Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[UNRELEASED]

Added

  • Add GVEC in the toolchains.

Fixed

  • Fix compilation errors due to use of std::make_tuple in functions that are GPU compatible.
  • Fix errors in InterfaceDerivCoeffs for uniform patches with a different \(\Delta x\).

Changed

  • Rename SingleInterfaceDerivativesCalculator to InterfaceDerivCoeffs.
  • Rename SingleInterfaceDerivativesCalculatorCollection to InterfaceDerivCoeffsCollection.
  • Rename SingleInterfaceDerivativesCalculator::get_function_coefficients() to InterfaceDerivCoeffs::get_approx_deriv().
  • Rename file single_interface_derivatives_calculator.hpp to interface_derivative_coefficients.hpp.
  • Rename file single_interface_derivatives_calculator_collection.hpp to interface_derivative_coefficients_collection.hpp.

Deprecated

Removed

  • Removed .span_view and .span_cview from Gyselalib++ objects.

Dependency requirements

  • Remove Dask dependency in the tests.
  • Update DDC to v0.16.0.

[v0.8.0] - 2026-09-09

Added

  • Add an extrapolation_rule argument to GMGPolarPoissonLikeSolver. Default no extrapolation.
  • Add a new constructor for GaussLegendre from an index range describing the cell edges.
  • Add a GradientCreator operator to group derivative calculations.
  • Add a NDLagrangeEvaluator class.
  • Add eval_basis_and_n_derivs function to Lagrange basis operators.
  • Add deriv function to Lagrange evaluation.
  • Add concepts::InterpolationBuilder1D and concepts::Interpolation1D.
  • Add new constructors for MultipatchFieldMem to allow labelling.
  • Add more labels to memory allocations.
  • Add a NDIdentityInterpolationBuilder class.
  • Add a new abstract class IPolarPoissonLikeSolver.
  • Add data type parametrisation to ConstantIdentityInterpolationExtrapolationRule.
  • Add support for arbitrary precision quadrature.
  • Add support for arbitrary precision norm calculations.
  • Add an element-wise version of SplinePolarFootFinder to reduce memory consumption.
  • Add a landau4d_fft_lagrange target for 3D Landau damping with Lagrange interpolation.
  • Add an optional IdxRange argument to the constructor of LagrangeInterpolator.
  • Add a new init_lagrange_dependent_idx_range method.
  • Add a MPI scope guard.
  • Add MI250X toolchain for LUMI.
  • Add a CMake option to control the discovery timeout of Gyselalib++ tests.
  • Added PolarFootFinder to replace SplinePolarFootFinder for clarity of Advection and Foot finding spaces.
  • Allow polar advection to modify a subset of the distribution function.
  • Add a LocalPartialDerivativeCreator concept to describe partial derivatives that can be called locally.
  • Add a method get_derivative to CentralFDMPartialDerivative so it respects the LocalPartialDerivativeCreator concept.
  • Add class GMGPolarPoissonLikeSolver to allow the use of GMGPolar as a polar Poisson solver.
  • Add GMGPolar in the toolchains.
  • Allow SplineInterpolator and LagrangeInterpolator to specify custom extrapolation rules.
  • Allow IdentityInterpolationBuilder class to take a field on a strided layout.
  • Add a DiscreteMapping class to handle ND mappings whose values are only known at the mesh points of a grid.
  • Add a CoordWithOPoint class to recognise radial and poloidal components of a 2D coordinate.
  • Add MI300 toolchain for Adastra.
  • Add CMake installation commands and call the package gyselalibxx.

