Update to nanoflann 1.6.1

- Add threading in kd three construction. It should improve runtime most
of the time.

Author: rjanvier

include/nn_search.hpp

@@ -36,7 +36,7 @@ static std::pair<nb::ndarray<nb::numpy, uint32_t, nb::ndim<2>>, nb::ndarray<nb::
 
     if (knn > data.rows()) { throw std::invalid_argument("knn size is greater than the data point cloud size"); }
 
-    kd_tree_t          kd_tree(3, data, 10);
+    kd_tree_t          kd_tree(3, data, 10, 0);
     const Eigen::Index n_points = query.rows();
     uint32_t*          indices  = new uint32_t[knn * n_points];
     nb::capsule        owner_indices(indices, [](void* p) noexcept { delete[] (uint32_t*)p; });
@@ -94,7 +94,7 @@ static std::pair<nb::ndarray<nb::numpy, int32_t, nb::ndim<2>>, nb::ndarray<nb::n
         throw std::invalid_argument("max knn size is greater than the data point cloud size");
     }
 
-    kd_tree_t    kd_tree(3, data, 10);
+    kd_tree_t    kd_tree(3, data, 10, 0);
     const real_t sq_search_radius = search_radius * search_radius;
 
     const Eigen::Index n_points = query.rows();

include/pgeof.hpp

@@ -4,7 +4,6 @@
 #include <nanobind/nanobind.h>
 #include <nanobind/ndarray.h>
 
-
 #include <algorithm>
 #include <cmath>
 #include <cstdio>
@@ -74,9 +73,10 @@ static inline void flush() { std::cout << std::endl; };
  * @return the geometric features associated with each point's neighborhood in a (num_points, features_count) nd::array.
  */
 template <typename real_t = float, const size_t feature_count = 11>
-static nb::ndarray<nb::numpy, real_t, nb::shape<-1, static_cast<nb::ssize_t>(feature_count)>> compute_geometric_features(
-    RefCloud<real_t> xyz, nb::ndarray<const uint32_t, nb::ndim<1>> nn, nb::ndarray<const uint32_t, nb::ndim<1>> nn_ptr,
-    const size_t k_min, const bool verbose)
+static nb::ndarray<nb::numpy, real_t, nb::shape<-1, static_cast<nb::ssize_t>(feature_count)>>
+    compute_geometric_features(
+        RefCloud<real_t> xyz, nb::ndarray<const uint32_t, nb::ndim<1>> nn,
+        nb::ndarray<const uint32_t, nb::ndim<1>> nn_ptr, const size_t k_min, const bool verbose)
 {
     if (k_min < 1) { throw std::invalid_argument("k_min should be > 1"); }
     // Each point can be treated in parallel
@@ -330,7 +330,7 @@ static nb::ndarray<nb::numpy, real_t, nb::shape<-1, -1>> compute_geometric_featu
     using kd_tree_t = nanoflann::KDTreeEigenMatrixAdaptor<RefCloud<real_t>, 3, nanoflann::metric_L2_Simple>;
     // TODO: where knn < num of points
 
-    kd_tree_t          kd_tree(3, xyz, 10);
+    kd_tree_t          kd_tree(3, xyz, 10, 0);
     const size_t       feature_count    = selected_features.size();
     const Eigen::Index n_points         = xyz.rows();
     real_t             sq_search_radius = search_radius * search_radius;

src/pgeof_ext.cpp

@@ -1,7 +1,6 @@
 
 #include <nanobind/nanobind.h>
 #include <nanobind/ndarray.h>
-#include <nanobind/stl/pair.h>
 #include <nanobind/stl/vector.h>
 
 #include "nn_search.hpp"
@@ -176,4 +175,4 @@ NB_MODULE(pgeof_ext, m)
             :param selected_features: List of selected features. See EFeatureID
             :return: Geometric features associated with each point's neighborhood in a (num_points, features_count) numpy array.
         )");
-}
+}