thrust::mr::pool_options

Defined in thrust/mr/pool_options.h

struct pool_options

A type used for configuring pooling resource adaptors, to fine-tune their behavior and parameters.

Public Functions

inline bool validate() const

Checks if the options are self-consistent.

/returns true if the options are self-consitent, false otherwise.

Public Members

std::size_t min_blocks_per_chunk

The minimal number of blocks, i.e. pieces of memory handed off to the user from a pool of a given size, in a single chunk allocated from upstream.

std::size_t min_bytes_per_chunk

The minimal number of bytes in a single chunk allocated from upstream.

std::size_t max_blocks_per_chunk

The maximal number of blocks, i.e. pieces of memory handed off to the user from a pool of a given size, in a single chunk allocated from upstream.

std::size_t max_bytes_per_chunk

The maximal number of bytes in a single chunk allocated from upstream.

std::size_t smallest_block_size

The size of blocks in the smallest pool covered by the pool resource. All allocation requests below this size will be rounded up to this size.

std::size_t largest_block_size

The size of blocks in the largest pool covered by the pool resource. All allocation requests above this size will be considered oversized, allocated directly from upstream (and not from a pool), and cached only of cache_oversized is true.

std::size_t alignment

The alignment of all blocks in internal pools of the pool resource. All allocation requests above this alignment will be considered oversized, allocated directly from upstream (and not from a pool), and cached only of cache_oversized is true.

bool cache_oversized

Decides whether oversized and overaligned blocks are cached for later use, or immediately return it to the upstream resource.

std::size_t cached_size_cutoff_factor

The size factor at which a cached allocation is considered too ridiculously oversized to use to fulfill an allocation request. For instance: the user requests an allocation of size 1024 bytes. A block of size 32 * 1024 bytes is cached. If cached_size_cutoff_factor is 32 or less, this block will be considered too big for that allocation request.

std::size_t cached_alignment_cutoff_factor

The alignment factor at which a cached allocation is considered too ridiculously overaligned to use to fulfill an allocation request. For instance: the user requests an allocation aligned to 32 bytes. A block aligned to 1024 bytes is cached. If cached_size_cutoff_factor is 32 or less, this block will be considered too overaligned for that allocation request.