VTK-m
2.2
|
Assigner for PartitionedDataSet partitions. More...
#include <AssignerPartitionedDataSet.h>
Public Member Functions | |
AssignerPartitionedDataSet (const vtkm::cont::PartitionedDataSet &pds) | |
Initialize the assigner using a partitioned dataset. More... | |
AssignerPartitionedDataSet (vtkm::Id num_partitions) | |
~AssignerPartitionedDataSet () override | |
void | local_gids (int rank, std::vector< int > &gids) const override |
int | rank (int gid) const override |
Private Attributes | |
std::vector< vtkm::Id > | IScanPartitionCounts |
Assigner for PartitionedDataSet partitions.
AssignerPartitionedDataSet
is a vtkmdiy::StaticAssigner
implementation that uses PartitionedDataSet
's partition distribution to build global-id/rank associations needed for several diy
operations. It uses a contiguous assignment strategy to map partitions to global ids, i.e. partitions on rank 0 come first, then rank 1, etc. Any rank may have 0 partitions.
AssignerPartitionedDataSet uses collectives in the constructor hence it is essential it gets created on all ranks irrespective of whether the rank has any partitions.
vtkm::cont::AssignerPartitionedDataSet::AssignerPartitionedDataSet | ( | const vtkm::cont::PartitionedDataSet & | pds | ) |
Initialize the assigner using a partitioned dataset.
This may initialize collective operations to populate the assigner with information about partitions on all ranks.
vtkm::cont::AssignerPartitionedDataSet::AssignerPartitionedDataSet | ( | vtkm::Id | num_partitions | ) |
|
override |
|
override |
vtkmdiy::Assigner API implementation.
|
override |
vtkmdiy::Assigner API implementation.
|
private |