APIs related to spatial indexing.
More...
|
template<class PointIterator , class T = typename cuspatial::iterator_value_type<PointIterator>> |
std::pair< rmm::device_uvector< uint32_t >, point_quadtree > | cuspatial::quadtree_on_points (PointIterator points_first, PointIterator points_last, vec_2d< T > vertex_1, vec_2d< T > vertex_2, T scale, int8_t max_depth, int32_t max_size, rmm::cuda_stream_view stream=rmm::cuda_stream_default, rmm::mr::device_memory_resource *mr=rmm::mr::get_current_device_resource()) |
| Construct a quadtree structure from points.
|
|
std::pair< std::unique_ptr< cudf::column >, std::unique_ptr< cudf::table > > | cuspatial::quadtree_on_points (cudf::column_view const &x, cudf::column_view const &y, double x_min, double x_max, double y_min, double y_max, double scale, int8_t max_depth, cudf::size_type max_size, rmm::mr::device_memory_resource *mr=rmm::mr::get_current_device_resource()) |
| Construct a quadtree structure from points.
|
|
APIs related to spatial indexing.
◆ quadtree_on_points() [1/2]
std::pair< std::unique_ptr< cudf::column >, std::unique_ptr< cudf::table > > cuspatial::quadtree_on_points |
( |
cudf::column_view const & | x, |
|
|
cudf::column_view const & | y, |
|
|
double | x_min, |
|
|
double | x_max, |
|
|
double | y_min, |
|
|
double | y_max, |
|
|
double | scale, |
|
|
int8_t | max_depth, |
|
|
cudf::size_type | max_size, |
|
|
rmm::mr::device_memory_resource * | mr = rmm::mr::get_current_device_resource() ) |
Construct a quadtree structure from points.
- See also
- http://www.adms-conf.org/2019-camera-ready/zhang_adms19.pdf for details.
- Note
scale
is applied to (x - x_min) and (y - y_min) to convert coordinates into a Morton code in 2D space.
-
max_depth
should be less than 16, since Morton codes are represented as uint32_t
. The eventual number of levels may be less than max_depth
if the number of points is small or max_size
is large.
-
All intermediate quadtree nodes will have fewer than
max_size
number of points. Leaf nodes are permitted (but not guaranteed) to have >= max_size
number of points.
- Parameters
-
x | Column of x-coordinates for each point. |
y | Column of y-coordinates for each point. |
x_min | The lower-left x-coordinate of the area of interest bounding box. |
x_max | The upper-right x-coordinate of the area of interest bounding box. |
y_min | The lower-left y-coordinate of the area of interest bounding box. |
y_max | The upper-right y-coordinate of the area of interest bounding box. |
scale | Scale to apply to each x and y distance from x_min and y_min. |
max_depth | Maximum quadtree depth. |
max_size | Maximum number of points allowed in a node before it's split into 4 leaf nodes. |
mr | The optional resource to use for output device memory allocations. |
- Exceptions
-
- Returns
- Pair of INT32 column of sorted keys to point indices, and cudf table with five columns for a complete quadtree: key - UINT32 column of quad node keys level - UINT8 column of quadtree levels is_internal_node - BOOL8 column indicating whether the node is a quad (true) or leaf (false) length - UINT32 column for the number of child nodes (if is_internal_node), or number of points offset - UINT32 column for the first child position (if is_internal_node), or first point position
◆ quadtree_on_points() [2/2]
template<class PointIterator , class T = typename cuspatial::iterator_value_type<PointIterator>>
std::pair< rmm::device_uvector< uint32_t >, point_quadtree > cuspatial::quadtree_on_points |
( |
PointIterator | points_first, |
|
|
PointIterator | points_last, |
|
|
vec_2d< T > | vertex_1, |
|
|
vec_2d< T > | vertex_2, |
|
|
T | scale, |
|
|
int8_t | max_depth, |
|
|
int32_t | max_size, |
|
|
rmm::cuda_stream_view | stream = rmm::cuda_stream_default, |
|
|
rmm::mr::device_memory_resource * | mr = rmm::mr::get_current_device_resource() ) |
Construct a quadtree structure from points.
- See also
- http://www.adms-conf.org/2019-camera-ready/zhang_adms19.pdf for details.
- Note
- 2D coordinates are converted into a 1D Morton code by dividing each x/y by the
scale
: ((x - min_x) / scale
and (y - min_y) / scale
).
-
max_depth
should be less than 16, since Morton codes are represented as uint32_t
. The eventual number of levels may be less than max_depth
if the number of points is small or max_size
is large.
-
All intermediate quadtree nodes will have fewer than
max_size
number of points. Leaf nodes are permitted (but not guaranteed) to have >= max_size
number of points.
- Template Parameters
-
PointIterator | Iterator over x/y points. Must meet the requirements of LegacyRandomAccessIterator and be device-accessible. |
T | the floating-point coordinate value type of the input x/y points. |
- Parameters
-
points_first | Iterator to the beginning of the range of (x, y) points. |
points_last | Iterator to the end of the range of (x, y) points. |
vertex_1 | Vertex of the area of interest bounding box |
vertex_2 | Vertex of the area of interest bounding box opposite vertex_1 |
scale | Scale to apply to each x and y distance from min.x and min.y. |
max_depth | Maximum quadtree depth. |
max_size | Maximum number of points allowed in a node before it's split into 4 leaf nodes. |
stream | The CUDA stream on which to perform computations |
mr | The optional resource to use for output device memory allocations. |
- Returns
- Pair of UINT32 column of sorted keys to point indices and a point_quadtree
- Precondition
- Point iterators must have the same
vec_2d
value type, with the same underlying floating-point coordinate type (e.g. cuspatial::vec_2d<float>
).