Fixed

  • Fix missing load of pdiplugin-pycall in some Spack-based toolchains.
  • Fix missing update of Spack repos before loading environments.
  • Fix missing PDIEvent(initialisation) in the guiding centre (X,Y) simulation.
  • Fix ddcHelper::assign_elements to allow a ND tensor to be truncated to a MD tensor (with M\<N).
  • Fix use of PolarSplineFEMPoissonLikeSolver with a ND metric tensor (with N>2).
  • Decrease memory usage in BslAdvection1D.
  • Fix compute_coeffs_on_mapping to allow integration when a coordinate change allows the determinant of the Jacobian to be calculated with less information than is required to calculate the Jacobian matrix.
  • Fix hardcoded Homogeneous Hermite boundary conditions in BslAdvection1D.
  • Decrease memory usage in SplinePolarFootFinder.
  • Fix GCC version on Adastra toolchains.
  • Fix use of BslAdvectionSpatial and BslAdvectionVelocity with non-double precision.
  • Fix H100 toolchain on Jean-Zay.
  • Fix Lagrange basis non-uniform initialisation for a sub-domain.
  • Fix use of ExtrapolationRule::Constant for 2D splines.
  • Fix use of a generic interpolator in FEM1DPoissonSolver, PolarFootFinder and BslAdvectionPolar.
  • Fix use of polar advection with DiscreteMapping class.

Changed

  • Changed the name of class SplineBuliderDerivField2D to fix typo (->SplineBuilderDerivField2D).
  • Update DDC to v0.15.0 (see also v0.14.0 and v0.13.0).
  • Changed FindLAPACKE CMake module to the version in DDC.
  • Renamed DiscreteToCartesian -> DiscretePoloidalCSSplineMapping.
  • Renamed DiscreteToCartesianBuilder -> DiscretePoloidalCSSplineMappingBuilder.
  • Changed BslAdvectionPolar template parameters and constructor to take a builder and evaluator instead of an Interpolator2D.
  • Changed new concepts::InterpolationEvaluator concept and associated InterpolationEvaluatorTraits class to generalise to ND.
  • Changed new concepts::InterpolationBuilder concept and associated InterpolationBuilderTraits class to generalise to ND.
  • Renamed ACTIVATE_RESTART_TESTS -> GYSELALIBXX_ACTIVATE_RESTART_TESTS.
  • Renamed POISSON_2D_BUILD_TESTING -> GYSELALIBXX_POISSON_2D_BUILD_TESTING.
  • Pin toolchains to Python 3.13.
  • Disable some costly variants in Spack based toolchains.
  • Changed the order of arguments to PolarSplineFEMPoissonLikeSolver to respect output argument first convention.
  • Use any callable in PolarSplineFEMPoissonLikeAssembler instead of spline coefficients.
  • Add interpolation operators to PolarSplineFEMPoissonLikeSolver constructor.
  • Take pointwise values for PolarSplineFEMPoissonLikeSolver::update_coefficients.
  • Allow pointwise values to be passed to PolarSplineFEMPoissonLikeSolver::operator().
  • Rename polarpoissonlikesolver.hpp -> polar_spline_fem_poisson_like_solver.hpp.
  • Allow the components and determinant of the Jacobian of a coordinate transformation to be any floating point precision.
  • Prefixed the name of the Kokkos region with "(GSLX)"
  • Use Simpson quadrature for XYVxVy Landau damping.
  • Update py-dask in the CPU Spack toolchain.
  • Disable python for PDI and pdiplugin-pycall on macOS.
  • Enforce version of Kokkos Tools in all toolchains for reproducibility.
  • Remove default O1 optimisation flag.
  • Changed type of species identifier read in read_species PDI event (from int to int64).
  • Remove version constraint on the indirect dependency Kokkos-FFT.
  • Use uppercase L suffix for long double literals.
  • Allow passing fields defined on strided domains to BslAdvection1D.
  • Allow run_cppcheck CI script to be run in parallel.
  • Simplify restrict_to_idx_range implementation.
  • Rename CMake options POLAR_SPLINES_TEST_DEGREE_[MIN/MAX] to GYSELALIBXX_POLAR_SPLINES_TEST_DEGREE_[MIN/MAX].
  • Changed ExtrapolationRule::RULENAME from an enum to ExtrapolationRule::RuleName, a struct tag in a namespace.
  • Remove the C compiler dependency.
  • Make the dependency on GoogleTest dependent on the CMake option GYSELALIBXX_BUILD_TESTING.
  • Group extrapolation rules by dimension in SplineInterpolator and LagrangeInterpolator.
  • Group spline boundary closure rules by dimension in SplineInterpolator.
  • Change the LagrangeInterpolator templates to allow ND cases to be handled.
  • Change the SplineInterpolator templates to allow 2D cases to be handled.
  • Setup GMGPolar in GMGPolarPoissonLikeSolver::update_coefficients instead of GMGPolarPoissonLikeSolver::operator().
  • Use Interpolator concept instead of Builder and Evaluator classes to simplify classes:
  • PolarFootFinder
  • DiscretePoloidalCSSplineMappingBuilder
  • RefinedDiscretePoloidalCSSplineMappingBuilder
  • BslPredCorrRTheta
  • BslExplicitPredCorrRTheta
  • BslImplicitPredCorrRTheta
  • PolarSplineFEMPoissonLikeSolver
  • GMGPolarPoissonLikeSolver
  • The CMake namespace gslx has been renamed gyselalibxx.

Deprecated

  • Deprecated SplinePolarFootFinder.

Removed

  • Remove DDC submodule.
  • Remove 2D Interpolator classes:
  • IInterpolator2D
  • SplineInterpolator2D
  • Remove Google Test submodule.
  • Remove ruche toolchain.
  • Remove GYSELALIBXX_VERSION_* CMake variables.
  • Remove out-of-date Leonardo toolchain.
  • Remove unuseful defaulted template parameters MinBound and MaxBound from LagrangeInterpolator.
  • Remove unused superclass IPolarFootFinder.
  • Remove BslAdvectionPolar::operator() taking advection_field_xy_centre.
  • Remove cartesian_tag_* aliases in mappings.
  • Remove cylindrical_tag_* and toroidal_tag_* aliases in ToroidalToCylindrical.
  • Remove LAPACK dependency (LAPACKE remains).

[v0.7.0] - 2026-03-18

Added

  • Add SplineInterpolator class to group a compatible DDC spline builder and evaluator.
  • Add LagrangeInterpolator class to group a compatible builder and evaluator.
  • Add constructors taking a concepts::Interpolator class instance for:
  • BslAdvectionSpatial
  • BslAdvectionVelocity
  • BslAdvection1D
  • FEM1DPoissonSolver
  • Add a constructor to BslAdvection1D to use the same interpolator type for both the advected function and the advection field.
  • Add new constructors for VectorFieldMem to support labels as used in DDC and Kokkos.
  • Added index range assertions in Quadrature.

Fixed

  • Add missing enable_tensor_type for CartesianLeviCivitaTensor and LeviCivita size().

Changed

  • Add Koliop in all toolchains.
  • Changed spline boundary condition in velocity dimensions in the XVx and XYVxVy geometries to use homogeneous Hermite boundary conditions.
  • Change BslAdvectionVelocity to stop providing values for the derivatives at the boundaries.
  • Add labels to all parallel constructs and many variable allocations.
  • Completed the porting of PolarSplineFEMPoissonLikeAssembler to GPU (less memory, fewer data transfers, faster execution).
  • Changed interface of PolarSplineFEMPoissonLikeSolver to split initialisation from coefficient setting.

Deprecated

  • Deprecate constructors taking both builders and evaluators n favour of constructors taking Interpolator classes for:
  • BslAdvectionSpatial
  • BslAdvectionVelocity
  • BslAdvection1D
  • FEM1DPoissonSolver

Removed

  • Remove deprecated Lagrange class.
  • Remove Koliop submodule.

[v0.6.2] - 2026-03-06

Fixed

  • Ensure CoordRTheta is defined in polarpoissonlikeassembler.hpp.

[v0.6.1] - 2026-03-06

Fixed

  • Allow neumann_spline_quadrature_coefficients_1d to work with HOMOGENEOUS_HERMITE boundary conditions.
  • Fix non-uniform periodic interpolation point selection.

Changed

  • Modified Quadrature to allow the batched operator to work on strided data.

Removed

  • Removed the mini-app and moved it to its own repository mini-app.

[v0.6.0] - 2026-03-05

Added

  • Add UniformLagrangeBasis and NonUniformLagrangeBasis classes describing Lagrange bases using the second barycentric formulation.
  • Add the fluid moments computation in pycall block to the mini-application
  • Add IdentityInterpolationBuilder to copy data required for an interpolation operator.
  • Add LagrangeEvaluator to evaluate a Lagrange polynomial centred on a given point.
  • Added error messages when wrong input is provided to a simulation.
  • Add concepts::InterpolationBuilder to make IdentityInterpolationBuilder and ddc::SplineBuilder interchangeable.
  • Add concepts::InterpolationEvaluator to make LagrangeEvaluator and ddc::SplineEvaluator interchangeable.
  • Allow BslAdvection1D, BslAdvectionSpatial, and BslAdvectionVelocity to use variable precision.

