Epilogue operator.
#include <direct_epilogue_tensor_op.h>
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
- Parameters
-
| params | Host-constructable params object |
| shared_storage | Shared storage object |
| thread_idx | ID of a thread within the threadblock |
| warp_idx | ID of warp within threadblock |
| lane_idx | Id of thread within warp |
template<typename Shape_ , typename Operator_ , int PartitionsK, typename Element_ , typename OutputOp_ , typename ConvertOp_ >
< Accumulator tile
Number of mma operations performed
- Parameters
-
| problem_size | Problem size of GEMM (units of ElementC) |
| tb_tile_coord | Threadblock tile coordinate in GEMM (in units of threadblock tiles) |
The documentation for this class was generated from the following file: