tilelang.language.loop
======================

.. py:module:: tilelang.language.loop

.. autoapi-nested-parse::

   The language interface for tl programs.



Functions
---------

.. autoapisummary::

   tilelang.language.loop.Parallel
   tilelang.language.loop.Persistent
   tilelang.language.loop.Pipelined
   tilelang.language.loop.serial


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

.. py:function:: Parallel(*extents, coalesced_width = None)

   Tools to construct nested parallel for loop.
      This can be used to create element-wise tensor expression.

   :param extents: The extents of the iteration.
   :type extents: PrimExpr
   :param coalesced_width: The coalesced width of the parallel loop.
   :type coalesced_width: Optional[int]

   :returns: **res** -- The ForFrame.
   :rtype: frame.ForFrame


.. py:function:: Persistent(domain, wave_size, index, group_size = 8)

   Tools to construct persistent for loop.

   :param domain: The list of dominators.
   :type domain: List[tir.PrimExpr]
   :param wave_size: The wave size.
   :type wave_size: int
   :param index: The tile index in one wave.
   :type index: int
   :param group_size: The group size.
   :type group_size: tir.PrimExpr


.. py:function:: Pipelined(start, stop = None, num_stages = 0, order = None, stage = None, sync = None, group = None)

   Tools to construct pipelined for loop.

   :param start: The minimum value of iteration.
   :type start: PrimExpr
   :param stop: The maximum value of iteration.
   :type stop: PrimExpr
   :param num_stages: The max number of buffer used between pipeline producers and consumers.
                      if num_stages is 0, pipeline will not be enabled.
   :type num_stages: int

   :returns: **res** -- The ForFrame.
   :rtype: frame.ForFrame


.. py:function:: serial(start, stop = None, step = None, *, annotations = None)