tilelang.language.gemm_op
=========================

.. py:module:: tilelang.language.gemm_op

.. autoapi-nested-parse::

   GEMM (General Matrix Multiplication) operators exposed on the TileLang language surface.



Functions
---------

.. autoapisummary::

   tilelang.language.gemm_op.gemm_v1
   tilelang.language.gemm_op.gemm_v2
   tilelang.language.gemm_op.gemm


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

.. py:function:: gemm_v1(A, B, C, transpose_A = False, transpose_B = False, policy = GemmWarpPolicy.Square, clear_accum = False, k_pack = 1, wg_wait = 0, mbar = None)

   GEMM v1: use op tl.gemm.


.. py:function:: gemm_v2(A, B, C, transpose_A = False, transpose_B = False, policy = GemmWarpPolicy.Square, clear_accum = False, k_pack = 1, wg_wait = 0, mbar = None)

   GEMM v2: use op tl.gemm_py.


.. py:function:: gemm(A, B, C, transpose_A = False, transpose_B = False, policy = GemmWarpPolicy.Square, clear_accum = False, k_pack = 1, wg_wait = 0, mbar = None)

   TileLang GEMM operator.

   :param A: Input buffer A.
   :type A: tir.Buffer | tir.Var
   :param B: Input buffer B.
   :type B: tir.Buffer | tir.Var
   :param C: Output buffer C.
   :type C: tir.Buffer | tir.Var
   :param transpose_A: Whether to transpose A. Defaults to False.
   :type transpose_A: bool
   :param transpose_B: Whether to transpose B. Defaults to False.
   :type transpose_B: bool
   :param policy: GEMM warp partition policy.
   :type policy: GemmWarpPolicy
   :param clear_accum: Whether to clear the accumulator.
   :type clear_accum: bool
   :param k_pack: Numbers of packed matrix cores, for ROCm only. Defaults to 1.
   :type k_pack: int
   :param wg_wait: Int identifier of the warpgroup MMA batch to wait on.. Defaults to 0.
   :type wg_wait: int
   :param mbar: Mbarrier in Blackwell. Defaults to None.
   :type mbar: tir.Buffer | None, optional

   :returns: A handle to the GEMM operation.
   :rtype: tir.Call