Release v0.8.5

Features

• GPU In-Place Laplacian Mesh Smoothing (conquer3d.data_structure):

  • Implemented smooth(iterations=10, damping=0.5, mode=1) on TriangleMesh performing explicit Euler Laplacian smoothing on GPU.
  • Accelerated via edge-parallel CUDA kernels (assemble_edge_weighted_laplacian, smooth_laplacian).
  • Supports uniform weights (mode 0, equilateral relaxation) and cotangent weights (mode 1, shape and feature preservation).
  • Automatic boundary vertex pinning to preserve open mesh border geometry.
  • Coherent geometric cache invalidation (normals, Voronoi areas, curvature, BVH, flood fills) on vertex updates.

• Multi-Mode Discrete Mean Curvature & Laplacian Fields (conquer3d.data_structure):

  • Enhanced get_mean_curvature(mode) with 3 discrete modes:
    • Mode 0: Signed scalar mean curvature $H$ (sphere with outward normals reports $H = +1/R$).
    • Mode 1: Vector curvature normal $\Delta_{LB} x = -2H\hat{n}$.
    • Mode 2: Absolute discrete mean curvature $\tfrac{1}{2}|\Delta_{LB} x|$ (Botsch et al., Polygon Mesh Processing, Eq. 3.13).
  • Enhanced compute_laplacian(mode) supporting uniform (mode 0), cotangent area-normalised (mode 1), and cotangent weight-normalised (mode 2, dimensionless scale-invariant convex combination) vector fields.
  • Enhanced get_principal_curvatures() and refined get_gaussian_curvature() closed-ring documentation.

• Conda Toolchain & Compiler Pinning (install.sh):

  • Added repository-level install.sh script automating toolchain discovery, CUDA architecture list configuration, and compiler resolution for the active conda environment.

Refactors & Build

• Build System: Bumped package version to 0.8.5 in pyproject.toml.
• PyBind11 Extension: Exposed smooth, multi-mode get_mean_curvature, and compute_laplacian over TriangleMesh in conquer3d/csrc/binds/data_structure/triangle_mesh.cpp.

Examples

• In-place GPU mesh fairing, feature-preserving smoothing, and multi-mode curvature analysis on high-resolution triangle meshes.