cuda::experimental::stf::blocked_partition#
-
using cuda::experimental::stf::blocked_partition = blocked_partition_custom<>#
Partitions a multidimensional box or shape into contiguous blocks along a selected dimension.
This partitioning strategy divides the data space into contiguous blocks, distributing them across execution places. By default, partitioning occurs along the last dimension, but a specific dimension can be selected using the template parameter. This approach ensures good spatial locality and is particularly effective for regular data access patterns.