rapidsmpf::BloomFilter Struct Reference

A bloom filter, used for approximate set membership queries. More...

#include <bloom_filter.hpp>

Public Member Functions
	BloomFilter (std::size_t num_blocks, std::uint64_t seed, rmm::cuda_stream_view stream, rmm::device_async_resource_ref mr)
	Create a filter. More...
void	add (cudf::table_view const &values_to_hash, rmm::cuda_stream_view stream, rmm::device_async_resource_ref mr)
	Add values to the filter. More...
void	merge (BloomFilter &other, rmm::cuda_stream_view stream)
	Merge two filters, computing their union. More...
rmm::device_uvector< bool >	contains (cudf::table_view const &values, rmm::cuda_stream_view stream, rmm::device_async_resource_ref mr)
	Return a mask of which rows are contained in the filter. More...
rmm::cuda_stream_view	stream () const noexcept
void *	data () noexcept
std::size_t	size () const noexcept

Static Public Member Functions
static std::size_t	fitting_num_blocks (std::size_t l2size) noexcept

Detailed Description

A bloom filter, used for approximate set membership queries.

Definition at line 19 of file bloom_filter.hpp.

Constructor & Destructor Documentation

BloomFilter()

rapidsmpf::BloomFilter::BloomFilter	(	std::size_t	num_blocks,
		std::uint64_t	seed,
		rmm::cuda_stream_view	stream,
		rmm::device_async_resource_ref	mr
	)

Create a filter.

Parameters

num_blocks	Number of blocks in the filter.
seed	Seed used for hashing each value.
stream	CUDA stream for allocations and device operations.
mr	Memory resource for allocations.

Member Function Documentation

add()

void rapidsmpf::BloomFilter::add	(	cudf::table_view const &	values_to_hash,
		rmm::cuda_stream_view	stream,
		rmm::device_async_resource_ref	mr
	)

Add values to the filter.

Parameters

values_to_hash	table of values to hash (with cudf::hashing::xxhash_64())
stream	CUDA stream for allocations and device operations.
mr	Memory resource for allocations.

contains()

rmm::device_uvector<bool> rapidsmpf::BloomFilter::contains	(	cudf::table_view const &	values,
		rmm::cuda_stream_view	stream,
		rmm::device_async_resource_ref	mr
	)

Return a mask of which rows are contained in the filter.

Parameters

values	Value to check for set membership
stream	CUDA stream for allocations and device operations.
mr	Memory resource for allocations.

Returns

Mask vector to be used for filtering the table.

data()

void* rapidsmpf::BloomFilter::data ( )

noexcept

Returns

Pointer to the underlying storage.

fitting_num_blocks()

static std::size_t rapidsmpf::BloomFilter::fitting_num_blocks ( std::size_t  l2size )

staticnoexcept

Returns

Number of blocks to use if the filter should fit in a given L2 cache size.

Parameters

l2size

Size of the L2 cache in bytes.

merge()

void rapidsmpf::BloomFilter::merge	(	BloomFilter &	other,
		rmm::cuda_stream_view	stream
	)

Merge two filters, computing their union.

Parameters

other	Other filter to merge into this one.
stream	CUDA stream for device operations.

Exceptions

std::logic_error

If other is not compatible with this filter.

size()

std::size_t rapidsmpf::BloomFilter::size ( ) const

noexcept

Returns

Size in bytes of the underlying storage.

stream()

rmm::cuda_stream_view rapidsmpf::BloomFilter::stream ( ) const

noexcept

Returns

The stream the underlying storage is valid on.

---

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

• /__w/rapidsmpf/rapidsmpf/cpp/include/rapidsmpf/integrations/cudf/bloom_filter.hpp