bragTDiffusion`` and diffusivity type ``constant``,               |
+|        |                       |                         | | value of the normal diffusivity. Should be a real number.                           |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 7      | alpha collisionless   | real                    | | If the diffusivity type is ``constant`` and saturation is ``wcless``.               |
+|        |                       |                         | | Set to 1 if not provided.                                                           |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| 8      | beta collisionless    | real                    | | If the diffusivity type is ``constant`` and saturation is ``wcless``.               |
+|        |                       |                         | | Set to 0 if not provided.                                                           |
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+
+for the *bragViscosity*:
+
++--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
+| Column |  Entry name           | Parameter type          | Comment                                                                               |
++========+=======================+=========================+=======================================================================================+
+| 0      |  bragViscosity        | string                  | | Activates Braginskii viscosity.                                                     |
 +--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
 | 1      | integration           | string                  | | Specifies the type of scheme to be used to integrate the parabolic term.            |
 |        |                       |                         | | Can be ``rkl`` or ``explicit``.                                                     |
@@ -101,7 +151,7 @@ associated to the activation of the Braginskii module:
 +--------+-----------------------+-------------------------+---------------------------------------------------------------------------------------+
 
 Numerical checks
----------------
+----------------
 
 In Cartesian geometry, the ``Braginskii`` module has been tested with many setups
 and in all configurations of magnetic polarisation:
@@ -119,3 +169,5 @@ The same goes for the anisotropic heat flux in Cylindrical/Polar geometry while
 the anisotropic viscosity has *never* been tested in this geometry.
 In spherical geometry, both ``Braginskii`` operators have been partially validated
 (diffusion along the polar axis has not been directly tested).
+
+The collisionless saturation has been tested in 1D and 2D spherical geometry.