/* * Copyright (C) 2008 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ /* * Hash table. The dominant calls are add and lookup, with removals * happening very infrequently. We use probing, and don't worry much * about tombstone removal. */ #include "Dalvik.h" #include <stdlib.h> /* table load factor, i.e. how full can it get before we resize */ //#define LOAD_NUMER 3 // 75% //#define LOAD_DENOM 4 #define LOAD_NUMER 5 // 62.5% #define LOAD_DENOM 8 //#define LOAD_NUMER 1 // 50% //#define LOAD_DENOM 2 /* * Compute the capacity needed for a table to hold "size" elements. */ size_t dvmHashSize(size_t size) { return (size * LOAD_DENOM) / LOAD_NUMER +1; } /* * Create and initialize a hash table. */ HashTable* dvmHashTableCreate(size_t initialSize, HashFreeFunc freeFunc) { HashTable* pHashTable; assert(initialSize > 0); pHashTable = (HashTable*) malloc(sizeof(*pHashTable)); if (pHashTable == NULL) return NULL; dvmInitMutex(&pHashTable->lock); pHashTable->tableSize = dexRoundUpPower2(initialSize); pHashTable->numEntries = pHashTable->numDeadEntries = 0; pHashTable->freeFunc = freeFunc; pHashTable->pEntries = (HashEntry*) calloc(pHashTable->tableSize, sizeof(HashEntry)); if (pHashTable->pEntries == NULL) { free(pHashTable); return NULL; } return pHashTable; } /* * Clear out all entries. */ void dvmHashTableClear(HashTable* pHashTable) { HashEntry* pEnt; int i; pEnt = pHashTable->pEntries; for (i = 0; i < pHashTable->tableSize; i++, pEnt++) { if (pEnt->data == HASH_TOMBSTONE) { // nuke entry pEnt->data = NULL; } else if (pEnt->data != NULL) { // call free func then nuke entry if (pHashTable->freeFunc != NULL) (*pHashTable->freeFunc)(pEnt->data); pEnt->data = NULL; } } pHashTable->numEntries = 0; pHashTable->numDeadEntries = 0; } /* * Free the table. */ void dvmHashTableFree(HashTable* pHashTable) { if (pHashTable == NULL) return; dvmHashTableClear(pHashTable); free(pHashTable->pEntries); free(pHashTable); } #ifndef NDEBUG /* * Count up the number of tombstone entries in the hash table. */ static int countTombStones(HashTable* pHashTable) { int i, count; for (count = i = 0; i < pHashTable->tableSize; i++) { if (pHashTable->pEntries[i].data == HASH_TOMBSTONE) count++; } return count; } #endif /* * Resize a hash table. We do this when adding an entry increased the * size of the table beyond its comfy limit. * * This essentially requires re-inserting all elements into the new storage. * * If multiple threads can access the hash table, the table's lock should * have been grabbed before issuing the "lookup+add" call that led to the * resize, so we don't have a synchronization problem here. */ static bool resizeHash(HashTable* pHashTable, int newSize) { HashEntry* pNewEntries; int i; assert(countTombStones(pHashTable) == pHashTable->numDeadEntries); //ALOGI("before: dead=%d", pHashTable->numDeadEntries); pNewEntries = (HashEntry*) calloc(newSize, sizeof(HashEntry)); if (pNewEntries == NULL) return false; for (i = 0; i < pHashTable->tableSize; i++) { void* data = pHashTable->pEntries[i].data; if (data != NULL && data != HASH_TOMBSTONE) { int hashValue = pHashTable->pEntries[i].hashValue; int newIdx; /* probe for new spot, wrapping around */ newIdx = hashValue & (newSize-1); while (pNewEntries[newIdx].data != NULL) newIdx = (newIdx + 1) & (newSize-1); pNewEntries[newIdx].hashValue = hashValue; pNewEntries[newIdx].data = data; } } free(pHashTable->pEntries); pHashTable->pEntries = pNewEntries; pHashTable->tableSize = newSize; pHashTable->numDeadEntries = 0; assert(countTombStones(pHashTable) == 0); return true; } /* * Look up an entry. * * We probe on collisions, wrapping around the table. */ void* dvmHashTableLookup(HashTable* pHashTable, u4 itemHash, void* item, HashCompareFunc cmpFunc, bool doAdd) { HashEntry* pEntry; HashEntry* pEnd; void* result = NULL; assert(pHashTable->tableSize > 0); assert(item != HASH_TOMBSTONE); assert(item != NULL); /* jump to the first entry and probe for a match */ pEntry = &pHashTable->pEntries[itemHash & (pHashTable->tableSize-1)]; pEnd = &pHashTable->pEntries[pHashTable->tableSize]; while (pEntry->data != NULL) { if (pEntry->data != HASH_TOMBSTONE && pEntry->hashValue == itemHash && (*cmpFunc)(pEntry->data, item) == 0) { /* match */ //ALOGD("+++ match on entry %d", pEntry - pHashTable->pEntries); break; } pEntry++; if (pEntry == pEnd) { /* wrap around to start */ if (pHashTable->tableSize == 1) break; /* edge case - single-entry table */ pEntry = pHashTable->pEntries; } //ALOGI("+++ look probing %d...", pEntry - pHashTable->pEntries); } if (pEntry->data == NULL) { if (doAdd) { pEntry->hashValue = itemHash; pEntry->data = item; pHashTable->numEntries++; /* * We've added an entry. See if this brings us too close to full. */ if ((pHashTable->numEntries+pHashTable->numDeadEntries) * LOAD_DENOM > pHashTable->tableSize * LOAD_NUMER) { if (!resizeHash(pHashTable, pHashTable->tableSize * 2)) { /* don't really have a way to indicate failure */ ALOGE("Dalvik hash resize failure"); dvmAbort(); } /* note "pEntry" is now invalid */ } else { //ALOGW("okay %d/%d/%d", // pHashTable->numEntries, pHashTable->tableSize, // (pHashTable->tableSize * LOAD_NUMER) / LOAD_DENOM); } /* full table is bad -- search for nonexistent never halts */ assert(pHashTable->numEntries < pHashTable->tableSize); result = item; } else { assert(result == NULL); } } else { result = pEntry->data; } return result; } /* * Remove an entry from the table. * * Does NOT invoke the "free" function on the item. */ bool dvmHashTableRemove(HashTable* pHashTable, u4 itemHash, void* item) { HashEntry* pEntry; HashEntry* pEnd; assert(pHashTable->tableSize > 0); /* jump to the first entry and probe for a match */ pEntry = &pHashTable->pEntries[itemHash & (pHashTable->tableSize-1)]; pEnd = &pHashTable->pEntries[pHashTable->tableSize]; while (pEntry->data != NULL) { if (pEntry->data == item) { //ALOGI("+++ stepping on entry %d", pEntry - pHashTable->pEntries); pEntry->data = HASH_TOMBSTONE; pHashTable->numEntries--; pHashTable->numDeadEntries++; return true; } pEntry++; if (pEntry == pEnd) { /* wrap around to start */ if (pHashTable->tableSize == 1) break; /* edge case - single-entry table */ pEntry = pHashTable->pEntries; } //ALOGI("+++ del probing %d...", pEntry - pHashTable->pEntries); } return false; } /* * Scan every entry in the hash table and evaluate it with the specified * indirect function call. If the function returns 1, remove the entry from * the table. * * Does NOT invoke the "free" function on the item. * * Returning values other than 0 or 1 will abort the routine. */ int dvmHashForeachRemove(HashTable* pHashTable, HashForeachRemoveFunc func) { int i, val, tableSize; tableSize = pHashTable->tableSize; for (i = 0; i < tableSize; i++) { HashEntry* pEnt = &pHashTable->pEntries[i]; if (pEnt->data != NULL && pEnt->data != HASH_TOMBSTONE) { val = (*func)(pEnt->data); if (val == 1) { pEnt->data = HASH_TOMBSTONE; pHashTable->numEntries--; pHashTable->numDeadEntries++; } else if (val != 0) { return val; } } } return 0; } /* * Execute a function on every entry in the hash table. * * If "func" returns a nonzero value, terminate early and return the value. */ int dvmHashForeach(HashTable* pHashTable, HashForeachFunc func, void* arg) { int i, val, tableSize; tableSize = pHashTable->tableSize; for (i = 0; i < tableSize; i++) { HashEntry* pEnt = &pHashTable->pEntries[i]; if (pEnt->data != NULL && pEnt->data != HASH_TOMBSTONE) { val = (*func)(pEnt->data, arg); if (val != 0) return val; } } return 0; } /* * Look up an entry, counting the number of times we have to probe. * * Returns -1 if the entry wasn't found. */ static int countProbes(HashTable* pHashTable, u4 itemHash, const void* item, HashCompareFunc cmpFunc) { HashEntry* pEntry; HashEntry* pEnd; int count = 0; assert(pHashTable->tableSize > 0); assert(item != HASH_TOMBSTONE); assert(item != NULL); /* jump to the first entry and probe for a match */ pEntry = &pHashTable->pEntries[itemHash & (pHashTable->tableSize-1)]; pEnd = &pHashTable->pEntries[pHashTable->tableSize]; while (pEntry->data != NULL) { if (pEntry->data != HASH_TOMBSTONE && pEntry->hashValue == itemHash && (*cmpFunc)(pEntry->data, item) == 0) { /* match */ break; } pEntry++; if (pEntry == pEnd) { /* wrap around to start */ if (pHashTable->tableSize == 1) break; /* edge case - single-entry table */ pEntry = pHashTable->pEntries; } count++; } if (pEntry->data == NULL) return -1; return count; } /* * Evaluate the amount of probing required for the specified hash table. * * We do this by running through all entries in the hash table, computing * the hash value and then doing a lookup. * * The caller should lock the table before calling here. */ void dvmHashTableProbeCount(HashTable* pHashTable, HashCalcFunc calcFunc, HashCompareFunc cmpFunc) { int numEntries, minProbe, maxProbe, totalProbe; HashIter iter; numEntries = maxProbe = totalProbe = 0; minProbe = 65536*32767; for (dvmHashIterBegin(pHashTable, &iter); !dvmHashIterDone(&iter); dvmHashIterNext(&iter)) { const void* data = (const void*)dvmHashIterData(&iter); int count; count = countProbes(pHashTable, (*calcFunc)(data), data, cmpFunc); numEntries++; if (count < minProbe) minProbe = count; if (count > maxProbe) maxProbe = count; totalProbe += count; } ALOGI("Probe: min=%d max=%d, total=%d in %d (%d), avg=%.3f", minProbe, maxProbe, totalProbe, numEntries, pHashTable->tableSize, (float) totalProbe / (float) numEntries); }