rocPRIM
|
Configuration of device-level reduce-by-key operation. More...
#include <device_reduce_by_key_config.hpp>
Static Public Attributes | |
static constexpr unsigned int | block_size = BlockSize |
Number of threads in a block. | |
static constexpr unsigned int | tiles_per_block = TilesPerBlock |
Number of tiles (BlockSize * ItemsPerThread items) to process per block. | |
static constexpr unsigned int | items_per_thread = ItemsPerThread |
Number of items processed by each thread per tile. | |
static constexpr block_load_method | load_keys_method = LoadKeysMethod |
A rocprim::block_load_method emum value indicating how the keys should be loaded. More... | |
static constexpr block_load_method | load_values_method = LoadValuesMethod |
A rocprim::block_load_method emum value indicating how the values should be loaded. More... | |
static constexpr block_scan_algorithm | scan_algorithm = ScanAlgorithm |
A rocprim::block_scan_algorithm enum value indicating how the reduction should be done. More... | |
static constexpr unsigned int | size_limit = SizeLimit |
Maximum possible number of values. Defaults to ROCPRIM_GRID_SIZE_LIMIT. | |
Configuration of device-level reduce-by-key operation.
BlockSize | number of threads in a block. |
ItemsPerThread | number of items processed by each thread per tile. |
LoadKeysMethod | method of loading keys |
LoadValuesMethod | method of loading values |
ScanAlgorithm | block level scan algorithm to use |
TilesPerBlock | number of tiles (BlockSize * ItemsPerThread items) to process per block |
SizeLimit | limit on the number of items for a single reduce_by_key kernel launch. |
|
static |
A rocprim::block_load_method emum value indicating how the keys should be loaded.
Defaults to block_load_method::block_load_transpose
|
static |
A rocprim::block_load_method emum value indicating how the values should be loaded.
Defaults to block_load_method::block_load_transpose
|
static |
A rocprim::block_scan_algorithm enum value indicating how the reduction should be done.
Defaults to block_scan_algorithm::using_warp_scan