X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/ca9bf6a02e31f2a107f1963ba9724bb9a6affd77..0d2220ca1c127fc19c249e6e59dec2ac47107854:/src/xbt/dict.c diff --git a/src/xbt/dict.c b/src/xbt/dict.c index 6c13d1c483..97cfedca84 100644 --- a/src/xbt/dict.c +++ b/src/xbt/dict.c @@ -7,7 +7,11 @@ /* This program is free software; you can redistribute it and/or modify it * under the terms of the license (GNU LGPL) which comes with this package. */ +#define DJB2_HASH_FUNCTION +//#define FNV_HASH_FUNCTION + #include +#include #include "xbt/ex.h" #include "xbt/log.h" #include "xbt/mallocator.h" @@ -34,16 +38,6 @@ static void dict_mallocator_reset_f(void* dict); * Creates and initialize a new dictionnary with a default hashtable size. */ xbt_dict_t xbt_dict_new(void) { - return xbt_dict_new_ext(256); -} - -/** - * \brief Create a new dictionary with the specified hashtable size - * \param hashsize the hashtable size - * \return a pointer to the created object - * \see xbt_dict_new(), xbt_dict_free() - */ -xbt_dict_t xbt_dict_new_ext(int hashsize) { xbt_dict_t dict; if (dict_mallocator == NULL) { @@ -59,9 +53,10 @@ xbt_dict_t xbt_dict_new_ext(int hashsize) { } dict = xbt_mallocator_get(dict_mallocator); - dict->table_size = hashsize; - dict->table = xbt_new0(xbt_dictelm_t, dict->table_size); + dict->table_size = 127; + dict->table = xbt_new0(xbt_dictelm_t, dict->table_size+1); dict->count = 0; + dict->fill = 0; return dict; } @@ -78,6 +73,8 @@ void xbt_dict_free(xbt_dict_t *dict) { int table_size; xbt_dictelm_t *table; +// if ( *dict ) xbt_dict_dump_sizes(*dict); + if (dict != NULL && *dict != NULL) { table_size = (*dict)->table_size; table = (*dict)->table; @@ -95,63 +92,123 @@ void xbt_dict_free(xbt_dict_t *dict) { } } -/** - * \brief Change the hashtable size - * \param dict a dictionary - * \param hashsize the new hashtable size - * - * Change the hashtable size is a long operation, so it's better - * to use xbt_dict_new_ext or to call xbt_dict_hashsize_set when - * the dictionary is empty. - */ -void xbt_dict_hashsize_set(xbt_dict_t dict, int hashsize) { - xbt_dict_t new_dict = xbt_dict_new_ext(hashsize); - xbt_dictelm_t element, next; - int i; - - for (i = 0; i < dict->table_size; i++) { - element = dict->table[i]; - while (element != NULL) { - next = element->next; /* save the next because it will be lost */ - xbt_dict_add_element(new_dict, element); /* no new xbt_dictelm_t is mallocated */ - element = next; - } - } - - xbt_free(dict->table); - dict->table = new_dict->table; - dict->table_size = hashsize; - xbt_free(new_dict); -} - /** * Returns the hash code of a string. */ -static unsigned int xbt_dict_hash_ext(const char *str, int str_len) { +static XBT_INLINE unsigned int xbt_dict_hash_ext(const char *str, int str_len) { + + +#ifdef DJB2_HASH_FUNCTION /* fast implementation of djb2 algorithm */ - unsigned int hash = 5381; int c; + register unsigned int hash = 5381; while (str_len--) { c = *str++; hash = ((hash << 5) + hash) + c; /* hash * 33 + c */ } - +# elif defined(FNV_HASH_FUNCTION) + register unsigned int hash = 0x811c9dc5; + unsigned char *bp = (unsigned char *)str; /* start of buffer */ + unsigned char *be = bp + str_len; /* beyond end of buffer */ + + while (bp < be) { + /* multiply by the 32 bit FNV magic prime mod 2^32 */ + hash += (hash<<1) + (hash<<4) + (hash<<7) + (hash<<8) + (hash<<24); + + /* xor the bottom with the current octet */ + hash ^= (unsigned int)*bp++; + } + +# else + register unsigned int hash = 0; + + while (str_len--) { + hash += (*str) * (*str); + str++; + } +#endif + return hash; } -static unsigned int xbt_dict_hash(const char *str) { +static XBT_INLINE unsigned int xbt_dict_hash(const char *str) { +#ifdef DJB2_HASH_FUNCTION /* fast implementation of djb2 algorithm */ - unsigned int hash = 5381; int c; + register unsigned