Closes #231

See merge request gysela-developpers/gyselalibxx!499

--------------------------------------------

The function `is_spline_interpolation_mesh_uniform` assumes that `GrevilleInterpolationPoints` is used to calculate the interpolation points. This is not always the case so this function should not be used. Rather the interpolation mesh type can be extracted from the interpolation point calculation method.

Closes #223

See merge request gysela-developpers/gyselalibxx!511

--------------------------------------------

Closes #242

See merge request gysela-developpers/gyselalibxx!509

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

…es,vpar,mu)

Closes #200

See merge request gysela-developpers/gyselalibxx!424

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Closes #239

See merge request gysela-developpers/gyselalibxx!516

--------------------------------------------

Closes #238

See merge request gysela-developpers/gyselalibxx!517

--------------------------------------------

Closes #240

See merge request gysela-developpers/gyselalibxx!515

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Closes #247

See merge request gysela-developpers/gyselalibxx!518

--------------------------------------------

See merge request gysela-developpers/gyselalibxx!522

--------------------------------------------

Closes #241

See merge request gysela-developpers/gyselalibxx!527

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Closes #249

See merge request gysela-developpers/gyselalibxx!528

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Closes #230

See merge request gysela-developpers/gyselalibxx!525

--------------------------------------------

… meshes in collisions intra (geometryXVx)

See merge request gysela-developpers/gyselalibxx!523

--------------------------------------------

Closes #251

See merge request gysela-developpers/gyselalibxx!530

--------------------------------------------

In order to handle patches derivatives will be needed at the boundaries of the domain. This MR introduces a structure to link values and derivatives of a field in one structure which fixes #224. DDC does not provide a sub-domain type. Such a type is important as `DiscreteElement`s are linked to coordinates and we want values on the same grid. This MR therefore implements a sub-domain type and stores data internally on mdspans instead of `ddc::Chunk`s.

See merge request gysela-developpers/gyselalibxx!484

--------------------------------------------

Related to #229

See merge request gysela-developpers/gyselalibxx!533

--------------------------------------------

Related to #229

See merge request gysela-developpers/gyselalibxx!532

--------------------------------------------

Add LAPACKE to the docker generated by github. Necessary for the CI of !531

See merge request gysela-developpers/gyselalibxx!534

--------------------------------------------

Add an advection operator which can solve the following equation : 

$\\partial_t f  + A \\partial\_{x_i} f  = 0$

The advection direction can be spatial or along the velocity dimension. We need to provide an advection field. The function \\( f \\) can be only on a full domain (with Species, Space and Velocity), on a space/velocity domain (Space and Velocity) or on a space domain (Space).

See merge request gysela-developpers/gyselalibxx!481

--------------------------------------------

Co-authored-by: Pauline Vidal <pavi@nb-pavi-1.ipp.mpg.de>

Closes #255

See merge request gysela-developpers/gyselalibxx!538

--------------------------------------------

Fixes #207

See merge request gysela-developpers/gyselalibxx!539

--------------------------------------------

…evice

… case to fix

Add some missing links to the source folder of some README to facilitate the access to the right  README on Doxygen.

See merge request gysela-developpers/gyselalibxx!551

--------------------------------------------

Co-authored-by: Pauline Vidal <pavi@nb-pavi-1.ipp.mpg.de>

Closes #219 Add a guiding center in XY geometry for further tests on multipatch geometry. Any new operator to implement, only the simulation. Implementation on CPU in this branch. Maybe an update to GPU or it will be made in another branch.

**New comment:**

* Add a guiding center simulation on XY geometry with Kelvin-Helmholtz test case.
* KelvinHelmholtzInstabilityInitialisation, PredCorrRK2XY operators implemented. 
* Add a geometryXY folder in source. 
* Simulation runs on CPU and GPU.
* Add some python scripts for the diagnosis in a geometryXY/ folder.

See merge request gysela-developpers/gyselalibxx!467

--------------------------------------------

Co-authored-by: Pauline Vidal <pavi@nb-pavi-1.ipp.mpg.de>

…evice

Closes #261.

* Add the option to use `device_t` on `VectorField` and `VectorFieldSpan` too. 
* Add a test `device_host_t.cpp` which tests `device_t` and `host_t` on Chunk, ChunkSpan, VectorField and VectorFieldSpan.  
* Fix a compilation error due to redeclaration of `deepcopy`  in `vector_field_common.hpp` and `derivative_field.hpp`.

