#include <reduction_functions_threadwise.hpp>
|
template<typename SrcValueBufferType , typename SrcIndexBufferType , typename DstValueBufferType , typename DstIndexBufferType > |
static __device__ void | Reduce (const SrcValueBufferType &src_val_buf, const SrcIndexBufferType &src_idx_buf, DstValueBufferType &dst_val_buf, DstIndexBufferType &dst_idx_buf) |
|
◆ Reduce()
template<typename AccDataType , typename IndexDataType , typename SrcThreadDesc_M_K , typename DstThreadDesc_M , typename OpReduce , bool PropagateNan, typename Accumulation = detail::AccumulateWithIndexAndNanCheck<PropagateNan, OpReduce, AccDataType, IndexDataType>>
template<typename SrcValueBufferType , typename SrcIndexBufferType , typename DstValueBufferType , typename DstIndexBufferType >
static __device__ void ck::ThreadwiseReductionWithIndex< AccDataType, IndexDataType, SrcThreadDesc_M_K, DstThreadDesc_M, OpReduce, PropagateNan, Accumulation >::Reduce |
( |
const SrcValueBufferType & |
src_val_buf, |
|
|
const SrcIndexBufferType & |
src_idx_buf, |
|
|
DstValueBufferType & |
dst_val_buf, |
|
|
DstIndexBufferType & |
dst_idx_buf |
|
) |
| |
|
inlinestatic |
◆ dst_length_m
template<typename AccDataType , typename IndexDataType , typename SrcThreadDesc_M_K , typename DstThreadDesc_M , typename OpReduce , bool PropagateNan, typename Accumulation = detail::AccumulateWithIndexAndNanCheck<PropagateNan, OpReduce, AccDataType, IndexDataType>>
constexpr auto ck::ThreadwiseReductionWithIndex< AccDataType, IndexDataType, SrcThreadDesc_M_K, DstThreadDesc_M, OpReduce, PropagateNan, Accumulation >::dst_length_m = dst_thread_desc_m.GetLength(Number<0>{}) |
|
staticconstexpr |
◆ dst_thread_desc_m
template<typename AccDataType , typename IndexDataType , typename SrcThreadDesc_M_K , typename DstThreadDesc_M , typename OpReduce , bool PropagateNan, typename Accumulation = detail::AccumulateWithIndexAndNanCheck<PropagateNan, OpReduce, AccDataType, IndexDataType>>
constexpr auto ck::ThreadwiseReductionWithIndex< AccDataType, IndexDataType, SrcThreadDesc_M_K, DstThreadDesc_M, OpReduce, PropagateNan, Accumulation >::dst_thread_desc_m = DstThreadDesc_M{} |
|
staticconstexpr |
◆ src_length_k
template<typename AccDataType , typename IndexDataType , typename SrcThreadDesc_M_K , typename DstThreadDesc_M , typename OpReduce , bool PropagateNan, typename Accumulation = detail::AccumulateWithIndexAndNanCheck<PropagateNan, OpReduce, AccDataType, IndexDataType>>
constexpr auto ck::ThreadwiseReductionWithIndex< AccDataType, IndexDataType, SrcThreadDesc_M_K, DstThreadDesc_M, OpReduce, PropagateNan, Accumulation >::src_length_k = src_thread_desc_m_k.GetLength(Number<1>{}) |
|
staticconstexpr |
◆ src_length_m
template<typename AccDataType , typename IndexDataType , typename SrcThreadDesc_M_K , typename DstThreadDesc_M , typename OpReduce , bool PropagateNan, typename Accumulation = detail::AccumulateWithIndexAndNanCheck<PropagateNan, OpReduce, AccDataType, IndexDataType>>
constexpr auto ck::ThreadwiseReductionWithIndex< AccDataType, IndexDataType, SrcThreadDesc_M_K, DstThreadDesc_M, OpReduce, PropagateNan, Accumulation >::src_length_m = src_thread_desc_m_k.GetLength(Number<0>{}) |
|
staticconstexpr |
◆ src_thread_desc_m_k
template<typename AccDataType , typename IndexDataType , typename SrcThreadDesc_M_K , typename DstThreadDesc_M , typename OpReduce , bool PropagateNan, typename Accumulation = detail::AccumulateWithIndexAndNanCheck<PropagateNan, OpReduce, AccDataType, IndexDataType>>
constexpr auto ck::ThreadwiseReductionWithIndex< AccDataType, IndexDataType, SrcThreadDesc_M_K, DstThreadDesc_M, OpReduce, PropagateNan, Accumulation >::src_thread_desc_m_k = SrcThreadDesc_M_K{} |
|
staticconstexpr |
The documentation for this struct was generated from the following file:
- /home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck/tensor_operation/gpu/thread/reduction_functions_threadwise.hpp