More efficient bitsliced rolling Bloom filter
This patch changes the implementation from one that stores 16 2-bit integers in one uint32_t's, to one that stores the first bit of 64 2-bit integers in one uint64_t and the second bit in another. This allows for 450x faster refreshing and 2.2x faster average speed.
This commit is contained in:
parent
aa62b68745
commit
1953c40aa9
3 changed files with 32 additions and 26 deletions
|
@ -234,14 +234,18 @@ CRollingBloomFilter::CRollingBloomFilter(unsigned int nElements, double fpRate)
|
|||
*/
|
||||
uint32_t nFilterBits = (uint32_t)ceil(-1.0 * nHashFuncs * nMaxElements / log(1.0 - exp(logFpRate / nHashFuncs)));
|
||||
data.clear();
|
||||
/* We store up to 16 'bits' per data element. */
|
||||
data.resize((nFilterBits + 15) / 16);
|
||||
/* For each data element we need to store 2 bits. If both bits are 0, the
|
||||
* bit is treated as unset. If the bits are (01), (10), or (11), the bit is
|
||||
* treated as set in generation 1, 2, or 3 respectively.
|
||||
* These bits are stored in separate integers: position P corresponds to bit
|
||||
* (P & 63) of the integers data[(P >> 6) * 2] and data[(P >> 6) * 2 + 1]. */
|
||||
data.resize(((nFilterBits + 63) / 64) << 1);
|
||||
reset();
|
||||
}
|
||||
|
||||
/* Similar to CBloomFilter::Hash */
|
||||
inline unsigned int CRollingBloomFilter::Hash(unsigned int nHashNum, const std::vector<unsigned char>& vDataToHash) const {
|
||||
return MurmurHash3(nHashNum * 0xFBA4C795 + nTweak, vDataToHash) % (data.size() * 16);
|
||||
static inline uint32_t RollingBloomHash(unsigned int nHashNum, uint32_t nTweak, const std::vector<unsigned char>& vDataToHash) {
|
||||
return MurmurHash3(nHashNum * 0xFBA4C795 + nTweak, vDataToHash);
|
||||
}
|
||||
|
||||
void CRollingBloomFilter::insert(const std::vector<unsigned char>& vKey)
|
||||
|
@ -252,18 +256,25 @@ void CRollingBloomFilter::insert(const std::vector<unsigned char>& vKey)
|
|||
if (nGeneration == 4) {
|
||||
nGeneration = 1;
|
||||
}
|
||||
uint64_t nGenerationMask1 = -(uint64_t)(nGeneration & 1);
|
||||
uint64_t nGenerationMask2 = -(uint64_t)(nGeneration >> 1);
|
||||
/* Wipe old entries that used this generation number. */
|
||||
for (uint32_t p = 0; p < data.size() * 16; p++) {
|
||||
if (get(p) == nGeneration) {
|
||||
put(p, 0);
|
||||
}
|
||||
for (uint32_t p = 0; p < data.size(); p += 2) {
|
||||
uint64_t p1 = data[p], p2 = data[p + 1];
|
||||
uint64_t mask = (p1 ^ nGenerationMask1) | (p2 ^ nGenerationMask2);
|
||||
data[p] = p1 & mask;
|
||||
data[p + 1] = p2 & mask;
|
||||
}
|
||||
}
|
||||
nEntriesThisGeneration++;
|
||||
|
||||
for (int n = 0; n < nHashFuncs; n++) {
|
||||
uint32_t h = Hash(n, vKey);
|
||||
put(h, nGeneration);
|
||||
uint32_t h = RollingBloomHash(n, nTweak, vKey);
|
||||
int bit = h & 0x3F;
|
||||
uint32_t pos = (h >> 6) % data.size();
|
||||
/* The lowest bit of pos is ignored, and set to zero for the first bit, and to one for the second. */
|
||||
data[pos & ~1] = (data[pos & ~1] & ~(((uint64_t)1) << bit)) | ((uint64_t)(nGeneration & 1)) << bit;
|
||||
data[pos | 1] = (data[pos | 1] & ~(((uint64_t)1) << bit)) | ((uint64_t)(nGeneration >> 1)) << bit;
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -276,8 +287,11 @@ void CRollingBloomFilter::insert(const uint256& hash)
|
|||
bool CRollingBloomFilter::contains(const std::vector<unsigned char>& vKey) const
|
||||
{
|
||||
for (int n = 0; n < nHashFuncs; n++) {
|
||||
uint32_t h = Hash(n, vKey);
|
||||
if (get(h) == 0) {
|
||||
uint32_t h = RollingBloomHash(n, nTweak, vKey);
|
||||
int bit = h & 0x3F;
|
||||
uint32_t pos = (h >> 6) % data.size();
|
||||
/* If the relevant bit is not set in either data[pos & ~1] or data[pos | 1], the filter does not contain vKey */
|
||||
if (!(((data[pos & ~1] | data[pos | 1]) >> bit) & 1)) {
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
@ -295,7 +309,7 @@ void CRollingBloomFilter::reset()
|
|||
nTweak = GetRand(std::numeric_limits<unsigned int>::max());
|
||||
nEntriesThisGeneration = 0;
|
||||
nGeneration = 1;
|
||||
for (std::vector<uint32_t>::iterator it = data.begin(); it != data.end(); it++) {
|
||||
for (std::vector<uint64_t>::iterator it = data.begin(); it != data.end(); it++) {
|
||||
*it = 0;
|
||||
}
|
||||
}
|
||||
|
|
13
src/bloom.h
13
src/bloom.h
|
@ -135,20 +135,9 @@ private:
|
|||
int nEntriesPerGeneration;
|
||||
int nEntriesThisGeneration;
|
||||
int nGeneration;
|
||||
std::vector<uint32_t> data;
|
||||
std::vector<uint64_t> data;
|
||||
unsigned int nTweak;
|
||||
int nHashFuncs;
|
||||
|
||||
unsigned int Hash(unsigned int nHashNum, const std::vector<unsigned char>& vDataToHash) const;
|
||||
|
||||
inline int get(uint32_t position) const {
|
||||
return (data[(position >> 4) % data.size()] >> (2 * (position & 0xF))) & 0x3;
|
||||
}
|
||||
|
||||
inline void put(uint32_t position, uint32_t val) {
|
||||
uint32_t& cell = data[(position >> 4) % data.size()];
|
||||
cell = (cell & ~(((uint32_t)3) << (2 * (position & 0xF)))) | (val << (2 * (position & 0xF)));
|
||||
}
|
||||
};
|
||||
|
||||
#endif // BITCOIN_BLOOM_H
|
||||
|
|
|
@ -514,11 +514,14 @@ BOOST_AUTO_TEST_CASE(rolling_bloom)
|
|||
if (i >= 100)
|
||||
BOOST_CHECK(rb1.contains(data[i-100]));
|
||||
rb1.insert(data[i]);
|
||||
BOOST_CHECK(rb1.contains(data[i]));
|
||||
}
|
||||
|
||||
// Insert 999 more random entries:
|
||||
for (int i = 0; i < 999; i++) {
|
||||
rb1.insert(RandomData());
|
||||
std::vector<unsigned char> d = RandomData();
|
||||
rb1.insert(d);
|
||||
BOOST_CHECK(rb1.contains(d));
|
||||
}
|
||||
// Sanity check to make sure the filter isn't just filling up:
|
||||
nHits = 0;
|
||||
|
|
Loading…
Reference in a new issue