See merge request gysela-developpers/gyselalibxx!542

--------------------------------------------

Co-authored-by: Pauline Vidal <pavi@nb-pavi-1.ipp.mpg.de>

Unitary tests were not discovered by gtest and thus not run by the CI.

The `neumann` test is also fixed as it was using the wrong memory and execution space for the spline builder.

See merge request gysela-developpers/gyselalibxx!556

--------------------------------------------

Fixes #265

See merge request gysela-developpers/gyselalibxx!554

--------------------------------------------

There were a few minor issues in the READMEs making the docs harder to read. While fixing #267 I have cleaned these up. The issues were:
- READMEs that were not linked to as described here: https://gyselax.github.io/gyselalibxx/docs_Adding_docs.html#docs_Adding_docs__Documentation_describing_general_methods leading to a large number of sections in the index tab
- Fixes #267 - a bad link when an apostrophe is used in the section header
- Bad links to classes only defined via template specialisation. This is fixed through the use of anchors
- Python syntax used to comment markdown code in the collision documentation

See merge request gysela-developpers/gyselalibxx!555

--------------------------------------------

- Add README in /geometryRTheta/geometry/
- Clean geometry file.
- Use aliases for VectorFieldSpan.
- Remove depencies on species_info.hpp.
- Reorder some #include in advection test folder.

See merge request gysela-developpers/gyselalibxx!558

--------------------------------------------

Co-authored-by: Pauline Vidal <pavi@nb-pavi-1.ipp.mpg.de>

…ed to fix execspace devicet mismatch

`fluid_masses` was defined on wrong domain in `species_init`

See merge request gysela-developpers/gyselalibxx!557

--------------------------------------------

Add a spline builder on multipatch geometry. 

This builder calls the builders on each patch. 

It uses std::tuple.

See merge request gysela-developpers/gyselalibxx!549

--------------------------------------------

Co-authored-by: Pauline Vidal <pavi@nb-pavi-1.ipp.mpg.de>

…evice

* Splines LAPACK backend
* `ddc::parallel_for_each` now can take a label! Usefull for profiling.
* Align `UniformBSplines` API to `NonUniformBSplines` API
* Change terminology in splines (ie. `spline_domain` -\> `batched_spline_domain`... complete list: CExA-project/ddc#382)
* Ginkgo throws error in case of non-convergency.
* Bsplines API is based on `mdspan` in place of `std::array`.
* Deprecate `nbe_xmin` and `nbe_xmax` in splines (it is redundant with `nbc_xmin`and `nbc_xmax` for all boundary conditions except NATURAL which has been removed because unused).
* Update supposedly-ddc-internal SplinesLinearProblem API (which is used by quadrature).
* Lot of minor internal changes.

See merge request gysela-developpers/gyselalibxx!548

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Now that the splines have been migrated to DDC, DDC's tests are quite slow. Gysela is the only place where these tests are run so they should not be removed completely, however currently DDC's tests are slowing down the pipeline dramatically.
The following is an example of the times currently taken by an example pipeline:
- cmake_tests : 13m
- cmake_tests_Release : 16m
- cmake_tests_Release_OMP : 19m
- cmake_tests_Release_GPU : 38m

This PR treats DDC in a similar way to SLL and the Polar Poisson solver and only runs these tests if the contents of the associated folder have been modified

See merge request gysela-developpers/gyselalibxx!561

--------------------------------------------

Closes #270

See merge request gysela-developpers/gyselalibxx!559

--------------------------------------------

Adds realistic calculations for the reactions rates of neutrals for three processes: charge exchange, recombination and ionization.

See merge request gysela-developpers/gyselalibxx!503

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>
Co-authored-by: Thomas Padioleau <thomas.padioleau@cea.fr>
Co-authored-by: Baptiste LEGOUIX <baptiste.legouix@cea.fr>

An index is incorrectly named as though it were a discrete dimension

See merge request gysela-developpers/gyselalibxx!560

--------------------------------------------

See merge request gysela-developpers/gyselalibxx!563

--------------------------------------------

Fixes #264 . The reference commit remains unchanged at 7f12edc26261ef3b97dd451cddb5cd8964ff57ca

See merge request gysela-developpers/gyselalibxx!568

--------------------------------------------

Updates neutrals temperature to be equal to the ion one.

