afe.backends.ev.sima_ev_kernels
===============================

.. py:module:: afe.backends.ev.sima_ev_kernels

.. autoapi-nested-parse::

   SiMa pre-built OpenVX/OpenCL kernels for EV

   Each kernel has an equivalent EV transform implemented in the ev-transforms repo,
   and the mapping from EV transform to SiMa prebuilt kernel is documented here.

   Steps to add a new prebuilt kernel:
   1) implement a corresponding EV transform in ev-transform repo and verifies
      function equivalence to the prebuilt kernel;
   2) add a corresponding transform class in afe/apis/transform.py;
   3) add operator converter mapping to _OPERATOR_CONVERTERS in afe/backends/mpk/operator.py;
   4) update this file.



Attributes
----------

.. autoapisummary::

   afe.backends.ev.sima_ev_kernels.SIMA_EV_KERNELS


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

.. py:data:: SIMA_EV_KERNELS
   :type:  Mapping[Type[afe.ir.operations.AwesomeOperation], str]