1 //===--- StringMap.cpp - String Hash table map implementation -------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements the StringMap class.
11 //
12 //===----------------------------------------------------------------------===//
13
14 #include "llvm/ADT/StringMap.h"
15 #include "llvm/ADT/StringExtras.h"
16 #include "llvm/Support/Compiler.h"
17 #include "llvm/Support/DJB.h"
18 #include "llvm/Support/MathExtras.h"
19 #include <cassert>
20
21 using namespace llvm;
22
23 /// Returns the number of buckets to allocate to ensure that the DenseMap can
24 /// accommodate \p NumEntries without need to grow().
getMinBucketToReserveForEntries(unsigned NumEntries)25 static unsigned getMinBucketToReserveForEntries(unsigned NumEntries) {
26 // Ensure that "NumEntries * 4 < NumBuckets * 3"
27 if (NumEntries == 0)
28 return 0;
29 // +1 is required because of the strict equality.
30 // For example if NumEntries is 48, we need to return 401.
31 return NextPowerOf2(NumEntries * 4 / 3 + 1);
32 }
33
StringMapImpl(unsigned InitSize,unsigned itemSize)34 StringMapImpl::StringMapImpl(unsigned InitSize, unsigned itemSize) {
35 ItemSize = itemSize;
36
37 // If a size is specified, initialize the table with that many buckets.
38 if (InitSize) {
39 // The table will grow when the number of entries reach 3/4 of the number of
40 // buckets. To guarantee that "InitSize" number of entries can be inserted
41 // in the table without growing, we allocate just what is needed here.
42 init(getMinBucketToReserveForEntries(InitSize));
43 return;
44 }
45
46 // Otherwise, initialize it with zero buckets to avoid the allocation.
47 TheTable = nullptr;
48 NumBuckets = 0;
49 NumItems = 0;
50 NumTombstones = 0;
51 }
52
init(unsigned InitSize)53 void StringMapImpl::init(unsigned InitSize) {
54 assert((InitSize & (InitSize-1)) == 0 &&
55 "Init Size must be a power of 2 or zero!");
56
57 unsigned NewNumBuckets = InitSize ? InitSize : 16;
58 NumItems = 0;
59 NumTombstones = 0;
60
61 TheTable = static_cast<StringMapEntryBase **>(
62 safe_calloc(NewNumBuckets+1,
63 sizeof(StringMapEntryBase **) + sizeof(unsigned)));
64
65 // Set the member only if TheTable was successfully allocated
66 NumBuckets = NewNumBuckets;
67
68 // Allocate one extra bucket, set it to look filled so the iterators stop at
69 // end.
70 TheTable[NumBuckets] = (StringMapEntryBase*)2;
71 }
72
73 /// LookupBucketFor - Look up the bucket that the specified string should end
74 /// up in. If it already exists as a key in the map, the Item pointer for the
75 /// specified bucket will be non-null. Otherwise, it will be null. In either
76 /// case, the FullHashValue field of the bucket will be set to the hash value
77 /// of the string.
LookupBucketFor(StringRef Name)78 unsigned StringMapImpl::LookupBucketFor(StringRef Name) {
79 unsigned HTSize = NumBuckets;
80 if (HTSize == 0) { // Hash table unallocated so far?
81 init(16);
82 HTSize = NumBuckets;
83 }
84 unsigned FullHashValue = djbHash(Name, 0);
85 unsigned BucketNo = FullHashValue & (HTSize-1);
86 unsigned *HashTable = (unsigned *)(TheTable + NumBuckets + 1);
87
88 unsigned ProbeAmt = 1;
89 int FirstTombstone = -1;
90 while (true) {
91 StringMapEntryBase *BucketItem = TheTable[BucketNo];
92 // If we found an empty bucket, this key isn't in the table yet, return it.
93 if (LLVM_LIKELY(!BucketItem)) {
94 // If we found a tombstone, we want to reuse the tombstone instead of an
95 // empty bucket. This reduces probing.
96 if (FirstTombstone != -1) {
97 HashTable[FirstTombstone] = FullHashValue;
98 return FirstTombstone;
99 }
100
101 HashTable[BucketNo] = FullHashValue;
102 return BucketNo;
103 }
104
105 if (BucketItem == getTombstoneVal()) {
106 // Skip over tombstones. However, remember the first one we see.
107 if (FirstTombstone == -1) FirstTombstone = BucketNo;
108 } else if (LLVM_LIKELY(HashTable[BucketNo] == FullHashValue)) {
109 // If the full hash value matches, check deeply for a match. The common
110 // case here is that we are only looking at the buckets (for item info
111 // being non-null and for the full hash value) not at the items. This
112 // is important for cache locality.
113
114 // Do the comparison like this because Name isn't necessarily
115 // null-terminated!
116 char *ItemStr = (char*)BucketItem+ItemSize;
117 if (Name == StringRef(ItemStr, BucketItem->getKeyLength())) {
118 // We found a match!
119 return BucketNo;
120 }
121 }
122
123 // Okay, we didn't find the item. Probe to the next bucket.
124 BucketNo = (BucketNo+ProbeAmt) & (HTSize-1);
125
126 // Use quadratic probing, it has fewer clumping artifacts than linear
127 // probing and has good cache behavior in the common case.
128 ++ProbeAmt;
129 }
130 }
131
132 /// FindKey - Look up the bucket that contains the specified key. If it exists
133 /// in the map, return the bucket number of the key. Otherwise return -1.
134 /// This does not modify the map.
FindKey(StringRef Key) const135 int StringMapImpl::FindKey(StringRef Key) const {
136 unsigned HTSize = NumBuckets;
137 if (HTSize == 0) return -1; // Really empty table?
