From 3324db06000233ce50467527faeaf7d89d6a6008 Mon Sep 17 00:00:00 2001 From: fc_botelho Date: Mon, 5 Sep 2005 17:32:21 +0000 Subject: [PATCH] BMZ8 - A 8 bit version of BMZ has been added --- src/Makefile.am | 1 + src/bmz.c | 6 +- src/bmz8.c | 547 +++++++++++++++++++++++++++++++++++++++++++++ src/bmz8.h | 18 ++ src/bmz8_structs.h | 25 +++ src/cmph.c | 39 +++- src/cmph_types.h | 2 +- 7 files changed, 628 insertions(+), 10 deletions(-) create mode 100644 src/bmz8.c create mode 100644 src/bmz8.h create mode 100644 src/bmz8_structs.h diff --git a/src/Makefile.am b/src/Makefile.am index 73ee4ae..1d05357 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -16,6 +16,7 @@ libcmph_la_SOURCES = debug.h\ cmph_structs.h cmph_structs.c\ chm.h chm_structs.h chm.c\ bmz.h bmz_structs.h bmz.c\ + bmz8.h bmz8_structs.h bmz8.c\ brz.h brz_structs.h brz.c libcmph_la_LDFLAGS = -version-info 0:0:0 diff --git a/src/bmz.c b/src/bmz.c index 87c3aeb..7a3d069 100644 --- a/src/bmz.c +++ b/src/bmz.c @@ -31,13 +31,13 @@ bmz_config_data_t *bmz_config_new() { bmz_config_data_t *bmz; bmz = (bmz_config_data_t *)malloc(sizeof(bmz_config_data_t)); + assert(bmz); memset(bmz, 0, sizeof(bmz_config_data_t)); bmz->hashfuncs[0] = CMPH_HASH_JENKINS; bmz->hashfuncs[1] = CMPH_HASH_JENKINS; bmz->g = NULL; bmz->graph = NULL; bmz->hashes = NULL; - assert(bmz); return bmz; } @@ -378,7 +378,7 @@ static void bmz_traverse(bmz_config_data_t *bmz, cmph_uint8 * used_edges, cmph_u DEBUGP("Visiting neighbor %u\n", neighbor); *unused_edge_index = next_unused_edge(bmz, used_edges, *unused_edge_index); bmz->g[neighbor] = *unused_edge_index - bmz->g[v]; - if (bmz->g[neighbor] >= bmz->m) bmz->g[neighbor] += bmz->m; + //if (bmz->g[neighbor] >= bmz->m) bmz->g[neighbor] += bmz->m; SETBIT(visited, neighbor); (*unused_edge_index)++; bmz_traverse(bmz, used_edges, neighbor, unused_edge_index, visited); @@ -535,7 +535,7 @@ cmph_uint32 bmz_search(cmph_t *mphf, const char *key, cmph_uint32 keylen) DEBUGP("key: %s h1: %u h2: %u\n", key, h1, h2); if (h1 == h2 && ++h2 > bmz->n) h2 = 0; DEBUGP("key: %s g[h1]: %u g[h2]: %u edges: %u\n", key, bmz->g[h1], bmz->g[h2], bmz->m); - return ((bmz->g[h1] + bmz->g[h2]) % bmz->m); + return (bmz->g[h1] + bmz->g[h2]); } void bmz_destroy(cmph_t *mphf) { diff --git a/src/bmz8.c b/src/bmz8.c new file mode 100644 index 0000000..65de52f --- /dev/null +++ b/src/bmz8.c @@ -0,0 +1,547 @@ +#include "graph.h" +#include "bmz8.h" +#include "cmph_structs.h" +#include "bmz8_structs.h" +#include "hash.h" +#include "vqueue.h" +#include "bitbool.h" + +#include +#include +#include +#include +#include + +//#define DEBUG +#include "debug.h" + +static int bmz8_gen_edges(cmph_config_t *mph); +static cmph_uint8 bmz8_traverse_critical_nodes(bmz8_config_data_t *bmz8, cmph_uint8 v, cmph_uint8 * biggest_g_value, cmph_uint8 * biggest_edge_value, cmph_uint8 * used_edges, cmph_uint8 * visited); +static cmph_uint8 bmz8_traverse_critical_nodes_heuristic(bmz8_config_data_t *bmz8, cmph_uint8 v, cmph_uint8 * biggest_g_value, cmph_uint8 * biggest_edge_value, cmph_uint8 * used_edges, cmph_uint8 * visited); +static void bmz8_traverse_non_critical_nodes(bmz8_config_data_t *bmz8, cmph_uint8 * used_edges, cmph_uint8 * visited); + +bmz8_config_data_t *bmz8_config_new() +{ + bmz8_config_data_t *bmz8; + bmz8 = (bmz8_config_data_t *)malloc(sizeof(bmz8_config_data_t)); + assert(bmz8); + memset(bmz8, 0, sizeof(bmz8_config_data_t)); + bmz8->hashfuncs[0] = CMPH_HASH_JENKINS; + bmz8->hashfuncs[1] = CMPH_HASH_JENKINS; + bmz8->g = NULL; + bmz8->graph = NULL; + bmz8->hashes = NULL; + return bmz8; +} + +void bmz8_config_destroy(cmph_config_t *mph) +{ + bmz8_config_data_t *data = (bmz8_config_data_t *)mph->data; + DEBUGP("Destroying algorithm dependent data\n"); + free(data); +} + +void bmz8_config_set_hashfuncs(cmph_config_t *mph, CMPH_HASH *hashfuncs) +{ + bmz8_config_data_t *bmz8 = (bmz8_config_data_t *)mph->data; + CMPH_HASH *hashptr = hashfuncs; + cmph_uint8 i = 0; + while(*hashptr != CMPH_HASH_COUNT) + { + if (i >= 2) break; //bmz8 only uses two hash functions + bmz8->hashfuncs[i] = *hashptr; + ++i, ++hashptr; + } +} + +cmph_t *bmz8_new(cmph_config_t *mph, float c) +{ + cmph_t *mphf = NULL; + bmz8_data_t *bmz8f = NULL; + cmph_uint8 i; + cmph_uint8 iterations; + cmph_uint8 iterations_map = 20; + cmph_uint8 *used_edges = NULL; + cmph_uint8 restart_mapping = 0; + cmph_uint8 * visited = NULL; + bmz8_config_data_t *bmz8 = (bmz8_config_data_t *)mph->data; + + + if (mph->key_source->nkeys >= 256) + { + if (mph->verbosity) fprintf(stderr, "The number of keys in BMZ8 must be lower than 256.\n"); + return NULL; + } + + DEBUGP("c: %f\n", c); + bmz8->m = mph->key_source->nkeys; + bmz8->n = ceil(c * mph->key_source->nkeys); + DEBUGP("m (edges): %u n (vertices): %u c: %f\n", bmz8->m, bmz8->n, c); + bmz8->graph = graph_new(bmz8->n, bmz8->m); + DEBUGP("Created graph\n"); + + bmz8->hashes = (hash_state_t **)malloc(sizeof(hash_state_t *)*3); + for(i = 0; i < 3; ++i) bmz8->hashes[i] = NULL; + + do + { + // Mapping step + cmph_uint8 biggest_g_value = 0; + cmph_uint8 biggest_edge_value = 1; + iterations = 100; + if (mph->verbosity) + { + fprintf(stderr, "Entering mapping step for mph creation of %u keys with graph sized %u\n", bmz8->m, bmz8->n); + } + while(1) + { + int ok; + DEBUGP("hash function 1\n"); + bmz8->hashes[0] = hash_state_new(bmz8->hashfuncs[0], bmz8->n); + DEBUGP("hash function 2\n"); + bmz8->hashes[1] = hash_state_new(bmz8->hashfuncs[1], bmz8->n); + DEBUGP("Generating edges\n"); + ok = bmz8_gen_edges(mph); + if (!ok) + { + --iterations; + hash_state_destroy(bmz8->hashes[0]); + bmz8->hashes[0] = NULL; + hash_state_destroy(bmz8->hashes[1]); + bmz8->hashes[1] = NULL; + DEBUGP("%u iterations remaining\n", iterations); + if (mph->verbosity) + { + fprintf(stderr, "simple graph creation failure - %u iterations remaining\n", iterations); + } + if (iterations == 0) break; + } + else break; + } + if (iterations == 0) + { + graph_destroy(bmz8->graph); + return NULL; + } + + // Ordering step + if (mph->verbosity) + { + fprintf(stderr, "Starting ordering step\n"); + } + + graph_obtain_critical_nodes(bmz8->graph); + + // Searching step + if (mph->verbosity) + { + fprintf(stderr, "Starting Searching step.\n"); + fprintf(stderr, "\tTraversing critical vertices.\n"); + } + DEBUGP("Searching step\n"); + visited = (char *)malloc(bmz8->n/8 + 1); + memset(visited, 0, bmz8->n/8 + 1); + used_edges = (cmph_uint8 *)malloc(bmz8->m/8 + 1); + memset(used_edges, 0, bmz8->m/8 + 1); + free(bmz8->g); + bmz8->g = calloc(bmz8->n, sizeof(cmph_uint8)); + assert(bmz8->g); + for (i = 0; i < bmz8->n; ++i) // critical nodes + { + if (graph_node_is_critical(bmz8->graph, i) && (!GETBIT(visited,i))) + { + if(c > 1.14) restart_mapping = bmz8_traverse_critical_nodes(bmz8, i, &biggest_g_value, &biggest_edge_value, used_edges, visited); + else restart_mapping = bmz8_traverse_critical_nodes_heuristic(bmz8, i, &biggest_g_value, &biggest_edge_value, used_edges, visited); + if(restart_mapping) break; + } + } + if(!restart_mapping) + { + if (mph->verbosity) + { + fprintf(stderr, "\tTraversing non critical vertices.\n"); + } + bmz8_traverse_non_critical_nodes(bmz8, used_edges, visited); // non_critical_nodes + } + else + { + iterations_map--; + if (mph->verbosity) fprintf(stderr, "Restarting mapping step. %u iterations remaining.\n", iterations_map); + } + free(used_edges); + free(visited); + }while(restart_mapping && iterations_map > 0); + graph_destroy(bmz8->graph); + bmz8->graph = NULL; + if (iterations_map == 0) + { + return NULL; + } + mphf = (cmph_t *)malloc(sizeof(cmph_t)); + mphf->algo = mph->algo; + bmz8f = (bmz8_data_t *)malloc(sizeof(bmz8_data_t)); + bmz8f->g = bmz8->g; + bmz8->g = NULL; //transfer memory ownership + bmz8f->hashes = bmz8->hashes; + bmz8->hashes = NULL; //transfer memory ownership + bmz8f->n = bmz8->n; + bmz8f->m = bmz8->m; + mphf->data = bmz8f; + mphf->size = bmz8->m; + DEBUGP("Successfully generated minimal perfect hash\n"); + if (mph->verbosity) + { + fprintf(stderr, "Successfully generated minimal perfect hash function\n"); + } + return mphf; +} + +static cmph_uint8 bmz8_traverse_critical_nodes(bmz8_config_data_t *bmz8, cmph_uint8 v, cmph_uint8 * biggest_g_value, cmph_uint8 * biggest_edge_value, cmph_uint8 * used_edges, cmph_uint8 * visited) +{ + cmph_uint8 next_g; + cmph_uint32 u; /* Auxiliary vertex */ + cmph_uint32 lav; /* lookahead vertex */ + cmph_uint8 collision; + vqueue_t * q = vqueue_new((cmph_uint32)(graph_ncritical_nodes(bmz8->graph))); + graph_iterator_t it, it1; + + DEBUGP("Labelling critical vertices\n"); + bmz8->g[v] = (cmph_uint8)ceil ((double)(*biggest_edge_value)/2) - 1; + SETBIT(visited, v); + next_g = (cmph_uint8)floor((double)(*biggest_edge_value/2)); /* next_g is incremented in the do..while statement*/ + vqueue_insert(q, v); + while(!vqueue_is_empty(q)) + { + v = vqueue_remove(q); + it = graph_neighbors_it(bmz8->graph, v); + while ((u = graph_next_neighbor(bmz8->graph, &it)) != GRAPH_NO_NEIGHBOR) + { + if (graph_node_is_critical(bmz8->graph, u) && (!GETBIT(visited,u))) + { + collision = 1; + while(collision) // lookahead to resolve collisions + { + next_g = *biggest_g_value + 1; + it1 = graph_neighbors_it(bmz8->graph, u); + collision = 0; + while((lav = graph_next_neighbor(bmz8->graph, &it1)) != GRAPH_NO_NEIGHBOR) + { + if (graph_node_is_critical(bmz8->graph, lav) && GETBIT(visited,lav)) + { + if(next_g + bmz8->g[lav] >= bmz8->m) + { + vqueue_destroy(q); + return 1; // restart mapping step. + } + if (GETBIT(used_edges, next_g + bmz8->g[lav])) + { + collision = 1; + break; + } + } + } + if (next_g > *biggest_g_value) *biggest_g_value = next_g; + } + // Marking used edges... + it1 = graph_neighbors_it(bmz8->graph, u); + while((lav = graph_next_neighbor(bmz8->graph, &it1)) != GRAPH_NO_NEIGHBOR) + { + if (graph_node_is_critical(bmz8->graph, lav) && GETBIT(visited, lav)) + { + SETBIT(used_edges,next_g + bmz8->g[lav]); + if(next_g + bmz8->g[lav] > *biggest_edge_value) *biggest_edge_value = next_g + bmz8->g[lav]; + } + } + bmz8->g[u] = next_g; // Labelling vertex u. + SETBIT(visited,u); + vqueue_insert(q, u); + } + } + + } + vqueue_destroy(q); + return 0; +} + +static cmph_uint8 bmz8_traverse_critical_nodes_heuristic(bmz8_config_data_t *bmz8, cmph_uint8 v, cmph_uint8 * biggest_g_value, cmph_uint8 * biggest_edge_value, cmph_uint8 * used_edges, cmph_uint8 * visited) +{ + cmph_uint8 next_g; + cmph_uint32 u; /* Auxiliary vertex */ + cmph_uint32 lav; /* lookahead vertex */ + cmph_uint8 collision; + cmph_uint8 * unused_g_values = NULL; + cmph_uint8 unused_g_values_capacity = 0; + cmph_uint8 nunused_g_values = 0; + vqueue_t * q = vqueue_new((cmph_uint32)(graph_ncritical_nodes(bmz8->graph))); + graph_iterator_t it, it1; + + DEBUGP("Labelling critical vertices\n"); + bmz8->g[v] = (cmph_uint8)ceil ((double)(*biggest_edge_value)/2) - 1; + SETBIT(visited, v); + next_g = (cmph_uint8)floor((double)(*biggest_edge_value/2)); /* next_g is incremented in the do..while statement*/ + vqueue_insert(q, v); + while(!vqueue_is_empty(q)) + { + v = vqueue_remove(q); + it = graph_neighbors_it(bmz8->graph, v); + while ((u = graph_next_neighbor(bmz8->graph, &it)) != GRAPH_NO_NEIGHBOR) + { + if (graph_node_is_critical(bmz8->graph, u) && (!GETBIT(visited,u))) + { + cmph_uint8 next_g_index = 0; + collision = 1; + while(collision) // lookahead to resolve collisions + { + if (next_g_index < nunused_g_values) + { + next_g = unused_g_values[next_g_index++]; + } + else + { + next_g = *biggest_g_value + 1; + next_g_index = 255;//UINT_MAX; + } + it1 = graph_neighbors_it(bmz8->graph, u); + collision = 0; + while((lav = graph_next_neighbor(bmz8->graph, &it1)) != GRAPH_NO_NEIGHBOR) + { + if (graph_node_is_critical(bmz8->graph, lav) && GETBIT(visited,lav)) + { + if(next_g + bmz8->g[lav] >= bmz8->m) + { + vqueue_destroy(q); + free(unused_g_values); + return 1; // restart mapping step. + } + if (GETBIT(used_edges, next_g + bmz8->g[lav])) + { + collision = 1; + break; + } + } + } + if(collision && (next_g > *biggest_g_value)) // saving the current g value stored in next_g. + { + if(nunused_g_values == unused_g_values_capacity) + { + unused_g_values = realloc(unused_g_values, (unused_g_values_capacity + BUFSIZ)*sizeof(cmph_uint8)); + unused_g_values_capacity += BUFSIZ; + } + unused_g_values[nunused_g_values++] = next_g; + + } + if (next_g > *biggest_g_value) *biggest_g_value = next_g; + } + next_g_index--; + if (next_g_index < nunused_g_values) unused_g_values[next_g_index] = unused_g_values[--nunused_g_values]; + + // Marking used edges... + it1 = graph_neighbors_it(bmz8->graph, u); + while((lav = graph_next_neighbor(bmz8->graph, &it1)) != GRAPH_NO_NEIGHBOR) + { + if (graph_node_is_critical(bmz8->graph, lav) && GETBIT(visited, lav)) + { + SETBIT(used_edges,next_g + bmz8->g[lav]); + if(next_g + bmz8->g[lav] > *biggest_edge_value) *biggest_edge_value = next_g + bmz8->g[lav]; + } + } + bmz8->g[u] = next_g; // Labelling vertex u. + SETBIT(visited, u); + vqueue_insert(q, u); + } + } + + } + vqueue_destroy(q); + free(unused_g_values); + return 0; +} + +static cmph_uint8 next_unused_edge(bmz8_config_data_t *bmz8, cmph_uint8 * used_edges, cmph_uint8 unused_edge_index) +{ + while(1) + { + assert(unused_edge_index < bmz8->m); + if(GETBIT(used_edges, unused_edge_index)) unused_edge_index ++; + else break; + } + return unused_edge_index; +} + +static void bmz8_traverse(bmz8_config_data_t *bmz8, cmph_uint8 * used_edges, cmph_uint8 v, cmph_uint8 * unused_edge_index, cmph_uint8 * visited) +{ + graph_iterator_t it = graph_neighbors_it(bmz8->graph, v); + cmph_uint32 neighbor = 0; + while((neighbor = graph_next_neighbor(bmz8->graph, &it)) != GRAPH_NO_NEIGHBOR) + { + if(GETBIT(visited,neighbor)) continue; + DEBUGP("Visiting neighbor %u\n", neighbor); + *unused_edge_index = next_unused_edge(bmz8, used_edges, *unused_edge_index); + bmz8->g[neighbor] = *unused_edge_index - bmz8->g[v]; + SETBIT(visited, neighbor); + (*unused_edge_index)++; + bmz8_traverse(bmz8, used_edges, neighbor, unused_edge_index, visited); + + } +} + +static void bmz8_traverse_non_critical_nodes(bmz8_config_data_t *bmz8, cmph_uint8 * used_edges, cmph_uint8 * visited) +{ + + cmph_uint8 i, v1, v2, unused_edge_index = 0; + DEBUGP("Labelling non critical vertices\n"); + for(i = 0; i < bmz8->m; i++) + { + v1 = graph_vertex_id(bmz8->graph, i, 0); + v2 = graph_vertex_id(bmz8->graph, i, 1); + if((GETBIT(visited,v1) && GETBIT(visited,v2)) || (!GETBIT(visited,v1) && !GETBIT(visited,v2))) continue; + if(GETBIT(visited,v1)) bmz8_traverse(bmz8, used_edges, v1, &unused_edge_index, visited); + else bmz8_traverse(bmz8, used_edges, v2, &unused_edge_index, visited); + + } + + for(i = 0; i < bmz8->n; i++) + { + if(!GETBIT(visited,i)) + { + bmz8->g[i] = 0; + SETBIT(visited, i); + bmz8_traverse(bmz8, used_edges, i, &unused_edge_index, visited); + } + } + +} + +static int bmz8_gen_edges(cmph_config_t *mph) +{ + cmph_uint8 e; + bmz8_config_data_t *bmz8 = (bmz8_config_data_t *)mph->data; + cmph_uint8 multiple_edges = 0; + DEBUGP("Generating edges for %u vertices\n", bmz8->n); + graph_clear_edges(bmz8->graph); + mph->key_source->rewind(mph->key_source->data); + for (e = 0; e < mph->key_source->nkeys; ++e) + { + cmph_uint8 h1, h2; + cmph_uint32 keylen; + char *key = NULL; + mph->key_source->read(mph->key_source->data, &key, &keylen); + +// if (key == NULL)fprintf(stderr, "key = %s -- read BMZ\n", key); + h1 = hash(bmz8->hashes[0], key, keylen) % bmz8->n; + h2 = hash(bmz8->hashes[1], key, keylen) % bmz8->n; + if (h1 == h2) if (++h2 >= bmz8->n) h2 = 0; + if (h1 == h2) + { + if (mph->verbosity) fprintf(stderr, "Self loop for key %u\n", e); + mph->key_source->dispose(mph->key_source->data, key, keylen); + return 0; + } + DEBUGP("Adding edge: %u -> %u for key %s\n", h1, h2, key); + mph->key_source->dispose(mph->key_source->data, key, keylen); +// fprintf(stderr, "key = %s -- dispose BMZ\n", key); + multiple_edges = graph_contains_edge(bmz8->graph, h1, h2); + if (mph->verbosity && multiple_edges) fprintf(stderr, "A non simple graph was generated\n"); + if (multiple_edges) return 0; // checking multiple edge restriction. + graph_add_edge(bmz8->graph, h1, h2); + } + return !multiple_edges; +} + +int bmz8_dump(cmph_t *mphf, FILE *fd) +{ + char *buf = NULL; + cmph_uint32 buflen; + cmph_uint8 i; + cmph_uint8 two = 2; //number of hash functions + bmz8_data_t *data = (bmz8_data_t *)mphf->data; + cmph_uint8 nn, nm; + __cmph_dump(mphf, fd); + + fwrite(&two, sizeof(cmph_uint8), 1, fd); + + hash_state_dump(data->hashes[0], &buf, &buflen); + DEBUGP("Dumping hash state with %u bytes to disk\n", buflen); + fwrite(&buflen, sizeof(cmph_uint32), 1, fd); + fwrite(buf, buflen, 1, fd); + free(buf); + + hash_state_dump(data->hashes[1], &buf, &buflen); + DEBUGP("Dumping hash state with %u bytes to disk\n", buflen); + fwrite(&buflen, sizeof(cmph_uint32), 1, fd); + fwrite(buf, buflen, 1, fd); + free(buf); + + fwrite(&(data->n), sizeof(cmph_uint8), 1, fd); + fwrite(&(data->m), sizeof(cmph_uint8), 1, fd); + + fwrite(data->g, sizeof(cmph_uint8)*(data->n), 1, fd); + #ifdef DEBUG + fprintf(stderr, "G: "); + for (i = 0; i < data->n; ++i) fprintf(stderr, "%u ", data->g[i]); + fprintf(stderr, "\n"); + #endif + return 1; +} + +void bmz8_load(FILE *f, cmph_t *mphf) +{ + cmph_uint8 nhashes; + char *buf = NULL; + cmph_uint32 buflen; + cmph_uint8 i; + bmz8_data_t *bmz8 = (bmz8_data_t *)malloc(sizeof(bmz8_data_t)); + + DEBUGP("Loading bmz8 mphf\n"); + mphf->data = bmz8; + fread(&nhashes, sizeof(cmph_uint8), 1, f); + bmz8->hashes = (hash_state_t **)malloc(sizeof(hash_state_t *)*(nhashes + 1)); + bmz8->hashes[nhashes] = NULL; + DEBUGP("Reading %u hashes\n", nhashes); + for (i = 0; i < nhashes; ++i) + { + hash_state_t *state = NULL; + fread(&buflen, sizeof(cmph_uint32), 1, f); + DEBUGP("Hash state has %u bytes\n", buflen); + buf = (char *)malloc(buflen); + fread(buf, buflen, 1, f); + state = hash_state_load(buf, buflen); + bmz8->hashes[i] = state; + free(buf); + } + + DEBUGP("Reading m and n\n"); + fread(&(bmz8->n), sizeof(cmph_uint8), 1, f); + fread(&(bmz8->m), sizeof(cmph_uint8), 1, f); + + bmz8->g = (cmph_uint8 *)malloc(sizeof(cmph_uint8)*bmz8->n); + fread(bmz8->g, bmz8->n*sizeof(cmph_uint8), 1, f); + #ifdef DEBUG + fprintf(stderr, "G: "); + for (i = 0; i < bmz8->n; ++i) fprintf(stderr, "%u ", bmz8->g[i]); + fprintf(stderr, "\n"); + #endif + return; +} + + +cmph_uint8 bmz8_search(cmph_t *mphf, const char *key, cmph_uint32 keylen) +{ + bmz8_data_t *bmz8 = mphf->data; + cmph_uint8 h1 = hash(bmz8->hashes[0], key, keylen) % bmz8->n; + cmph_uint8 h2 = hash(bmz8->hashes[1], key, keylen) % bmz8->n; + DEBUGP("key: %s h1: %u h2: %u\n", key, h1, h2); + if (h1 == h2 && ++h2 > bmz8->n) h2 = 0; + DEBUGP("key: %s g[h1]: %u g[h2]: %u edges: %u\n", key, bmz8->g[h1], bmz8->g[h2], bmz8->m); + return (bmz8->g[h1] + bmz8->g[h2]); +} +void bmz8_destroy(cmph_t *mphf) +{ + bmz8_data_t *data = (bmz8_data_t *)mphf->data; + free(data->g); + hash_state_destroy(data->hashes[0]); + hash_state_destroy(data->hashes[1]); + free(data->hashes); + free(data); + free(mphf); +} diff --git a/src/bmz8.h b/src/bmz8.h new file mode 100644 index 0000000..cb35cd4 --- /dev/null +++ b/src/bmz8.h @@ -0,0 +1,18 @@ +#ifndef __CMPH_BMZ8_H__ +#define __CMPH_BMZ8_H__ + +#include "cmph.h" + +typedef struct __bmz8_data_t bmz8_data_t; +typedef struct __bmz8_config_data_t bmz8_config_data_t; + +bmz8_config_data_t *bmz8_config_new(); +void bmz8_config_set_hashfuncs(cmph_config_t *mph, CMPH_HASH *hashfuncs); +void bmz8_config_destroy(cmph_config_t *mph); +cmph_t *bmz8_new(cmph_config_t *mph, float c); + +void bmz8_load(FILE *f, cmph_t *mphf); +int bmz8_dump(cmph_t *mphf, FILE *f); +void bmz8_destroy(cmph_t *mphf); +cmph_uint8 bmz8_search(cmph_t *mphf, const char *key, cmph_uint32 keylen); +#endif diff --git a/src/bmz8_structs.h b/src/bmz8_structs.h new file mode 100644 index 0000000..408b529 --- /dev/null +++ b/src/bmz8_structs.h @@ -0,0 +1,25 @@ +#ifndef __CMPH_BMZ8_STRUCTS_H__ +#define __CMPH_BMZ8_STRUCTS_H__ + +#include "hash_state.h" + +struct __bmz8_data_t +{ + cmph_uint8 m; //edges (words) count + cmph_uint8 n; //vertex count + cmph_uint8 *g; + hash_state_t **hashes; +}; + + +struct __bmz8_config_data_t +{ + CMPH_HASH hashfuncs[2]; + cmph_uint8 m; //edges (words) count + cmph_uint8 n; //vertex count + graph_t *graph; + cmph_uint8 *g; + hash_state_t **hashes; +}; + +#endif diff --git a/src/cmph.c b/src/cmph.c index 203d60d..8f86f4e 100644 --- a/src/cmph.c +++ b/src/cmph.c @@ -2,6 +2,7 @@ #include "cmph_structs.h" #include "chm.h" #include "bmz.h" +#include "bmz8.h" /* included -- Fabiano */ #include "brz.h" /* included -- Fabiano */ #include @@ -10,7 +11,7 @@ //#define DEBUG #include "debug.h" -const char *cmph_names[] = { "bmz", "chm", "brz", NULL }; /* included -- Fabiano */ +const char *cmph_names[] = { "bmz", "bmz8", "chm", "brz", NULL }; /* included -- Fabiano */ static cmph_uint32 position; // access position when data is a vector @@ -154,6 +155,9 @@ void cmph_config_set_algo(cmph_config_t *mph, CMPH_ALGO algo) case CMPH_BMZ: bmz_config_destroy(mph); break; + case CMPH_BMZ8: + bmz8_config_destroy(mph); + break; case CMPH_BRZ: brz_config_destroy(mph); break; @@ -168,9 +172,11 @@ void cmph_config_set_algo(cmph_config_t *mph, CMPH_ALGO algo) case CMPH_BMZ: mph->data = bmz_config_new(); break; + case CMPH_BMZ8: + mph->data = (void*)bmz8_config_new(); + break; case CMPH_BRZ: mph->data = brz_config_new(); - break; default: assert(0); } @@ -186,6 +192,8 @@ void cmph_config_set_tmp_dir(cmph_config_t *mph, cmph_uint8 *tmp_dir) break; case CMPH_BMZ: /* included -- Fabiano */ break; + case CMPH_BMZ8: /* included -- Fabiano */ + break; case CMPH_BRZ: /* included -- Fabiano */ brz_config_set_tmp_dir(mph, tmp_dir); break; @@ -206,9 +214,11 @@ void cmph_config_destroy(cmph_config_t *mph) case CMPH_BMZ: /* included -- Fabiano */ bmz_config_destroy(mph); break; + case CMPH_BMZ8: /* included -- Fabiano */ + bmz8_config_destroy(mph); + break; case CMPH_BRZ: /* included -- Fabiano */ brz_config_destroy(mph); - break; default: assert(0); } @@ -230,6 +240,9 @@ void cmph_config_set_hashfuncs(cmph_config_t *mph, CMPH_HASH *hashfuncs) case CMPH_BMZ: /* included -- Fabiano */ bmz_config_set_hashfuncs(mph, hashfuncs); break; + case CMPH_BMZ8: /* included -- Fabiano */ + bmz8_config_set_hashfuncs(mph, hashfuncs); + break; case CMPH_BRZ: /* included -- Fabiano */ brz_config_set_hashfuncs(mph, hashfuncs); break; @@ -262,6 +275,11 @@ cmph_t *cmph_new(cmph_config_t *mph) if (c == 0) c = 1.15; mphf = bmz_new(mph, c); break; + case CMPH_BMZ8: /* included -- Fabiano */ + DEBUGP("Creating bmz8 hash\n"); + if (c == 0) c = 1.15; + mphf = bmz8_new(mph, c); + break; case CMPH_BRZ: /* included -- Fabiano */ DEBUGP("Creating brz hash\n"); if (c == 0) c = 1.15; @@ -279,13 +297,12 @@ int cmph_dump(cmph_t *mphf, FILE *f) { case CMPH_CHM: return chm_dump(mphf, f); - break; case CMPH_BMZ: /* included -- Fabiano */ return bmz_dump(mphf, f); - break; + case CMPH_BMZ8: /* included -- Fabiano */ + return bmz8_dump(mphf, f); case CMPH_BRZ: /* included -- Fabiano */ return brz_dump(mphf, f); - break; default: assert(0); } @@ -309,6 +326,10 @@ cmph_t *cmph_load(FILE *f) DEBUGP("Loading bmz algorithm dependent parts\n"); bmz_load(f, mphf); break; + case CMPH_BMZ8: /* included -- Fabiano */ + DEBUGP("Loading bmz8 algorithm dependent parts\n"); + bmz8_load(f, mphf); + break; case CMPH_BRZ: /* included -- Fabiano */ DEBUGP("Loading brz algorithm dependent parts\n"); brz_load(f, mphf); @@ -331,6 +352,9 @@ cmph_uint32 cmph_search(cmph_t *mphf, const char *key, cmph_uint32 keylen) case CMPH_BMZ: /* included -- Fabiano */ DEBUGP("bmz algorithm search\n"); return bmz_search(mphf, key, keylen); + case CMPH_BMZ8: /* included -- Fabiano */ + DEBUGP("bmz8 algorithm search\n"); + return bmz8_search(mphf, key, keylen); case CMPH_BRZ: /* included -- Fabiano */ DEBUGP("brz algorithm search\n"); return brz_search(mphf, key, keylen); @@ -356,6 +380,9 @@ void cmph_destroy(cmph_t *mphf) case CMPH_BMZ: /* included -- Fabiano */ bmz_destroy(mphf); return; + case CMPH_BMZ8: /* included -- Fabiano */ + bmz8_destroy(mphf); + return; case CMPH_BRZ: /* included -- Fabiano */ brz_destroy(mphf); return; diff --git a/src/cmph_types.h b/src/cmph_types.h index 9359e33..be976ee 100644 --- a/src/cmph_types.h +++ b/src/cmph_types.h @@ -9,7 +9,7 @@ typedef float cmph_float32; typedef enum { CMPH_HASH_DJB2, CMPH_HASH_FNV, CMPH_HASH_JENKINS, CMPH_HASH_SDBM, CMPH_HASH_COUNT } CMPH_HASH; extern const char *cmph_hash_names[]; -typedef enum { CMPH_BMZ, CMPH_CHM, CMPH_BRZ, CMPH_COUNT } CMPH_ALGO; /* included -- Fabiano */ +typedef enum { CMPH_BMZ, CMPH_BMZ8, CMPH_CHM, CMPH_BRZ, CMPH_COUNT } CMPH_ALGO; /* included -- Fabiano */ extern const char *cmph_names[]; #endif