#include "../hash.h"
#include "../minmax.h"
#include "../crc/xxhash.h"
-#include "../lib/murmur3.h"
+#include "../crc/murmur3.h"
+#include "../crc/crc32c.h"
+#include "../crc/fnv.h"
struct bloom {
uint64_t nentries;
uint32_t (*fn)(const void *, uint32_t, uint32_t);
};
-struct bloom_hash hashes[] = {
+static uint32_t bloom_crc32c(const void *buf, uint32_t len, uint32_t seed)
+{
+ return fio_crc32c(buf, len);
+}
+
+static uint32_t bloom_fnv(const void *buf, uint32_t len, uint32_t seed)
+{
+ return fnv(buf, len, seed);
+}
+
+#define BLOOM_SEED 0x8989
+
+static struct bloom_hash hashes[] = {
{
- .seed = 0x8989,
+ .seed = BLOOM_SEED,
.fn = jhash,
},
{
- .seed = 0x8989,
+ .seed = BLOOM_SEED,
.fn = XXH32,
},
{
- .seed = 0x8989,
+ .seed = BLOOM_SEED,
.fn = murmurhash3,
},
+ {
+ .seed = BLOOM_SEED,
+ .fn = bloom_crc32c,
+ },
+ {
+ .seed = BLOOM_SEED,
+ .fn = bloom_fnv,
+ },
};
-#define N_HASHES 3
-
-#define MIN_ENTRIES 1073741824UL
+#define N_HASHES 5
struct bloom *bloom_new(uint64_t entries)
{
struct bloom *b;
size_t no_uints;
+ crc32c_arm64_probe();
+ crc32c_intel_probe();
+
b = malloc(sizeof(*b));
b->nentries = entries;
no_uints = (entries + BITS_PER_INDEX - 1) / BITS_PER_INDEX;
- no_uints = max((unsigned long) no_uints, MIN_ENTRIES);
b->map = calloc(no_uints, sizeof(uint32_t));
if (!b->map) {
free(b);
free(b);
}
-static int __bloom_check(struct bloom *b, uint32_t *data, unsigned int nwords,
- int set)
+static bool __bloom_check(struct bloom *b, const void *data, unsigned int len,
+ bool set)
{
uint32_t hash[N_HASHES];
int i, was_set;
for (i = 0; i < N_HASHES; i++) {
- hash[i] = hashes[i].fn(data, nwords, hashes[i].seed);
+ hash[i] = hashes[i].fn(data, len, hashes[i].seed);
hash[i] = hash[i] % b->nentries;
}
if (b->map[index] & (1U << bit))
was_set++;
- if (set)
+ else if (set)
b->map[index] |= 1U << bit;
+ else
+ break;
}
return was_set == N_HASHES;
}
-int bloom_check(struct bloom *b, uint32_t *data, unsigned int nwords)
+bool bloom_set(struct bloom *b, uint32_t *data, unsigned int nwords)
{
- return __bloom_check(b, data, nwords, 0);
+ return __bloom_check(b, data, nwords * sizeof(uint32_t), true);
}
-int bloom_set(struct bloom *b, uint32_t *data, unsigned int nwords)
+bool bloom_string(struct bloom *b, const char *data, unsigned int len,
+ bool set)
{
- return __bloom_check(b, data, nwords, 1);
+ return __bloom_check(b, data, len, set);
}