spectral_galerkin_heat.backends.base
====================================

.. py:module:: spectral_galerkin_heat.backends.base

.. autoapi-nested-parse::

   Container pairing an array module with its physics kernels.



Classes
-------

.. autoapisummary::

   spectral_galerkin_heat.backends.base.MathBackend


Functions
---------

.. autoapisummary::

   spectral_galerkin_heat.backends.base.to_host


Module Contents
---------------

.. py:function:: to_host(arr)

   Return *arr* as a :class:`numpy.ndarray` on the host.

   Calls ``numpy.asarray`` if *arr* is already a host array, or ``arr.get()``
   if it is a CuPy device array.


.. py:class:: MathBackend(name, xp, kernels)

   Hold the array module and physics kernels for one execution target.

   NumPy and CuPy share the same API, so the solver sets ``xp = backend.xp``
   and calls ``xp.zeros(...)`` directly. Host transfer is the only operation
   that differs between the two, which is why :meth:`to_numpy` is the only
   method.

   :param name: Backend identifier, ``"numpy"`` or ``"cupy"``.
   :type name: str
   :param xp: :mod:`numpy` or :mod:`cupy`.
   :type xp: module
   :param kernels: Kernel module matching *xp* (``spectral_cpu_kernels`` or
                   ``spectral_gpu_kernels``).
   :type kernels: module


   .. py:method:: to_numpy(arr)

      Return *arr* as a :class:`numpy.ndarray` on the host.

      Calls ``numpy.asarray`` if *arr* is already a host array, or
      ``arr.get()`` if it is a CuPy device array.

      :param arr: A NumPy or CuPy array.
      :type arr: ndarray

      :returns: The array on the host.
      :rtype: numpy.ndarray