This module contains various auxiliary functions which are used throughout the library.


Returns the minimal number of the form 2**m such that it is greater or equal to n.

reikna.helpers.default_strides(shape, itemsize)

Return the default strides (corresponding to a contiguous array) for an array of shape shape and elements of size itemsize bytes.

reikna.helpers.factors(num, limit=None)

Returns the list of pairs (factor, num/factor) for all factors of num (including 1 and num), sorted by factor. If limit is set, only pairs with factor <= limit are returned.

class reikna.helpers.ignore_integer_overflow

Context manager for ignoring integer overflow in numpy operations on scalars (not ignored by default because of a bug in numpy).


Integer-valued logarigthm with base 2. If n is not a power of 2, the result is rounded to the smallest number.

reikna.helpers.make_axes_innermost(ndim, axes)

Given the total number of array axes and a list of axes in this range, produce a transposition plan (suitable e.g. for numpy.transpose()) that will move make the given axes innermost (in the order they’re given). Returns the transposition plan, and the plan to transpose the resulting array back to the original axes order.

reikna.helpers.min_blocks(length, block)

Returns minimum number of blocks with length block necessary to cover the array with length length.

reikna.helpers.min_buffer_size(shape, itemsize, strides=None, offset=0)

Return the minimum memory buffer size (in bytes) that can fit an array with given parameters, starting at an offset bytes from the beginning of the buffer.

reikna.helpers.normalize_axes(ndim, axes)

Transform an iterable of array axes (which can be negative) or a single axis into a tuple of non-negative axes.

reikna.helpers.padded_buffer_parameters(shape, itemsize, pad=0)

For an array of shape shape, padded from all sizes with pad elements, return a tuple of (strides, offset, size (in bytes) of the required memory buffer), which would have to be requested when allocating such an array.


Returns the product of elements in the iterable seq.

reikna.helpers.template_def(signature, code)

Returns a Mako template with the given signature.

Parameters:signature – a list of postitional argument names, or a Signature object from funcsigs module.
Code:a body of the template.

Returns the Mako template object created from the file which has the same name as filename and the extension .mako. Typically used in computation modules as template_for(__filename__).


Creates a Mako template object from a given string. If template already has render() method, does nothing.


If seq_or_elem is a sequence, converts it to a tuple, otherwise returns a tuple with a single element seq_or_elem.