partition.h File Reference

Cell partitioning for distributed runs. More...

#include <array>
#include <cstdint>
#include <vector>
#include "common_typedef.h"

Go to the source code of this file.

Functions
uint64_t	hilbert_key (const std::array< double, N_DIM > &x, const std::array< double, N_DIM > &lo, const std::array< double, N_DIM > &hi)
	Position along the Hilbert curve of a point in the box [lo, hi] (any dimension; 63 / N_DIM bits per axis).
std::vector< int >	partition_hilbert (const DistributedMesh &mesh, int n_parts)
	Owner rank of every block cell of a distributed mesh: cells sorted along the Hilbert curve of their vertex averages (ties by global id) by a distributed sample sort, split into n_parts contiguous pieces of nearly equal size (collective).
std::array< int, N_DIM >	grid_shape (int n_parts, const std::array< double, N_DIM > &extent)
	Parts per axis of the grid of n_parts blocks of a box with these extents that has the least total area between blocks.
std::vector< int >	partition_multijagged (const DistributedMesh &mesh, int n_parts)
	Owner rank of every block cell of a distributed mesh by multi-jagged coordinate partitioning (collective): the cells, keyed by the coordinates of their vertex averages (ties by global id), are split along x into equal slabs, each slab along y, and so on, with the number of pieces per axis from grid_shape() on their bounding box. A box of cuboid cells gives cuboid parts.
std::vector< int >	partition_graph (const DistributedMesh &mesh, int n_parts)
	Owner rank of every block cell of a distributed mesh from dKaMinPar on its dual graph (collective). Requires Mallard_ENABLE_KAMINPAR.
bool	have_graph_partitioner ()
	Whether this build has a graph partitioner.

Detailed Description

Cell partitioning for distributed runs.

Author

Matthew Bonanni (mbona.nosp@m.nni0.nosp@m.01@gm.nosp@m.ail..nosp@m.com)

Version

0.3

Date

2026-10-02

Copyright

Copyright (c) 2026 Matthew Bonanni

Function Documentation

grid_shape()

std::array< int, N_DIM > grid_shape	(	int	n_parts,
		const std::array< double, N_DIM > &	extent
	)

Parts per axis of the grid of n_parts blocks of a box with these extents that has the least total area between blocks.

have_graph_partitioner()

bool have_graph_partitioner

(

)

Whether this build has a graph partitioner.

hilbert_key()

uint64_t hilbert_key	(	const std::array< double, N_DIM > &	x,
		const std::array< double, N_DIM > &	lo,
		const std::array< double, N_DIM > &	hi
	)

Position along the Hilbert curve of a point in the box [lo, hi] (any dimension; 63 / N_DIM bits per axis).

partition_graph()

std::vector< int > partition_graph	(	const DistributedMesh &	mesh,
		int	n_parts
	)

Owner rank of every block cell of a distributed mesh from dKaMinPar on its dual graph (collective). Requires Mallard_ENABLE_KAMINPAR.

partition_hilbert()

std::vector< int > partition_hilbert	(	const DistributedMesh &	mesh,
		int	n_parts
	)

Owner rank of every block cell of a distributed mesh: cells sorted along the Hilbert curve of their vertex averages (ties by global id) by a distributed sample sort, split into n_parts contiguous pieces of nearly equal size (collective).

partition_multijagged()

std::vector< int > partition_multijagged	(	const DistributedMesh &	mesh,
		int	n_parts
	)

Owner rank of every block cell of a distributed mesh by multi-jagged coordinate partitioning (collective): the cells, keyed by the coordinates of their vertex averages (ties by global id), are split along x into equal slabs, each slab along y, and so on, with the number of pieces per axis from grid_shape() on their bounding box. A box of cuboid cells gives cuboid parts.