int hash = 5381; while ( (c = *str++) ) { hash = ((hash << 5) + hash) + c; /* hash * 33 + c */ } - + +# elif defined(FNV_HASH_FUNCTION) + register unsigned int hash = 0x811c9dc5; + + while (*str) { + /* multiply by the 32 bit FNV magic prime mod 2^32 */ + hash += (hash<<1) + (hash<<4) + (hash<<7) + (hash<<8) + (hash<<24); + + /* xor the bottom with the current octet */ + hash ^= (unsigned int)*str++; + } + +# else + register unsigned int hash = 0; + + while (*str) { + hash += (*str) * (*str); + str++; + } +#endif return hash; } +/* Expend the size of the dict */ +static void xbt_dict_rehash(xbt_dict_t dict) { + const int oldsize = dict->table_size + 1; + register int newsize = oldsize * 2; + register int i; + register xbt_dictelm_t *currcell; + register xbt_dictelm_t *twincell; + register xbt_dictelm_t bucklet; + register xbt_dictelm_t *pprev; + + currcell = (xbt_dictelm_t*) xbt_realloc((char*)dict->table, newsize * sizeof(xbt_dictelm_t)); + memset(&currcell[oldsize], 0, oldsize * sizeof(xbt_dictelm_t)); /* zero second half */ + dict->table_size = --newsize; + dict->table = currcell; + DEBUG2("REHASH (%d->%d)",oldsize,newsize); + + for (i=0; ihash_code & newsize) != i) { /* Move to b */ + *pprev = bucklet->next; + bucklet->next = *twincell; + if (!*twincell) + dict->fill++; + *twincell = bucklet; + continue; + } else { + pprev = &bucklet->next; + } + + } + + if (!*currcell) /* everything moved */ + dict->fill--; + } +} + /** * \brief Add data to the dict (arbitrary key) * \param dict the container @@ -164,36 +221,43 @@ static unsigned int xbt_dict_hash(const char *str) { * Set the \a data in the structure under the \a key, which can be any kind * of data, as long as its length is provided in \a key_len. */ -void xbt_dict_set_ext(xbt_dict_t dict, - const char *key, - int key_len, - void *data, - void_f_pvoid_t *free_ctn) { - xbt_assert(dict); +XBT_INLINE void xbt_dict_set_ext(xbt_dict_t dict, + const char *key, int key_len, + void *data, void_f_pvoid_t free_ctn) { - unsigned int hash_code = xbt_dict_hash_ext(key,key_len) % dict->table_size; - xbt_dictelm_t current, previous = NULL; + unsigned int hash_code = xbt_dict_hash_ext(key,key_len); - current = dict->table[hash_code]; + xbt_dictelm_t current, previous = NULL; + xbt_assert(dict); + + DEBUG5("ADD %.*s hash = %d, size = %d, & = %d",key_len,key,hash_code, dict->table_size, hash_code & dict->table_size); + current = dict->table[hash_code & dict->table_size]; while (current != NULL && - (key_len != current->key_len || strncmp(key, current->key, key_len))) { + (hash_code != current->hash_code || key_len != current->key_len || memcmp(key, current->key, key_len))) { previous = current; current = current->next; } if (current == NULL) { /* this key doesn't exist yet */ - current = xbt_dictelm_new(key, key_len, data, free_ctn, NULL); + current = xbt_dictelm_new(key, key_len, hash_code, data, free_ctn); + dict->count++; if (previous == NULL) { - dict->table[hash_code] = current; - } - else { + dict->table[hash_code & dict->table_size] = current; + dict->fill++; + if ((dict->fill * 100) / (dict->table_size + 1) > MAX_FILL_PERCENT) + xbt_dict_rehash(dict); + } else { previous->next = current; } - dict->count++; } else { - /* there is already an element with the same key: we overwrite it */ + + DEBUG6("Replace %.*s by %.*s under key %.*s", + key_len,(char*)current->content, + key_len,(char*)data, + key_len,(char*)key); + /* there is already an element with the same key: overwrite it */ if (current->content != NULL && current->free_f != NULL) { current->free_f(current->content); } @@ -214,13 +278,11 @@ void xbt_dict_set_ext(xbt_dict_t dict, * set the \a data in the structure under the \a key, which is a * null terminated string. */ -void xbt_dict_set(xbt_dict_t dict, +void xbt_dict_set(xbt_dict_t dict, const char *key, void *data, - void_f_pvoid_t *free_ctn) { + void_f_pvoid_t free_ctn) { - xbt_assert(dict); - xbt_dict_set_ext(dict, key, strlen(key), data, free_ctn); } @@ -234,23 +296,23 @@ void xbt_dict_set(xbt_dict_t dict, * * Search the given \a key. Throws not_found_error when not found. */ -void *xbt_dict_get_ext(xbt_dict_t dict, - const char *key, - int key_len) { - xbt_assert(dict); +void *xbt_dict_get_ext(xbt_dict_t dict, + const char *key, int key_len) { - unsigned int hash_code = xbt_dict_hash_ext(key,key_len) % dict->table_size; + + unsigned int hash_code = xbt_dict_hash_ext(key,key_len); xbt_dictelm_t current; - current = dict->table[hash_code]; + xbt_assert(dict); + + current = dict->table[hash_code & dict->table_size]; while (current != NULL && - (key_len != current->key_len || strncmp(key, current->key, key_len))) { + (hash_code != current->hash_code || key_len != current->key_len || memcmp(key, current->key, key_len))) { current = current->next; } - if (current == NULL) { + if (current == NULL) THROW2(not_found_error, 0, "key %.*s not found", key_len, key); - } return current->content; } @@ -268,19 +330,18 @@ void *xbt_dict_get_ext(xbt_dict_t dict, */ void *xbt_dict_get(xbt_dict_t dict, const char *key) { - xbt_assert(dict); - unsigned int hash_code = xbt_dict_hash(key) % dict->table_size; + unsigned int hash_code = xbt_dict_hash(key); xbt_dictelm_t current; - current = dict->table[hash_code]; - while (current != NULL && (strcmp(key, current->key))) { + xbt_assert(dict); + + current = dict->table[hash_code & dict->table_size]; + while (current != NULL && (hash_code != current->hash_code || strcmp(key, current->key))) current = current->next; - } - if (current == NULL) { + if (current == NULL) THROW1(not_found_error, 0, "key %s not found", key); - } return current->content; } @@ -288,19 +349,22 @@ void *xbt_dict_get(xbt_dict_t dict, /** * \brief like xbt_dict_get(), but returning NULL when not found */ -void *xbt_dict_get_or_null(xbt_dict_t dict, - const char *key) { - xbt_ex_t e; - void *result = NULL; - TRY { - result = xbt_dict_get(dict, key); - } CATCH(e) { - if (e.category != not_found_error) - RETHROW; - xbt_ex_free(e); - result = NULL; - } - return result; +void *xbt_dict_get_or_null(xbt_dict_t dict, + const char *key) { + unsigned int hash_code = xbt_dict_hash(key); + xbt_dictelm_t current; + + xbt_assert(dict); + + current = dict->table[hash_code & dict->table_size]; + while (current != NULL && + hash_code != current->hash_code && strcmp(key, current->key)) + current = current->next; + + if (current == NULL) + return NULL; + + return current->content; } @@ -316,29 +380,32 @@ void *xbt_dict_get_or_null(xbt_dict_t dict, void xbt_dict_remove_ext(xbt_dict_t dict, const char *key, int key_len) { - xbt_assert(dict); - unsigned int hash_code = xbt_dict_hash_ext(key,key_len) % dict->table_size; + + unsigned int hash_code = xbt_dict_hash_ext(key,key_len); xbt_dictelm_t current, previous = NULL; - current = dict->table[hash_code]; + xbt_assert(dict); + +// fprintf(stderr,"RM %.*s hash = %d, size = %d, & = %d\n",key_len,key,hash_code, dict->table_size, hash_code & dict->table_size); + current = dict->table[hash_code & dict->table_size]; while (current != NULL && - (key_len != current->key_len || strncmp(key, current->key, key_len))) { + (hash_code != current->hash_code || key_len != current->key_len || strncmp(key, current->key, key_len))) { previous = current; /* save the previous node */ current = current->next; } - if (current == NULL) { + if (current == NULL) THROW2(not_found_error, 0, "key %.*s not found", key_len, key); - } if (previous != NULL) { - xbt_assert0(previous->next == current, "previous-next != current"); previous->next = current->next; + } else { + dict->table[hash_code & dict->table_size] = current->next; } - else { - dict->table[hash_code] = current->next; - } + + if (!dict->table[hash_code & dict->table_size]) + dict->fill--; xbt_dictelm_free(current); dict->count--; @@ -352,10 +419,7 @@ void xbt_dict_remove_ext(xbt_dict_t dict, * * Remove the entry associated with the given \a key */ -void xbt_dict_remove(xbt_dict_t dict, - const char *key) { - xbt_assert(dict); - +void xbt_dict_remove(xbt_dict_t dict, const char *key) { xbt_dict_remove_ext(dict, key, strlen(key)); } @@ -364,15 +428,16 @@ void xbt_dict_remove(xbt_dict_t dict, * \param dict the dict */ void xbt_dict_reset(xbt_dict_t dict) { - xbt_assert(dict); int i; xbt_dictelm_t current, previous = NULL; + + xbt_assert(dict); if (dict->count == 0) return; - for (i = 0; i < dict->table_size; i++) { + for (i = 0; i <= dict->table_size; i++) { current = dict->table[i]; while (current != NULL) { previous = current; @@ -383,6 +448,7 @@ void xbt_dict_reset(xbt_dict_t dict) { } dict->count = 0; + dict->fill = 0; } /** @@ -395,17 +461,6 @@ int xbt_dict_length(xbt_dict_t dict) { return dict->count; } -/* - * Add an already mallocated element to a dictionary. - */ -void xbt_dict_add_element(xbt_dict_t dict, xbt_dictelm_t element) { - xbt_assert(dict); - - int hashcode = xbt_dict_hash_ext(element->key,element->key_len) % dict->table_size; - element->next = dict->table[hashcode]; - dict->table[hashcode] = element; -} - /** * \brief Outputs the content of the structure (debuging purpose) * @@ -417,23 +472,87 @@ void xbt_dict_add_element(xbt_dict_t dict, xbt_dictelm_t element) { */ void xbt_dict_dump(xbt_dict_t dict, - void_f_pvoid_t *output) { + void_f_pvoid_t output) { int i; xbt_dictelm_t element; printf("Dict %p:\n", dict); if (dict != NULL) { for (i = 0; i < dict->table_size; i++) { element = dict->table[i]; - while (element != NULL) { - printf("%s -> ", element->key); - if (output != NULL) { - output(element->content); - } - printf("\n"); - element = element->next; + if (element) { + printf("[\n"); + while (element != NULL) { + printf(" %s -> ", element->key); + if (output != NULL) { + (*output)(element->content); + } + printf("\n"); + element = element->next; + } + printf("]\n"); + } else { + printf("[]\n"); + } + } + } +} + +xbt_dynar_t all_sizes = NULL; +/** @brief shows some debugging info about the bucklet repartition */ +void xbt_dict_dump_sizes(xbt_dict_t dict) { + + int i; + unsigned int count; + unsigned int size; + xbt_dictelm_t element; + xbt_dynar_t sizes = xbt_dynar_new(sizeof(int),NULL); + + printf("Dict %p: %d bucklets, %d used cells (of %d) ", dict, dict->count, dict->fill,dict->table_size); + if (dict != NULL) { + for (i = 0; i < dict->table_size; i++) { + element = dict->table[i]; + size = 0; + if (element) { + while (element != NULL) { + size ++; + element = element->next; + } } + if (xbt_dynar_length(sizes) <= size) { + int prevsize = 1; + xbt_dynar_set(sizes,size,&prevsize); + } else { + int prevsize; + xbt_dynar_get_cpy(sizes,size,&prevsize); + prevsize++; + xbt_dynar_set(sizes,size,&prevsize); + } + } + if (!all_sizes) + all_sizes = xbt_dynar_new(sizeof(int), NULL); + + xbt_dynar_foreach(sizes,count,size) { + /* Copy values of this one into all_sizes */ + int prevcount; + if (xbt_dynar_length(all_sizes) <= count) { + prevcount = size; + xbt_dynar_set(all_sizes,count,&prevcount); + } else { + xbt_dynar_get_cpy(all_sizes,count,&prevcount); + prevcount += size; + xbt_dynar_set(all_sizes,count,&prevcount); + } + + /* Report current sizes */ + if (count==0) + continue; + if (size==0) + continue; + printf("%delm x %u cells; ",count,size); } } + printf("\n"); + xbt_dynar_free(&sizes); } /** @@ -443,7 +562,26 @@ void xbt_dict_dump(xbt_dict_t dict, void xbt_dict_exit(void) { if (dict_mallocator != NULL) { xbt_mallocator_free(dict_mallocator); + dict_mallocator = NULL; xbt_mallocator_free(dict_elm_mallocator); + dict_elm_mallocator = NULL; + } + if (all_sizes) { + unsigned int count; + int size; + double avg = 0; + int total_count = 0; + printf("Overall stats:"); + xbt_dynar_foreach(all_sizes,count,size) { + if (count==0) + continue; + if (size==0) + continue; + printf("%delm x %d cells; ",count,size); + avg += count * size; + total_count += size; + } + printf("; %f elm per cell\n",avg/(double)total_count); } } @@ -534,11 +672,16 @@ static void traverse(xbt_dict_t head) { xbt_dict_cursor_t cursor=NULL; char *key; char *data; + int i = 0; xbt_dict_foreach(head,cursor,key,data) { - xbt_test_log2("Seen: %s->%s",PRINTF_STR(key),PRINTF_STR(data)); + if (!key || !data || strcmp(key,data)) { + xbt_test_log3("Seen #%d: %s->%s",++i,PRINTF_STR(key),PRINTF_STR(data)); + } else { + xbt_test_log2("Seen #%d: %s",++i,PRINTF_STR(key)); + } xbt_test_assert2(!data || !strcmp(key,data), - "Key(%s) != value(%s). Abording\n",key,data); + "Key(%s) != value(%s). Abording",key,data); } } @@ -561,8 +704,19 @@ static void search_not_found(xbt_dict_t head, const char *data) { } static void count(xbt_dict_t dict, int length) { + xbt_dict_cursor_t cursor; + char *key; + void *data; + int effective = 0; + + xbt_test_add1("Count elements (expecting %d)", length); - xbt_test_assert2(xbt_dict_length(dict) == length, "Length(%d) != %d.", xbt_dict_length(dict), length); + xbt_test_assert2(xbt_dict_length(dict) == length, "Announced length(%d) != %d.", xbt_dict_length(dict), length); + + xbt_dict_foreach(dict,cursor,key,data) { + effective++; + } + xbt_test_assert2(effective == length, "Effective length(%d) != %d.", effective, length); } xbt_ex_t e; @@ -571,8 +725,20 @@ char *data; XBT_TEST_UNIT("basic",test_dict_basic,"Basic usage: change, retrieve, traverse"){ - xbt_test_add0("Traversal the empty dictionnary"); + xbt_test_add0("Traversal the null dictionnary"); + traverse(head); + + xbt_test_add0("Traversal and search the empty dictionnary"); + head = xbt_dict_new(); traverse(head); + TRY { + debuged_remove(head,"12346"); + } CATCH(e) { + if (e.category != not_found_error) + xbt_test_exception(e); + xbt_ex_free(e); + } + xbt_dict_free(&head); xbt_test_add0("Traverse the full dictionnary"); fill(&head); @@ -685,7 +851,7 @@ XBT_TEST_UNIT("remove",test_dict_remove,"Removing some values"){ xbt_ex_free(e); } traverse(head); - xbt_test_add0("Remove all values"); + xbt_test_add0("Free dict, create new fresh one, and then reset the dict"); xbt_dict_free(&head); fill(&head); xbt_dict_reset(head); @@ -711,7 +877,12 @@ XBT_TEST_UNIT("nulldata",test_dict_nulldata,"NULL data management"){ int found=0; xbt_dict_foreach(head,cursor,key,data) { - xbt_test_log2("Seen: %s->%s",PRINTF_STR(key),PRINTF_STR(data)); + if (!key || !data || strcmp(key,data)) { + xbt_test_log2("Seen: %s->%s",PRINTF_STR(key),PRINTF_STR(data)); + } else { + xbt_test_log1("Seen: %s",PRINTF_STR(key)); + } + if (!strcmp(key,"null")) found = 1; } @@ -733,7 +904,7 @@ static int countelems(xbt_dict_t head) { } return res; } - + XBT_TEST_UNIT("crash",test_dict_crash,"Crash test"){ xbt_dict_t head=NULL; int i,j,k, nb; @@ -742,26 +913,32 @@ XBT_TEST_UNIT("crash",test_dict_crash,"Crash test"){ srand((unsigned int)time(NULL)); - xbt_test_add0("CRASH test"); - xbt_test_log0("Fill the struct, count its elems and frees the structure (x10)"); - xbt_test_log1("using 1000 elements with %d chars long randomized keys.",SIZEOFKEY); - for (i=0;i<10;i++) { + xbt_test_add2("CRASH test number %d (%d to go)",i+1,10-i-1); + xbt_test_log0("Fill the struct, count its elems and frees the structure"); + xbt_test_log1("using 1000 elements with %d chars long randomized keys.",SIZEOFKEY); head=xbt_dict_new(); /* if (i%10) printf("."); else printf("%d",i/10); fflush(stdout); */ nb=0; for (j=0;j<1000;j++) { + char *data = NULL; key=xbt_malloc(SIZEOFKEY); - for (k=0;k