tilelang.language.experimental.gemm_sp
======================================

.. py:module:: tilelang.language.experimental.gemm_sp

.. autoapi-nested-parse::

   The language interface for tl programs.



Functions
---------

.. autoapisummary::

   tilelang.language.experimental.gemm_sp.gemm_sp


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

.. py:function:: gemm_sp(A_sparse, E, B, C, transpose_A = False, transpose_B = False, policy = GemmWarpPolicy.Square, clear_accum = False, k_pack = 1, wg_wait = 0)

   Perform a Sparse General Matrix Multiplication (GEMM-sp) operation.

   This function computes C = A @ B where A and B can optionally be transposed.
   The operation supports various warp policies and accumulation modes.

   :param A_sparse: First input matrix dense values
   :type A_sparse: Union[tir.Buffer, tir.Var]
   :param E: First input matrix sparse metadata
   :type E: Union[tir.Buffer, tir.Var]
   :param B: Second input matrix
   :type B: Union[tir.Buffer, tir.Var]
   :param C: Output matrix for results
   :type C: Union[tir.Buffer, tir.Var]
   :param transpose_A: Whether to transpose matrix A. Defaults to False.
   :type transpose_A: bool, optional
   :param transpose_B: Whether to transpose matrix B. Defaults to False.
   :type transpose_B: bool, optional
   :param policy: Warp execution policy. Defaults to GemmWarpPolicy.Square.
   :type policy: GemmWarpPolicy, optional
   :param clear_accum: Whether to clear accumulator before computation. Defaults to False.
   :type clear_accum: bool, optional
   :param k_pack: Number of k dimensions packed into a single warp. Defaults to 1.
   :type k_pack: int, optional
   :param wg_wait: Warp group wait count. Defaults to 0.
   :type wg_wait: int, optional

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

   :raises AssertionError: If the K dimensions of matrices A and B don't match