138 unsigned FullHashValue = djbHash(Key, 0);
139 unsigned BucketNo = FullHashValue & (HTSize-1);
140 unsigned *HashTable = (unsigned *)(TheTable + NumBuckets + 1);
141
142 unsigned ProbeAmt = 1;
143 while (true) {
144 StringMapEntryBase *BucketItem = TheTable[BucketNo];
145 // If we found an empty bucket, this key isn't in the table yet, return.
146 if (LLVM_LIKELY(!BucketItem))
147 return -1;
148
149 if (BucketItem == getTombstoneVal()) {
150 // Ignore tombstones.
151 } else if (LLVM_LIKELY(HashTable[BucketNo] == FullHashValue)) {
152 // If the full hash value matches, check deeply for a match. The common
153 // case here is that we are only looking at the buckets (for item info
154 // being non-null and for the full hash value) not at the items. This
155 // is important for cache locality.
156
157 // Do the comparison like this because NameStart isn't necessarily
158 // null-terminated!
159 char *ItemStr = (char*)BucketItem+ItemSize;
160 if (Key == StringRef(ItemStr, BucketItem->getKeyLength())) {
161 // We found a match!
162 return BucketNo;
163 }
164 }
165
166 // Okay, we didn't find the item. Probe to the next bucket.
167 BucketNo = (BucketNo+ProbeAmt) & (HTSize-1);
168
169 // Use quadratic probing, it has fewer clumping artifacts than linear
170 // probing and has good cache behavior in the common case.
171 ++ProbeAmt;
172 }
173 }
174
175 /// RemoveKey - Remove the specified StringMapEntry from the table, but do not
176 /// delete it. This aborts if the value isn't in the table.
RemoveKey(StringMapEntryBase * V)177 void StringMapImpl::RemoveKey(StringMapEntryBase *V) {
178 const char *VStr = (char*)V + ItemSize;
179 StringMapEntryBase *V2 = RemoveKey(StringRef(VStr, V->getKeyLength()));
180 (void)V2;
181 assert(V == V2 && "Didn't find key?");
182 }
183
184 /// RemoveKey - Remove the StringMapEntry for the specified key from the
185 /// table, returning it. If the key is not in the table, this returns null.
RemoveKey(StringRef Key)186 StringMapEntryBase *StringMapImpl::RemoveKey(StringRef Key) {
187 int Bucket = FindKey(Key);
188 if (Bucket == -1) return nullptr;
189
190 StringMapEntryBase *Result = TheTable[Bucket];
191 TheTable[Bucket] = getTombstoneVal();
192 --NumItems;
193 ++NumTombstones;
194 assert(NumItems + NumTombstones <= NumBuckets);
195
196 return Result;
197 }
198
199 /// RehashTable - Grow the table, redistributing values into the buckets with
200 /// the appropriate mod-of-hashtable-size.
RehashTable(unsigned BucketNo)201 unsigned StringMapImpl::RehashTable(unsigned BucketNo) {
202 unsigned NewSize;
203 unsigned *HashTable = (unsigned *)(TheTable + NumBuckets + 1);
204
205 // If the hash table is now more than 3/4 full, or if fewer than 1/8 of
206 // the buckets are empty (meaning that many are filled with tombstones),
207 // grow/rehash the table.
208 if (LLVM_UNLIKELY(NumItems * 4 > NumBuckets * 3)) {
209 NewSize = NumBuckets*2;
210 } else if (LLVM_UNLIKELY(NumBuckets - (NumItems + NumTombstones) <=
211 NumBuckets / 8)) {
212 NewSize = NumBuckets;
213 } else {
214 return BucketNo;
215 }
216
217 unsigned NewBucketNo = BucketNo;
218 // Allocate one extra bucket which will always be non-empty. This allows the
219 // iterators to stop at end.
220 auto NewTableArray = static_cast<StringMapEntryBase **>(
221 safe_calloc(NewSize+1, sizeof(StringMapEntryBase *) + sizeof(unsigned)));
222
223 unsigned *NewHashArray = (unsigned *)(NewTableArray + NewSize + 1);
224 NewTableArray[NewSize] = (StringMapEntryBase*)2;
225
226 // Rehash all the items into their new buckets. Luckily :) we already have
227 // the hash values available, so we don't have to rehash any strings.
228 for (unsigned I = 0, E = NumBuckets; I != E; ++I) {
229 StringMapEntryBase *Bucket = TheTable[I];
230 if (Bucket && Bucket != getTombstoneVal()) {
231 // Fast case, bucket available.
232 unsigned FullHash = HashTable[I];
233 unsigned NewBucket = FullHash & (NewSize-1);
234 if (!NewTableArray[NewBucket]) {
235 NewTableArray[FullHash & (NewSize-1)] = Bucket;
236 NewHashArray[FullHash & (NewSize-1)] = FullHash;
237 if (I == BucketNo)
238 NewBucketNo = NewBucket;
239 continue;
240 }
241
242 // Otherwise probe for a spot.
243 unsigned ProbeSize = 1;
244 do {
245 NewBucket = (NewBucket + ProbeSize++) & (NewSize-1);
246 } while (NewTableArray[NewBucket]);
247
248 // Finally found a slot. Fill it in.
249 NewTableArray[NewBucket] = Bucket;
250 NewHashArray[NewBucket] = FullHash;
251 if (I == BucketNo)
252 NewBucketNo = NewBucket;
253 }
254 }
255
256 free(TheTable);
257
258 TheTable = NewTableArray;
259 NumBuckets = NewSize;
260 NumTombstones = 0;
261 return NewBucketNo;
262 }
263