PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation > Struct Template Reference

• ck
• PartitionedBlockwiseReduction_v2

Static Public Member Functions | Static Public Attributes | List of all members

ck::PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation > Struct Template Reference

#include <reduction_functions_blockwise.hpp>

Static Public Member Functions
template<typename BufferType>
static __device__ void	Reduce (BufferType &work_buffer, AccDataType &in_out_value)

Static Public Attributes
static constexpr auto	BufferLength_M = ThreadClusterLengths_M_K::At(0)
static constexpr auto	BufferLength_K = ThreadClusterLengths_M_K::At(1)
static constexpr auto	block_buf_desc_m_k
static constexpr auto	thread_cluster_desc = ThreadClusterDesc{}

Member Function Documentation

◆ Reduce()

template<typename AccDataType, index_t BlockSize, typename ThreadClusterLengths_M_K, typename ThreadClusterDesc, typename OpReduce, bool PropagateNan, typename Accumulation = detail::AccumulateWithNanCheck<PropagateNan, OpReduce, AccDataType>>

template<typename BufferType>

__device__ void ck::PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation >::Reduce ( BufferType & work_buffer, AccDataType & in_out_value )

inlinestatic

Member Data Documentation

◆ block_buf_desc_m_k

template<typename AccDataType, index_t BlockSize, typename ThreadClusterLengths_M_K, typename ThreadClusterDesc, typename OpReduce, bool PropagateNan, typename Accumulation = detail::AccumulateWithNanCheck<PropagateNan, OpReduce, AccDataType>>

auto ck::PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation >::block_buf_desc_m_k

staticconstexpr

Initial value:

= make_naive_tensor_descriptor_packed(

make_tuple(Number<BufferLength_M>{}, Number<BufferLength_K>{}))

ck::Number

integral_constant< index_t, N > Number

Definition number.hpp:12

ck::make_naive_tensor_descriptor_packed

__host__ __device__ constexpr auto make_naive_tensor_descriptor_packed(const Tuple< Lengths... > &lengths)

Definition tensor_descriptor_helper.hpp:101

ck::make_tuple

__host__ __device__ constexpr auto make_tuple(Xs &&... xs)

Definition utility/tuple.hpp:211

◆ BufferLength_K

template<typename AccDataType, index_t BlockSize, typename ThreadClusterLengths_M_K, typename ThreadClusterDesc, typename OpReduce, bool PropagateNan, typename Accumulation = detail::AccumulateWithNanCheck<PropagateNan, OpReduce, AccDataType>>

auto ck::PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation >::BufferLength_K = ThreadClusterLengths_M_K::At(1)

staticconstexpr

◆ BufferLength_M

template<typename AccDataType, index_t BlockSize, typename ThreadClusterLengths_M_K, typename ThreadClusterDesc, typename OpReduce, bool PropagateNan, typename Accumulation = detail::AccumulateWithNanCheck<PropagateNan, OpReduce, AccDataType>>

auto ck::PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation >::BufferLength_M = ThreadClusterLengths_M_K::At(0)

staticconstexpr

◆ thread_cluster_desc

template<typename AccDataType, index_t BlockSize, typename ThreadClusterLengths_M_K, typename ThreadClusterDesc, typename OpReduce, bool PropagateNan, typename Accumulation = detail::AccumulateWithNanCheck<PropagateNan, OpReduce, AccDataType>>

auto ck::PartitionedBlockwiseReduction_v2< AccDataType, BlockSize, ThreadClusterLengths_M_K, ThreadClusterDesc, OpReduce, PropagateNan, Accumulation >::thread_cluster_desc = ThreadClusterDesc{}

staticconstexpr

---

The documentation for this struct was generated from the following file:

• reduction_functions_blockwise.hpp