Fixed

  • Fix incorrect memory_space type alias in DerivFieldMem.
  • Fix incorrect memory_space type alias in DerivField.
  • Fix a memory leak in DerivFieldMem.
  • Fix ddc::coordinate called on ddc::DiscreteElement outside of the domain of definition in single_interface_derivatives_calculator.hpp and tests.
  • Fix a memory leak related to an object of type PC_tree_t not destroyed.
  • Fix boundary conditions in single_interface_derivatives_calculator_collection_test.cpp.
  • Fix the finite differences method on a periodic domain.
  • Fix the warning for using the deprecated target PDI::pdi starting from PDI 1.10.1.

Changed

  • Remove ddc_sync utility.
  • Use generic binaries in the CPU Spack toolchain.
  • Remove Kokkos-related submodules.
  • Remove recursion from submodules synchronisation.
  • Add missing module purge in Persee toolchains.
  • Advection operators are now templated on InterpolationBuilder and InterpolationEvaluator concepts instead of Interpolator classes.
  • Use GCC 14 in the Jean-Zay toolchain.
  • Reduce memory footprint of FEM1DPoissonSolver.
  • Update DDC to v0.11.0.
  • Moved assembly of the stiffness matrix from PolarSplineFEMPoissonLikeSolver to new PolarSplineFEMPoissonLikeAssembler class.

Deprecated

  • Deprecated Lagrange class.

Removed

  • Remove the default PDI configuration file for the mini-app.
  • Remove Interpolator classes:
  • IInterpolator
  • LagrangeInterpolator
  • SplineInterpolator

[v0.5.0] - 2026-02-04

Added

  • Add a temporary SplineBuliderDerivField2D to allow building a 2D spline representation from data stored in a DerivField.
  • Allow FFTPoissonSolver to use variable precision.
  • Add SingleInterfaceDerivativesCalculator to compute an interface derivative of an equivalent global spline.
  • Add SingleInterfaceDerivativeCalculatorCollection to collect different SingleInterfaceDerivativesCalculator.
  • Add a mini-application gys_io.cpp to test I/O performance and in-situ diagnostics on 5D distribution functions
  • Add Kokkos Tools in the environment on Persee toolchains
  • Add a constructor of SingleInterfaceDerivativesCalculator for the approximated interface derivatives.
  • Added support for A100 Raven cluster Spack toolchain.
  • Add new function ddcHelper::assign_elements to assign elements of a tensor to another tensor containing the same elements.
  • Add a new coordinate transformation LinearCoordTransform.
  • Add a new coordinate transformation OrthogonalCoordTransforms.

Fixed

  • Fix derivative indexing of a DerivField object.
  • Fix transposition of arrays with more than 7 dimensions.
  • Fix missing guards in FindLAPACKE.cmake leading to duplicate target.
  • Fix spurious segfaults for toolchains based on Spack environment views.

Changed

  • Extract spline definitions from geometry.hpp files into files called spline_definitions_<geom_descriptor>.hpp which are dedicated to tests or simulations.
  • Rename geometry.hpp files to geometry_<geom_descriptor>.hpp to reflect the geometry that they describe.
  • DerivField objects are initialised on GPU by default.
  • Update DDC to v0.10.0.
  • Change some Python class scattered_coord arguments to keyword-only.
  • Increase C++ version to 20.
  • Use concepts to describe coordinate transformation classes.
  • Update koliop to v0.1.2.
  • Make koliop discoverable with find_package.
  • Simplify toolchains by disabling koliop LTO by default.
  • Update Persee toolchains to GCC 13.

Deprecated

  • PolarBSplines::eval_deriv_r, PolarBSplines::eval_deriv_theta, and PolarBSplines::eval_deriv_r_and_theta are deprecated in favour of PolarBSplines::eval_deriv.
  • PolarSplineEvaluator::deriv_dim_1, PolarSplineEvaluator::deriv_dim_2, and PolarSplineEvaluator::deriv_dim_1_and_2 are deprecated in favour of PolarSplineEvaluator::deriv.

Removed

  • Remove unused temporal dimension T in geometryXVx.

[v0.4.1] - 2025-12-12

Fixed

  • Allow access to slices extracted from const DerivField.
  • Fixed LAPACKE CMake target definition.
  • Uses patched cray-mpich package forcing using of the GTL on dependencies.

[v0.4.0] - 2025-12-10

Added

  • Add getters to DerivField types to access the associated index ranges.
  • Add H100 Jean-Zay toolchain.
  • Add an operator() in BslAdvectionPolar to advect a function with an advection field along <R, Theta>. The operator averages the values of the advection field on the first ring to get its value at the O-point.
  • Add static_assert expressions for spline builder/evaluator pairs.
  • Added equilibrium::init_from_input method to choose XVx equililibrium in input file.

Fixed

  • Ensure std::abs or Kokkos::abs is preferred over abs.
  • Specify return type for Lie-Poisson operator explicitly for better error messages.
  • Ensure copy_to_vector_space can be called for different memory layouts.

Changed

  • Use patched recipes for the CPU Spack toolchain.
  • Inject Kokkos Tools lib directory to LD_LIBRARY_PATH in the Adastra toolchains.
  • Update and reorganize the Persee toolchains.
  • Add pdiplugin-pycall to Persee toolchains.
  • Add an assertion to PolarSplines to ensure that the domain matches the assumptions.
  • Allow a local installation of Kokkos > v4.4.1 to be used by CMake.
  • Rename static variable is_curvilinear_2d_mapping_v to the more accurate: is_coord_transform_with_o_point_v.
  • Move Python cache files out of the Spack installation tree on Adastra and Jean-Zay.
  • Use CCFR environment variables on CINES and IDRIS machines when possible.
  • Update DDC dependency to v0.9.0.
  • Combine landau_fft and bumpontail_fft executables into a vlasovpoisson_xvx_fft executable.
  • Combine landau_fem_uniform and bumpontail_fem_uniform executables into a vlasovpoisson_xvx_fem_uniform executable.
  • Make the choice of equililibrium an input parameter for vlasovpoisson_xvx executables.
  • Move static class method BumpontailEquilibrium::init_from_input to namespace bumpontail_equilibrium::init_from_input.
  • Move static class method MaxwellianEquilibrium::init_from_input to namespace maxwellian_equilibrium::init_from_input.
  • Update CPU Spack toolchain to Spack v1 and update packages, for example Kokkos 4.7, Python 3.12:.
  • Update Adastra Spack toolchains to Spack v1 and update packages, for example Kokkos 4.7, Python 3.12:. Use OpenBLAS instead of Cray LibSci and raw GCC compilers.
  • Update Jean-Zay Spack toolchain to Spack v1 and update packages, for example Kokkos 4.7, Python 3.12:.
  • Use LAPACKE to call LAPACK functions.

[v0.3.0] - 2025-09-03

Added

  • Add a curl operator.
  • Port PolarSplineEvaluator methods to GPU.
  • Add methods to PolarSplineEvaluator to avoid unnecessary creation of fields of coordinates.
  • Allow a DerivField to be stored in a MultipatchField.
  • Add DerivFieldOnPatch and IdxRangeSliceOnPatch aliases for the MultipatchField in types.hpp.
  • Allow min and max from math_tools.hpp to be called from GPU.
  • Add a periodic_strips_non_uniform_2d_9patches geometry.

Fixed

  • Modify ruche.v100/environment.sh file to fix tokamaxi simulation segfault issues.
  • Fix type of derivatives stored in DerivFieldMem and DerivField types.
  • Fixed memory error in PolarSplineFEMPoissonLikeSolver.
  • Remove non-parallelisable loop in PolarSplineFEMPoissonLikeSolver::init_nnz_per_line.
  • Remove use of std::cyl_bessel_j which is not available in libc++.
  • Fix mi250.hipcc.adastra.spack toolchain.
  • Fix uninitialised values being used as an initial guess for the result of the matrix equation in PolarSplineFEMPoissonLikeSolver.
  • Fix missing grids when calling collect_grids_on_dim_t.
  • Fix is_borrowed_deriv_field_v<> in derivative_field_common.hpp file.
  • Fix phi_eq in diocotron and vortex-merger simulations.

Changed

  • Change interface of EdgeTransformation::search_for_match to return an out_of_bounds_idx instead of a boolean.
  • Change spack setup in CPU installation script (prepare.sh) to create and use independent spack installation.
  • Change template parameters of PolarSplineEvaluator to add execution and memory space information.
  • Allow get_idx_range to be called from a GPU execution space.
  • Uniformise toolchains.
  • Allow batch CSR convergence parameters to be specified in the constructor of PolarSplineFEMPoissonLikeSolver.
  • Change the internals of PolarSplineFEMPoissonLikeSolver to precalculate fewer values.
  • Change the internals of PolarSplineFEMPoissonLikeSolver to avoid calls to DDC's internals.
  • Clean up code in BslPredCorrRTheta::operator().
  • Clean up code in BslExplicitPredCorrRTheta::operator().
  • Clean up code in BslImplicitPredCorrRTheta::operator().
  • Ported PolarSplineFEMPoissonLikeSolver::operator() to GPU. The RHS function passed as argument will now be evaluated on GPU.
  • Change constructor arguments of BslImplicitPredCorrRTheta to pass a spline builder and evaluator which operate on GPU.
  • Remove the const version of get_values_field() and change it into get_values_const_field() in DerivFieldCommon.
  • Change alias in (r, theta) geometry DConstVectorFieldRTheta->DVectorConstFieldRTheta.
  • Change constructor arguments of VortexMergerEquilibria to pass a spline builder and evaluator which operate on GPU.
  • Port PoissonLikeRHSFunction from the (r, theta) geometry to GPU.
  • Change constructor argument of PoissonLikeRHSFunction from the (r, theta) geometry to pass spline coefficients on GPU.
  • Use Spack to install the Kokkos ecosystem in the MI250 Adastra toolchain.
  • Use Spack to install the Kokkos ecosystem in the GENOA Adastra toolchain.
  • Use a single rocm stack on Adastra.

Deprecated

Removed

  • Remove deprecated method PolarBSplines::integrals.
  • Remove unhelpful PolarSpline classes in favour of DField<IdxRange<PolarBSplines>> types.
  • Remove unused broken method PolarSplineEvaluator::integrate.
  • Remove min and max from math_tools.hpp (functionality is available in std and Kokkos namespaces).

[v0.2.0] - 2025-07-03

Added

  • Add a Gyroaverage operator with tests for circular geometry.
  • Curvilinear coordinate change classes have an O-point method to retrieve the O-point in the non-curvilinear coordinates.
  • Add a batched operator() to DiscreteToCartesian allowing a field of coordinates to be converted.
  • Add a LiePoissonBracket::operator() overload which takes a 2D tensor as the second argument to the bracket.
  • Add a function PDI_expose_vector_field to output a vector field via PDI.
  • Add a control_points method to DiscreteToCartesian to allow all control points to be retrieved at once.

Fixed

  • Fix uninitialized warning in the Tensor class.
  • Fix unused m_magnetic_field variable in MaxwellianEquilibrium class.
  • Fix break points incorrectly labelled as knots.
  • Fix minimum version requirement of Kokkos.
  • Fix tolerance of floating point comparisons in JacobianMatrixAndJacobianCoefficients and MultipatchSplineEvaluatorTest tests
  • Fix unnecessary std::move calls.
  • Fix missing assertion in LeviCivitaTensor to prevent division by 0 when Jacobian is calculated at singular point.
  • Fix bad result of is_tensor_v for IdentityTensor.

Changed

  • Bumped the minimum CMake version to 3.25 to benefit from the add_subdirectory(.. SYSTEM) feature.
  • Change the interface of IVlasovSolver and IQNSolver in geometryXYVxVy to store the electric field in a VectorField.
  • Integration of ddc::StridedDiscreteDomain by making IdxRangeSlice a type alias.
  • The parameter iter_start has been removed from the constructor of RestartInitialisation.
  • Generalise compute_coeffs_on_mapping to work with any mapping.
  • Rely on GPU-aware MPI to allow GPU-direct MPI for MPITransposeAllToAll.
  • Curvilinear coordinate change classes take a Coord type to specify the O-point in the constructor.
  • Allow init_discrete_space to be used to initialise PolarBSplines with a GPU-based DiscreteToCartesian coordinate change operator.