See merge request gysela-developpers/gyselalibxx!536

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>
Co-authored-by: jblanco <jblanco@login5.head.adastra.cines.fr>
Co-authored-by: Baptiste LEGOUIX <baptiste.legouix@cea.fr>

Fixes neutrals utils post-process to correctly compute the reaction rates.

See merge request gysela-developpers/gyselalibxx!570

--------------------------------------------

Fixes #218

See merge request gysela-developpers/gyselalibxx!521

--------------------------------------------

Removes a superfluous line from the script and the corresponding PDI as neutral temperature = ion temperature.

See merge request gysela-developpers/gyselalibxx!571

--------------------------------------------

Co-authored-by: jblanco <jblanco@login5.head.adastra.cines.fr>

Second to last step of #259 . DDC should be updated immediately after this MR as otherwise it will break the CI

See merge request gysela-developpers/gyselalibxx!566

--------------------------------------------

Closes #248

See merge request gysela-developpers/gyselalibxx!520

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Add toolchains and environment setup files for persee and the docker environment. This allows the CI commands to be simplified and removes duplication of code in the CI specifications

See merge request gysela-developpers/gyselalibxx!565

--------------------------------------------

Must be merged after #248.
Closes #260.

See merge request gysela-developpers/gyselalibxx!573

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Merge !566 first

Closes #259 

Update to Ginkgo 1.8.0

In splines multiple objects are renamed:

`interpolation_mesh_type` -> `interpolation_discrete_dimension_type`

`knot_mesh_type` -> `knot_discrete_dimension_type`

`is_uniform_mesh_v` -> `is_uniform_discrete_dimension_v`

`tag_type` -> `continuous_dimension_type`

additionally `continuous_dimension_type` (previously `tag_type`) is now publicly exposed (we previously needed to rely on `interpolation_mesh_type` to get access to `tag_type`)

See merge request gysela-developpers/gyselalibxx!572

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Closes #209

See merge request gysela-developpers/gyselalibxx!574

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Create csr batched solver in sll with associated tests, the class name is MatrixBatchCsr.

Closes #233

See merge request gysela-developpers/gyselalibxx!505

--------------------------------------------

Co-authored-by: Emily Bourne <emily.bourne@epfl.ch>

Closes #282

See merge request gysela-developpers/gyselalibxx!578

--------------------------------------------

Kokkos kernels backend

See merge request gysela-developpers/gyselalibxx!575

--------------------------------------------

See merge request gysela-developpers/gyselalibxx!579

--------------------------------------------

See merge request gysela-developpers/gyselalibxx!567

--------------------------------------------

Closes #286

See merge request gysela-developpers/gyselalibxx!581

--------------------------------------------

Closes #279

See merge request gysela-developpers/gyselalibxx!580

--------------------------------------------

Update spack environments on Persee

Closes #279

See merge request gysela-developpers/gyselalibxx!580

I've been able to successfully install spack and run simulations on Adastra with this toolchain from a fresh env.

- Rely on CINES's cmake module (for some reason the spack cmake does not anymore appears in `gcc/12.1/zen3/cmake/`. Maybe because the CINES' one is discovered by spack ?)
- Remove the Etienne's trick passing twice through the `module load`. It does not seem necessary.

See merge request gysela-developpers/gyselalibxx!553

--------------------------------------------

Co-authored-by: Etienne Malaboeuf <eti.malaboeuf@gmail.com>
Co-authored-by: blegouix <blegouix@login5.head.adastra.cines.fr>

Adastra update toolchain

See merge request gysela-developpers/gyselalibxx!553

Closes #278

See merge request gysela-developpers/gyselalibxx!583

--------------------------------------------

Create ScopeGuards after possible std::exit

Closes #278

See merge request gysela-developpers/gyselalibxx!583

Adapt fluid moments class to prepare for quadrature operator on GPU by ensuring that data is on the GPU when it is passed to the operators

See merge request gysela-developpers/gyselalibxx!585

--------------------------------------------

Co-authored-by: abdelhadi.kara@cea.fr <karaa@ruche01.cluster>

Adapt fluid moments operators to get GPU allocated data

See merge request gysela-developpers/gyselalibxx!585

Closes #127

See merge request gysela-developpers/gyselalibxx!526

--------------------------------------------

Implement MPI AlltoAll transposer

Closes #127

See merge request gysela-developpers/gyselalibxx!526