Decomposition Namespace Reference

Functions
	equidistant_splitting (Tree peano_tree, int number_of_subtrees, verbose=False)
	quantised_splitting (Tree peano_tree, int number_of_subtrees, int dimension, verbose=False)
	Quantised splitting.

Function Documentation

equidistant_splitting()

Decomposition.equidistant_splitting	(	Tree	peano_tree,
		int	number_of_subtrees,
			verbose = False
	)

Definition at line 8 of file Decomposition.py.

Referenced by test_hardcoded_output._split_regular_grid(), 35_offline-load-balancing.offline_decomposition(), 35_offline-load-balancing-split-modes.offline_decomposition(), 35_offline-load-balancing-plan-vs-run.offline_plan(), and 35_offline-load-balancing-quantised.owner_of_cells().

Here is the caller graph for this function:

quantised_splitting()

Decomposition.quantised_splitting	(	Tree	peano_tree,
		int	number_of_subtrees,
		int	dimension,
			verbose = False
	)

Quantised splitting.

This splitting starts from a perfect splitting as we find it in equidistant_splitting(), but then it picks partitions of the type \( 1 \\cdot 3^{kd}, 2 \\cdot 3^{kd}, ..., (3^d-1) \\cdot 3^{kd}, 1 \\cdot 3^{(k+1)d}, 2 \\cdot 3^{(k+1)d}, ... \)

Implementation

• Determine optimal number of splittings
• Introduce a residual counter. As the number of leaves typically does not exactly fit to number_of_subtrees, this is the remainder of the modulo division.
• Identify smallest chunk from sequence above that literally just fits.
• Add the cells that should have been added to this chunk (due to the optimal decomposition) but cannot be added (due to the quantisation from the sequence above) to the residual.
• Next step, increase the optimum threshold by the residual.

Definition at line 45 of file Decomposition.py.

Referenced by 35_offline-load-balancing-quantised.owner_of_cells().

Here is the caller graph for this function: