示例#1
0
static fch_buckets_t * mapping(cmph_config_t *mph)
{
	cmph_uint32 i = 0;
	fch_buckets_t *buckets = NULL;
	fch_config_data_t *fch = (fch_config_data_t *)mph->data;
	if (fch->h1) hash_state_destroy(fch->h1);
	fch->h1 = hash_state_new(fch->hashfuncs[0], fch->m);  
	fch->b = fch_calc_b(fch->c, fch->m);
	fch->p1 = fch_calc_p1(fch->m);
	fch->p2 = fch_calc_p2(fch->b);
	//DEBUGP("b:%u   p1:%f   p2:%f\n", fch->b, fch->p1, fch->p2);
	buckets = fch_buckets_new(fch->b);

	mph->key_source->rewind(mph->key_source->data);  
	for(i = 0; i < fch->m; i++)
	{
		cmph_uint32 h1, keylen;
		char *key = NULL;
		mph->key_source->read(mph->key_source->data, &key, &keylen);	
		h1 = hash(fch->h1, key, keylen) % fch->m;
		h1 = mixh10h11h12 (fch->b, fch->p1, fch->p2, h1);
		fch_buckets_insert(buckets, h1, key, keylen);
		key = NULL; // transger memory ownership
		
	}
	return buckets;  
}
示例#2
0
static cmph_uint32 brz_fch_search_packed(cmph_uint32 *packed_mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
	register CMPH_HASH h0_type = *packed_mphf++;
	
	register cmph_uint32 *h0_ptr = packed_mphf;
	packed_mphf = (cmph_uint32 *)(((cmph_uint8 *)packed_mphf) + hash_state_packed_size(h0_type)); 
	
	register cmph_uint32 k = *packed_mphf++;

	register double c = (double)(*((cmph_uint64*)packed_mphf));
	packed_mphf += 2;

	register CMPH_HASH h1_type = *packed_mphf++; 

	register CMPH_HASH h2_type = *packed_mphf++; 

	register cmph_uint8 * size = (cmph_uint8 *) packed_mphf;
	packed_mphf = (cmph_uint32 *)(size + k);  
	
	register cmph_uint32 * offset = packed_mphf;
	packed_mphf += k;
	
	register cmph_uint32 h0;
	
	hash_vector_packed(h0_ptr, h0_type, key, keylen, fingerprint);
	h0 = fingerprint[2] % k;
	
	register cmph_uint32 m = size[h0];
	register cmph_uint32 b = fch_calc_b(c, m);
	register double p1 = fch_calc_p1(m);
	register double p2 = fch_calc_p2(b);
	
	#if defined (__ia64) || defined (__x86_64__)
		register cmph_uint64 * g_is_ptr = (cmph_uint64 *)packed_mphf;
	#else
		register cmph_uint32 * g_is_ptr = packed_mphf;
	#endif
	
	register cmph_uint8 * h1_ptr = (cmph_uint8 *) g_is_ptr[h0];
	
	register cmph_uint8 * h2_ptr = h1_ptr + hash_state_packed_size(h1_type);

	register cmph_uint8 * g = h2_ptr + hash_state_packed_size(h2_type);
	
	register cmph_uint32 h1 = hash_packed(h1_ptr, h1_type, key, keylen) % m;
	register cmph_uint32 h2 = hash_packed(h2_ptr, h2_type, key, keylen) % m;

	register cmph_uint8 mphf_bucket = 0;
	h1 = mixh10h11h12(b, p1, p2, h1);
	mphf_bucket = (cmph_uint8)((h2 + g[h1]) % m);
	return (mphf_bucket + offset[h0]);
}
示例#3
0
static cmph_uint32 brz_fch_search(brz_data_t *brz, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
	register cmph_uint32 h0;

	hash_vector(brz->h0, key, keylen, fingerprint);
	h0 = fingerprint[2] % brz->k;

	register cmph_uint32 m = brz->size[h0];
	register cmph_uint32 b = fch_calc_b(brz->c, m);
	register double p1 = fch_calc_p1(m);
	register double p2 = fch_calc_p2(b);
	register cmph_uint32 h1 = hash(brz->h1[h0], key, keylen) % m;
	register cmph_uint32 h2 = hash(brz->h2[h0], key, keylen) % m;
	register cmph_uint8 mphf_bucket = 0;
	h1 = mixh10h11h12(b, p1, p2, h1);
	mphf_bucket = (cmph_uint8)((h2 + brz->g[h0][h1]) % m);
	return (mphf_bucket + brz->offset[h0]);
}
示例#4
0
/** \fn cmph_uint32 brz_packed_size(cmph_t *mphf);
 *  \brief Return the amount of space needed to pack mphf.
 *  \param mphf pointer to a mphf
 *  \return the size of the packed function or zero for failures
 */ 
cmph_uint32 brz_packed_size(cmph_t *mphf)
{
	cmph_uint32 i;
	cmph_uint32 size = 0;
	brz_data_t *data = (brz_data_t *)mphf->data;
	CMPH_HASH h0_type = hash_get_type(data->h0); 
	CMPH_HASH h1_type = hash_get_type(data->h1[0]); 
	CMPH_HASH h2_type = hash_get_type(data->h2[0]);
	size = (cmph_uint32)(2*sizeof(CMPH_ALGO) + 3*sizeof(CMPH_HASH) + hash_state_packed_size(h0_type) + sizeof(cmph_uint32) + 
			sizeof(double) + sizeof(cmph_uint8)*data->k + sizeof(cmph_uint32)*data->k);
	// pointers to g_is
	#if defined (__ia64) || defined (__x86_64__)
		size +=  (cmph_uint32) sizeof(cmph_uint64)*data->k;
	#else
		size +=  (cmph_uint32) sizeof(cmph_uint32)*data->k;
	#endif
	
	size += hash_state_packed_size(h1_type) * data->k;
	size += hash_state_packed_size(h2_type) * data->k;
	
	cmph_uint32 n = 0;
	for(i = 0; i < data->k; i++)
	{
   		switch(data->algo)
   		{
   			case CMPH_FCH:
   				n = fch_calc_b(data->c, data->size[i]);
   				break;
   			case CMPH_BMZ8:
   				n = (cmph_uint32)ceil(data->c * data->size[i]);
   				break;
   			default: assert(0);
   		}
		size += n;	
	}
	return size;
}
示例#5
0
/** \fn void brz_pack(cmph_t *mphf, void *packed_mphf);
 *  \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
 *  \param mphf pointer to the resulting mphf
 *  \param packed_mphf pointer to the contiguous memory area used to store the resulting mphf. The size of packed_mphf must be at least cmph_packed_size() 
 */
void brz_pack(cmph_t *mphf, void *packed_mphf)
{
	brz_data_t *data = (brz_data_t *)mphf->data;
	cmph_uint8 * ptr = packed_mphf;
	cmph_uint32 i,n;
	
	// packing internal algo type
	memcpy(ptr, &(data->algo), sizeof(data->algo));
	ptr += sizeof(data->algo);

	// packing h0 type
	CMPH_HASH h0_type = hash_get_type(data->h0); 
	memcpy(ptr, &h0_type, sizeof(h0_type));
	ptr += sizeof(h0_type);

	// packing h0
	hash_state_pack(data->h0, ptr);
	ptr += hash_state_packed_size(h0_type);
	
	// packing k
	memcpy(ptr, &(data->k), sizeof(data->k));
	ptr += sizeof(data->k);

	// packing c
	*((cmph_uint64 *)ptr) = (cmph_uint64)data->c; 
	ptr += sizeof(data->c);

	// packing h1 type
	CMPH_HASH h1_type = hash_get_type(data->h1[0]); 
	memcpy(ptr, &h1_type, sizeof(h1_type));
	ptr += sizeof(h1_type);

	// packing h2 type
	CMPH_HASH h2_type = hash_get_type(data->h2[0]); 
	memcpy(ptr, &h2_type, sizeof(h2_type));
	ptr += sizeof(h2_type);

	// packing size
	memcpy(ptr, data->size, sizeof(cmph_uint8)*data->k);	
	ptr += data->k;
	
	// packing offset
	memcpy(ptr, data->offset, sizeof(cmph_uint32)*data->k);	
	ptr += sizeof(cmph_uint32)*data->k;
	
	#if defined (__ia64) || defined (__x86_64__)
		cmph_uint64 * g_is_ptr = (cmph_uint64 *)ptr;
	#else
		cmph_uint32 * g_is_ptr = (cmph_uint32 *)ptr;
	#endif
	
	cmph_uint8 * g_i = (cmph_uint8 *) (g_is_ptr + data->k);
	
	for(i = 0; i < data->k; i++)
	{
		#if defined (__ia64) || defined (__x86_64__)
			*g_is_ptr++ = (cmph_uint64)g_i;
		#else
			*g_is_ptr++ = (cmph_uint32)g_i;
		#endif
		// packing h1[i]
		hash_state_pack(data->h1[i], g_i);
		g_i += hash_state_packed_size(h1_type);
		
		// packing h2[i]
		hash_state_pack(data->h2[i], g_i);
		g_i += hash_state_packed_size(h2_type);

		// packing g_i
		switch(data->algo)
		{
			case CMPH_FCH:
				n = fch_calc_b(data->c, data->size[i]);
				break;
			case CMPH_BMZ8:
				n = (cmph_uint32)ceil(data->c * data->size[i]);
				break;
			default: assert(0);
		}
		memcpy(g_i, data->g[i], sizeof(cmph_uint8)*n);	
		g_i += n;
		
	}

}
示例#6
0
void brz_load(FILE *f, cmph_t *mphf)
{
	char *buf = NULL;
	cmph_uint32 buflen;
	register size_t nbytes;
	cmph_uint32 i, n;
	brz_data_t *brz = (brz_data_t *)malloc(sizeof(brz_data_t));

	DEBUGP("Loading brz mphf\n");
	mphf->data = brz;
	nbytes = fread(&(brz->c), sizeof(double), (size_t)1, f);
	nbytes = fread(&(brz->algo), sizeof(brz->algo), (size_t)1, f); // Reading algo.
	nbytes = fread(&(brz->k), sizeof(cmph_uint32), (size_t)1, f);
	brz->size   = (cmph_uint8 *) malloc(sizeof(cmph_uint8)*brz->k);
	nbytes = fread(brz->size, sizeof(cmph_uint8)*(brz->k), (size_t)1, f);	
	brz->h1 = (hash_state_t **)malloc(sizeof(hash_state_t *)*brz->k);
	brz->h2 = (hash_state_t **)malloc(sizeof(hash_state_t *)*brz->k);
	brz->g  = (cmph_uint8 **)  calloc((size_t)brz->k, sizeof(cmph_uint8 *));
	DEBUGP("Reading c = %f   k = %u   algo = %u \n", brz->c, brz->k, brz->algo);
	//loading h_i1, h_i2 and g_i.
	for(i = 0; i < brz->k; i++)
	{
		// h1
		nbytes = fread(&buflen, sizeof(cmph_uint32), (size_t)1, f);
		DEBUGP("Hash state 1 has %u bytes\n", buflen);
		buf = (char *)malloc((size_t)buflen);
		nbytes = fread(buf, (size_t)buflen, (size_t)1, f);
		brz->h1[i] = hash_state_load(buf, buflen);
		free(buf);
		//h2
		nbytes = fread(&buflen, sizeof(cmph_uint32), (size_t)1, f);
		DEBUGP("Hash state 2 has %u bytes\n", buflen);
		buf = (char *)malloc((size_t)buflen);
		nbytes = fread(buf, (size_t)buflen, (size_t)1, f);
		brz->h2[i] = hash_state_load(buf, buflen);
		free(buf);
		switch(brz->algo)
		{
			case CMPH_FCH:
				n = fch_calc_b(brz->c, brz->size[i]);
				break;
			case CMPH_BMZ8:
				n = (cmph_uint32)ceil(brz->c * brz->size[i]);
				break;
			default: assert(0);
		}
		DEBUGP("g_i has %u bytes\n", n);
		brz->g[i] = (cmph_uint8 *)calloc((size_t)n, sizeof(cmph_uint8));
		nbytes = fread(brz->g[i], sizeof(cmph_uint8)*n, (size_t)1, f);
	}
	//loading h0
	nbytes = fread(&buflen, sizeof(cmph_uint32), (size_t)1, f);
	DEBUGP("Hash state has %u bytes\n", buflen);
	buf = (char *)malloc((size_t)buflen);
	nbytes = fread(buf, (size_t)buflen, (size_t)1, f);
	brz->h0 = hash_state_load(buf, buflen);
	free(buf);

	//loading c, m, and the vector offset.	
	nbytes = fread(&(brz->m), sizeof(cmph_uint32), (size_t)1, f);
	brz->offset = (cmph_uint32 *)malloc(sizeof(cmph_uint32)*brz->k);
	nbytes = fread(brz->offset, sizeof(cmph_uint32)*(brz->k), (size_t)1, f);
	return;
}