HMLP: High-performance Machine Learning Primitives
|
Data and setup that are shared with all nodes. More...
#include <tree_mpi.hpp>
Public Types | |
typedef DATATYPE | T |
Public Member Functions | |
vector< size_t > | ContainAny (vector< size_t > &queries, size_t target) |
Check if this node contain any query using morton. Notice that queries[] contains gids; thus, morton[] needs to be accessed using gids. More... | |
Public Attributes | |
size_t | m |
size_t | max_depth = 15 |
DistData< STAR, CBLK, pair< T, size_t > > * | NN_cblk = NULL |
DistData< STAR, CIDS, pair< T, size_t > > * | NN = NULL |
vector< size_t > | morton |
SPLITTER | splitter |
Data and setup that are shared with all nodes.
end class DistSplitTask
|
inline |
Check if this node contain any query using morton. Notice that queries[] contains gids; thus, morton[] needs to be accessed using gids.
notice that setup->morton only contains local morton ids
size_t hmlp::mpitree::Setup< SPLITTER, DATATYPE >::m |
end ContainAny() maximum leaf node size
size_t hmlp::mpitree::Setup< SPLITTER, DATATYPE >::max_depth = 15 |
by default we use 4 bits = 0-15 levels
vector<size_t> hmlp::mpitree::Setup< SPLITTER, DATATYPE >::morton |
morton ids
DistData<STAR, CBLK, pair<T, size_t> >* hmlp::mpitree::Setup< SPLITTER, DATATYPE >::NN_cblk = NULL |
coordinates (accessed with gids) neighbors<distance, gid> (accessed with gids)
SPLITTER hmlp::mpitree::Setup< SPLITTER, DATATYPE >::splitter |
tree splitter