DNDSR v0.1.0

🚀 DNDSR v0.1.0

Major milestone release — 701 files changed, ~110k insertions across 241 commits.
📖 Documentation: cfdlab-thu.github.io/DNDSR

✈️ New Solver Features

• 🔄 p-Multigrid (pMG): HM3 revamp with MG support, tpMG, incFScale for MG lower levels, positivity-preserving limiters in LimiterUGrad
• 🌀 SA-IDDES: DES → DDES → IDDES progression, psi term fixes, rotation correction variants, ft2 toggle
• 🧮 k-omega RANS: two-equation model with euler2EQ / euler2EQ3D executables
• ⏱️ ODE integrators: ESDIRK2, ESDIRK3, Trapezoidal; HM3 U2R2/U2R1/U3R1 modes
• 🔁 Axisymmetric flow with wedge-shaped metric in reconstruction
• 🧊 Isothermal wall BC (BCWallIsothermal)
• 📈 incFScale (incremental flux scaling) integrated into entropy fix
• ⚡ Roe_M8 flux; HLLE+ (experimental)
• 🎛️ source2nd, mergeMultiResidual, normOrd, restartOutAtInit, resBaseType options
• 🕸️ Overset grid exploration: hole creation, distance map, cell-cell connectivity (2D demo)

🐍 Python Modules (new)

• 📦 pybind11 bindings for DNDS, Geom, CFV, EulerP modules
• 📊 Python ModelEvaluator for CFV — Fourier dissipation-dispersion analysis
• 🗺️ create_mesh_from_CGNS, boundary mesh extraction, VTK output
• 🔗 ArrayPair, ArrayEigenMatrix/Vector/Batch exports
• 🛡️ py::classh holder for safe Python↔C++ ownership
• 🧩 Auto stubgen on install; DNDSR.XXX import structure

🏗️ Build & Packaging

• 🔧 Modular CMake with presets (release-test, debug, cuda)
• 🏷️ Centralized version from VERSION file + git describe (PEP 440)
• 📦 Python package via scikit-build-core with editable installs
• 📋 JSON Schema generation and config validation (--emit-schema CLI)
• ⚙️ DNDS_DECLARE_CONFIG registry for typed config sections
• 🧵 OMP support in ILU; DNDS_PYBIND11_NO_LTO option
• 🍎 macOS compilation fixes (fmtlib workaround)

🔬 CFD Module Refactoring

• Euler: RANSModelTraits, BC strategy pattern, shared gas physics, centralized Roe preamble / magic numbers / model constants
• CFV: 8-phase refactoring — safety, dedup, boundaries, functionals, data decomposition, DeviceTransferable, DOF factory, cleanup
• Geom: modular element traits with codegen shape functions, distributed mesh read with ParMetis repartition, extracted helpers (CGNS read, face interpolation, cell reorder, VTK output)

⚙️ Core Infrastructure

• 💾 Redistributable Array serialization with HDF5 checkpoint/restart
• 🧰 ObjectNaming, make_ssp, ArrayPair helpers
• 🎮 CUDA: device storage, device views for all derived array types, block-sparse MatVec benchmark, SoA layout tests
• 🚄 EulerP module with CUDA-optimized evaluator
• 🗂️ Mesh: cell2cellOrig / node2nodeOrig / bnd2bndOrig index tracking, node wall distance, boundary ghost via node2bnd

✅ Testing (new)

• 🧪 C++ (doctest): CFV limiters/reconstruction (43 tests), Euler gas/thermo/Riemann/RANS (59 tests), Solver ODE/linear/direct (24 tests), Geom elements/quadrature/mesh pipeline
• 🐍 Python (pytest): FV/VR correctness (32 tests), MPI communication
• 🔁 CTest registration at np=1, 2, 4 for all MPI-aware tests

📖 Documentation System (new)

• 🌐 Sphinx + Breathe site with Doxygen HTML embedded at /doxygen/
• 🔀 Shared Markdown for both Doxygen and Sphinx (doxygen_compat.py translator)
• ⚡ Incremental builds — no-op <1s, md-only ~10s, full ~2.5min
• 📚 Guides: building, style, arrays, serialization, geometry, examples, doc authoring
• 🏛️ Architecture pages, theory pages, test documentation
• 📐 C++ API with class diagrams (Breathe + Graphviz); Python API via autodoc
• 🤖 GitHub Actions workflow for manual Pages deployment with 3-layer caching
• 🎨 sphinxawesome-theme with colorful syntax highlighting

🐛 Notable Bug Fixes

• 🔒 MPI deadlock from collective globalSize() — made non-collective
• 🔗 Serializer false shared-ptr dedup from address reuse
• 🔄 InterpolateFace wrong face identity on double-periodic BCs
• 💥 Inner partitioning segfault in Metis reordering
• 📋 ArrayTransformer clone error
• 🔁 Periodic boundary RecoverCell2CellAndBnd2Cell on arbitrary partitions
• 💾 CSR pRowStart not written in serialization
• ⏱️ Multigrid fdtau/frhs ordering corrupting lambda estimates
• 🎮 EulerP accidental to_device in face buffer creation
• 🔧 CUDA thrust call fixes with CMAKE_CUDA_ARCHITECTURE=native