yarrow.array.cupy

A CuPy array backend.

Danger

Experimental Module

This code is not thoroughly tested. It’s included here as a proof-of-concept for GPU acceleration.

yarrow.array.cupy.array(*args, **kwargs)

yarrow.array.cupy.max(*args, **kwargs)

yarrow.array.cupy.arange(*args, **kwargs)

yarrow.array.cupy.all(*args, **kwargs)

yarrow.array.cupy.zeros(*args, **kwargs)

yarrow.array.cupy.ones(*args, **kwargs)

yarrow.array.cupy.cumsum(*args, **kwargs)

yarrow.array.cupy.sum(*args, **kwargs)

yarrow.array.cupy.repeat(*args, **kwargs)

yarrow.array.cupy.concatenate(*args, **kwargs)

yarrow.array.cupy.connected_components(source, target, n, dtype='int64')

Compute the connected components of a graph with N nodes, whose edges are encoded as a pair of arrays (source, target) such that the edges of the graph are source[i] → target[i].

Parameters:

source (array) – A length-N array with elements in the set {0 .. N - 1}.
target (array) – A length-N array with elements in the set {0 .. N - 1}.

Returns:

A pair (c, cc_ix) of the number of connected components c and a mapping from nodes to connected components cc_ix.

Return type:

(int, array)

yarrow.array.cupy.argsort(x)

yarrow.array.cupy.segmented_arange(x)

Given an array of sizes, [x₀, x₁, ...] output an array equal to the concatenation concatenate([arange(x₀), arange(x₁), ...])

>>> FiniteFunction._Array.segmented_arange([5, 2, 3, 1])
array([0, 1, 2, 3, 4, 0, 1, 0, 1, 2, 0])

Params:: x: An array of the sizes of each “segment” of the output

Returns:: segmented array with segment i equal to arange(i).
Return type:: array

yarrow.array.cupy.bincount(x, *args, **kwargs)

yarrow.array.cupy.full(n, x, *args, **kwargs)