Skip to content

CHANGELOG.md

Latest commit

 

History

History
376 lines (299 loc) · 17.3 KB

CHANGELOG.md

File metadata and controls

376 lines (299 loc) · 17.3 KB
Raw

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

Fixed

Changed

Deprecated

Removed

[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