marray¶

class minkit.marray(array, dtype, length=None, backend=None)[source]¶

Wrapper over the arrays to do operations in CPU or GPU devices.

Parameters:	array (numpy.ndarray or reikna.cluda.api.Array) – original array. dtype (numpy.dtype) – data type. length (int or None) – (GPU only) actual size of the array. backend (Backend or None) – backend where to put the array.

Attributes Summary

`aop`	Associated object to do array operations.
`backend`	Backend interface.
`dtype`	Data type.
`length`	Length as a `numpy.int32` instance.
`shape`	Shape of the array.
`ua`	Underlying array.

Methods Summary

`as_ndarray`()	Return the underlying array as a `numpy.ndarray` instance.
`copy`()	Copy this array.
`from_ndarray`(a, backend)	Create this class from a `numpy.ndarray` instance.
`get`(index)	Get an element of the array.
`to_backend`(backend)	Send the array to the given backend.

Attributes Documentation

Methods Documentation

as_ndarray()[source]¶

Return the underlying array as a numpy.ndarray instance. If the underlying array is already of this type, no copy is done.

Returns:	underlying array as a `numpy.ndarray`.
Return type:	numpy.ndarray

Copy this array.

Returns:	Copy of this array.

classmethod from_ndarray(a, backend)[source]¶

Create this class from a numpy.ndarray instance.

Parameters:	a (numpy.ndarray) – array. backend (Backend) – backend where to create this array.
Returns:	Newly created array.

get(index)[source]¶

Get an element of the array. The output type depends on the type of array.

Parameters:	index (int) – index to access.
Returns:	value at the given index.
Return type:	float, int or bool
Raises:	IndexError – If there is an attempt to access an element out of range.

to_backend(backend)[source]¶

Send the array to the given backend.

Parameters:	backend (Backend) – backend where to transfer the array.
Returns:	this array on a new backend.
Return type:	marray