cub::HistogramPolicy#

struct HistogramPolicy#

The tuning policy for all algorithms in DeviceHistogram.

Public Functions

inline friend ::std::ostream &operator<<(
::std::ostream &os,
const HistogramPolicy &p
)#

Public Members

int threads_per_block#

Number of threads in a CUDA block.

int pixels_per_thread#

Number of pixels processed per thread.

int vec_size#

Vectorization size for loading samples.

BlockLoadAlgorithm load_algorithm#

The BlockLoadAlgorithm used for loading samples from global memory.

CacheLoadModifier load_modifier#

The CacheLoadModifier used for loading samples from global memory.

bool rle_compress#

Whether to perform localized RLE to compress samples before histogramming.

BlockHistogramMemoryPreference mem_preference#

Whether to prefer privatized shared-memory or global-memory bins, or a mix of both.

bool use_work_stealing#

Whether to dequeue tiles from a global work queue.

int init_kernel_pdl_trigger_max_bins#

Maximum number of bins for the init kernel to trigger the histogram kernel early using PDL.

Friends

inline friend constexpr bool operator==(
const HistogramPolicy &lhs,
const HistogramPolicy &rhs
) noexcept#
inline friend constexpr bool operator!=(
const HistogramPolicy &lhs,
const HistogramPolicy &rhs
) noexcept#