Skip to content

CHANGELOG.md

Latest commit

 

History

History
95 lines (92 loc) · 9.76 KB

CHANGELOG.md

File metadata and controls

95 lines (92 loc) · 9.76 KB
Raw

Version 0.3.0 (unreleased)

Features

  • Add slip and inletOutlet volume boundary conditions #565
  • Add corrected and limited-corrected face-normal gradient schemes #514
  • Add experimental support for cell based assembly strategies #471,#517, #473
  • Add experimental support for COO and CSR Matrices #486
  • Add experimental umpire support #455
  • Add experimental MPI support #519, #520, #512, #525, #531, #522
  • Add python bindings via nanobind #382
  • Correct the Ginkgo version to 2.0.0 (unreleased) #493
  • Add tensor and symmTensor primitives, Su type sourceTerm and passing of fields to BCs #428
  • Add uniform mesh generator #475
  • Add distributed (processor-boundary) support: exact processor-face geometry and non-orthogonal corrected/limited snGrad correction, plus a one-time set() / per-iteration update() boundary-condition interface #528
  • Add DSL expression optimizer infrastructure for operator fusing #452
  • Add L1-norm stopping criterion #538

Misc

  • Added linearUpwind scheme #548
  • Include RHS / explicit source assembly #539
  • Build the distributed off-diagonal matrix with local row indices and widen non-local columns on the executor, removing host round-trips in the Ginkgo distributed matrix assembly
  • Remove stencilDataBase #416
  • Added backward ddtScheme and scheme selection mechanism #419
  • Support AssemblyType != ValueType for Vec3 fields #536
  • Bump Kokkos to 5.0.2 #476
  • Bump Ginkgo to 1.11 #409

Fixes

  • Fix distributed processor-face correctness: multi-patch (scotch) halo exchange, ddtFluxCorr proc-face correction, processor BC on coupled patches, and row-sorted non-local COO for correct CUDA distributed apply #528

Version 0.2.0 (2025/12/03)

Features

  • Add basic logging functionality #396
  • Improved support for assembling and solving expression of Field. #370,#374, #379

Refactoring

  • Vector split into an generic container class, Array, and the Vector class specialized for as a mathematical operations. #318
  • General refactor of span and spans to view and views. #311
  • View replaces std::span and Field uses view() rather than span() and returns a view #298

Implicit Capabilities

  • Implicit BCs and Laplacian operator #262
  • Support for implicit operators in the DSL #246

Linear Algebra Capabilities

  • Adds a minimal implementation linear algebra functionality #219
  • Add Ginkgo solver interface and config solver support #250, [#272] (#272)
  • Add Ginkgo SYCL support #384

Misc

  • Use current Ginkgo develop version #362
  • Removed subscript operators from Field for improved safety. #225
  • Templated Expression and Operator on ValueType #268
  • Ability to solve poisson equation similar required in the PISO algorithm #267
  • Add basic vectorField implementation #260
  • Add temporal operators inside expressions and the ability to solve linear system #259
  • Add segmentedField to represent vector of vectors #202
  • Add courant number calculation based on parallelFor #224
  • Add license automation #343

Fixes

Version 0.1.0

  • Improve build with MSVC and Clang on Windows #163
  • Add document based database #155
  • Implement a DSL interface #121
  • Implement a basic DSL interface #102
  • Faster project configuration #179
  • Improved error handling and addition of tokenList and Input #134
  • Disable span from temporary objects and simplification related to fields #139
  • Added launch json to debug unit test in vscode #135
  • Add a basic implementation of operators #100
  • Changes executor meanings, the CPUExecutor was renamed to SerialExecutor and the OMPExecutor was renamed to CPUExecutor. PR #120
  • Minor cleanup of MPI operator names, added vector version of allReduce, and updates mpi and thread support operations. PR #105
  • Implementation of surface BCs and improvements to volume BCs PR #104
  • Improvements to vector and dictionary; fix of the hanging issue with cuda PR #108
  • Field update: A number of operators and functions added, some bug fixes, and improved testing. PR #34
  • Provide wrapper for Catch2 that handles tests running with MPI #82
  • Added parallelFor and parallelReduce #62
  • Add runtimeselectionFactory to simplify the usage of the runtime selection mechanism #95
  • Implementation of boundary conditions and finite volume fields #74
  • Initial MPI support and infrastructure added, specifically focusing on not blocking communication #79.
  • Inject the boundaryconditionsstrategy with compile discovery into the boundary condition #83
  • Add compile time derived class discovery #80
  • Basic introduction macros for console printouts to screen (Debug, Info, and Error), as well as assert checking and throwing #63
  • Added boundaryField and domainField #54
  • Basic implementation of an unstructuredMesh #53
  • Enforce 80 characters limit #59
  • Improved and documented the workflow with vscode #58
  • Added dictionary class #52
  • Applied changes proposed by clang-tidy #57
  • Compiles with cuda support and gcc: Kokkos is now downloaded with CPM #50
  • Use clang-tidy check to enforce consistent naming scheme #46
  • Added equality operators for executors #45
  • Added pre-commit hooks, see PR #38
  • Added CPM package manager PR #37
  • Corrected/streamlined includes for Field.hpp. PR #33
  • Add basic field implementation, PR #17, PR #22
  • Check if Kokkos present or build it PR #10
  • Build documentation via Sphinx and Doxygen PR #11, PR #22
  • Add basic cmake setup PR #4, PR #21