CGAL’s exact kernel (using GMP rationals) is significantly slower than floating-point. For real-time Unity (60fps), you cannot run complex CGAL operations every frame. Run CGAL in a C# Job System or background Task . Show a loading spinner.
extern "C" __declspec(dllexport) float* ComputeConvexHull(float* points, int count, int* outCount) std::vector<Point_2> pts, result; for (int i = 0; i < count; i++) pts.push_back(Point_2(points[2 i], points[2 i+1])); CGAL::convex_hull_2(pts.begin(), pts.end(), std::back_inserter(result)); outCount = result.size(); float out = new float[2 * result.size()]; for (size_t i = 0; i < result.size(); i++) out[2 i] = CGAL::to_double(result[i].x()); out[2 i+1] = CGAL::to_double(result[i].y()); cgal unity
| Goal | Recommended Approach | |------|----------------------| | Real-time geometry processing (e.g., mesh simplification) | Native plugin | | One-time offline computation (level generation) | External process | | Complex CGAL features (Nef polygons, Minkowski sums) | External process | | Mobile / WebGL builds | External process (no native plugin support on those platforms) | CGAL’s exact kernel (using GMP rationals) is significantly
The hardest part of cgal unity integration is . CGAL uses exact rational numbers (e.g., CGAL::Exact_predicates_exact_constructions_kernel ). Unity uses 32-bit floats. Show a loading spinner