Deprecated

[v0.1.1] - 2025-05-30

Fixed

  • Fix paths in root CMakeLists.txt file to ensure it can be correctly used in a submodule.
  • Update remaining use of ddc::Coordinate to use Gyselalib++ conventions (Coord).
  • Update coding conventions to match what is applied.

[v0.1.0] - 2025-05-28

Added

  • First release of Gyselalib++
  • Advection operators
  • 1D Semi-Lagrangian spatial advection (\( \frac{df_s}{dt}= \sqrt{\frac{m_e}{m_s}} v \frac{\partial f_s}{\partial x} \))
  • 1D Semi-Lagrangian velocity advection (\( \frac{df_s}{dt}= q_s \sqrt{\frac{m_e}{m_s}} E \frac{\partial f_s}{\partial v} \))
  • 1D Semi-Lagrangian advection with a provided advection field
  • 2D Semi-Lagrangian advection on a polar plane with a provided advection field
  • Collisions
  • Collision operator in \((v_\parallel,\mu)\)
  • Coordinate transformation operators and tools
  • Coordinate transformation operators
    • Triangular Barycentric coordinates <-> Cartesian coordinates
    • Circular coordinates <-> Cartesian coordinates
    • Cylindrical coordinates <-> Cartesian coordinates
    • Tokamak-shaped Czarny coordinates <-> Cartesian coordinates
    • Toroidal coordinates -> Cylindrical coordinates
    • Discrete coordinates -> Cartesian coordinates
    • Identity transformation
    • Composite coordinate transformation
  • Tools to manage coordinate transformations by:
    • Getting the inverse Jacobian matrix at a given coordinate
    • Getting the inverse Jacobian matrix at the O-point (to provide explicit equations without an if)
    • Evaluate the metric tensor at a given coordinate
    • Map a vector from one vector space to another
  • Additional data types
  • DerivativeField to store a field and its boundary derivatives
  • VectorField
  • Tensor type and tools
    • Levi-Civita tensor
    • Identity tensor
    • Tensor multiplication operator
  • Interpolation operators
  • Lagrange interpolation
  • Spline interpolation
  • Polar spline evaluation
  • General Mathematical tools
  • Methods for calculating the L-norms
  • Derivative calculators
    • Finite differences method (with and without known boundary values)
    • Derivatives from 1D or 2D spline representations
    • Constant derivatives of a known value
  • Miscellaneous
    • sum
    • norm
    • modulo
    • pow for integer powers
    • factorial
    • min
    • max
    • determinant
    • inverse
    • scalar product
    • tensor product
  • Solvers for matrix equations with different sparsity patterns
  • Banded matrix
  • Batched CSR (compressed sparse row) matrix
  • Batched ELL matrix
  • Batched tridiagonal matrix
  • Matrix with dense corners
  • Matrix with dense bottom right-hand corner
  • Dense matrix
  • Positive-definite symmetric tridiagonal matrix
  • Periodic banded matrix (banded plus top-right and bottom-left corners)
  • MPI parallelisation tools
  • MPI all to all transposition
  • Solvers for partial differential equation (PDEs)
  • 1D Finite elements method (FEM)
  • 1D Fast Fourier transform (FFT)
  • Polar Poisson-like solver to solve \( - \nabla \cdot (\alpha \nabla \phi) + \beta \phi = \rho \) on a polar domain
  • Quadrature methods and tools
  • Definition of quadrature coefficients
    • Gauss-Legendre
    • Trapezoid
    • Simpson
    • Spline-based quadrature (with homogeneous Neumann boundary conditions or with an equal number of spline bases and interpolation points)
  • ND coefficients from multiple 1D quadrature coefficients
  • Definition of quadrature coefficients to calculate a volume on general coordinates
  • Timestepping methods
  • Euler
  • Crank-Nicolson
  • 2nd order Runge-Kutta (RK2)
  • 3rd order Runge-Kutta (RK3)
  • 4th order Runge-Kutta (RK4)

Notes

  • This is an early development release (v0.x); APIs may change