Split-K Reduction and Epilogue Fusion

[Maximilianxu]

Can Triton implement split-k gemm, followed by epilogue fusion? Like the following from cutlass:

 semaphore.wait(threadblock_tile_offset.k());


  // Execute the epilogue operator to update the destination tensor.
  epilogue(
    output_op, 
    iterator_D, 
    accumulators, 
    iterator_C);