//
// The LLVM Compiler Infrastructure
//
-// This file was developed by Chris Lattner and is distributed under
-// the University of Illinois Open Source License. See LICENSE.TXT for details.
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
//===----------------------------------------------------------------------===//
#include "llvm/ADT/StringMap.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/Support/Compiler.h"
#include <cassert>
using namespace llvm;
-StringMapVisitor::~StringMapVisitor() {
+StringMapImpl::StringMapImpl(unsigned InitSize, unsigned itemSize) {
+ ItemSize = itemSize;
+
+ // If a size is specified, initialize the table with that many buckets.
+ if (InitSize) {
+ init(InitSize);
+ return;
+ }
+
+ // Otherwise, initialize it with zero buckets to avoid the allocation.
+ TheTable = nullptr;
+ NumBuckets = 0;
+ NumItems = 0;
+ NumTombstones = 0;
}
-StringMapImpl::StringMapImpl(unsigned InitSize, unsigned itemSize) {
+void StringMapImpl::init(unsigned InitSize) {
assert((InitSize & (InitSize-1)) == 0 &&
"Init Size must be a power of 2 or zero!");
- NumBuckets = InitSize ? InitSize : 512;
- ItemSize = itemSize;
+ NumBuckets = InitSize ? InitSize : 16;
NumItems = 0;
+ NumTombstones = 0;
- TheTable = new ItemBucket[NumBuckets]();
- memset(TheTable, 0, NumBuckets*sizeof(ItemBucket));
-}
-
+ TheTable = (StringMapEntryBase **)calloc(NumBuckets+1,
+ sizeof(StringMapEntryBase **) +
+ sizeof(unsigned));
-/// HashString - Compute a hash code for the specified string.
-///
-static unsigned HashString(const char *Start, const char *End) {
- // Bernstein hash function.
- unsigned int Result = 0;
- // TODO: investigate whether a modified bernstein hash function performs
- // better: http://eternallyconfuzzled.com/tuts/algorithms/jsw_tut_hashing.aspx
- // X*33+c -> X*33^c
- while (Start != End)
- Result = Result * 33 + *Start++;
- Result = Result + (Result >> 5);
- return Result;
+ // Allocate one extra bucket, set it to look filled so the iterators stop at
+ // end.
+ TheTable[NumBuckets] = (StringMapEntryBase*)2;
}
+
/// LookupBucketFor - Look up the bucket that the specified string should end
/// up in. If it already exists as a key in the map, the Item pointer for the
/// specified bucket will be non-null. Otherwise, it will be null. In either
/// case, the FullHashValue field of the bucket will be set to the hash value
/// of the string.
-unsigned StringMapImpl::LookupBucketFor(const char *NameStart,
- const char *NameEnd) {
+unsigned StringMapImpl::LookupBucketFor(StringRef Name) {
unsigned HTSize = NumBuckets;
- unsigned FullHashValue = HashString(NameStart, NameEnd);
+ if (HTSize == 0) { // Hash table unallocated so far?
+ init(16);
+ HTSize = NumBuckets;
+ }
+ unsigned FullHashValue = HashString(Name);
unsigned BucketNo = FullHashValue & (HTSize-1);
-
+ unsigned *HashTable = (unsigned *)(TheTable + NumBuckets + 1);
+
unsigned ProbeAmt = 1;
+ int FirstTombstone = -1;
while (1) {
- ItemBucket &Bucket = TheTable[BucketNo];
- StringMapEntryBase *BucketItem = Bucket.Item;
+ StringMapEntryBase *BucketItem = TheTable[BucketNo];
// If we found an empty bucket, this key isn't in the table yet, return it.
- if (BucketItem == 0) {
- Bucket.FullHashValue = FullHashValue;
+ if (LLVM_LIKELY(!BucketItem)) {
+ // If we found a tombstone, we want to reuse the tombstone instead of an
+ // empty bucket. This reduces probing.
+ if (FirstTombstone != -1) {
+ HashTable[FirstTombstone] = FullHashValue;
+ return FirstTombstone;
+ }
+
+ HashTable[BucketNo] = FullHashValue;
return BucketNo;
}
- // If the full hash value matches, check deeply for a match. The common
- // case here is that we are only looking at the buckets (for item info
- // being non-null and for the full hash value) not at the items. This
- // is important for cache locality.
- if (Bucket.FullHashValue == FullHashValue) {
+ if (BucketItem == getTombstoneVal()) {
+ // Skip over tombstones. However, remember the first one we see.
+ if (FirstTombstone == -1) FirstTombstone = BucketNo;
+ } else if (LLVM_LIKELY(HashTable[BucketNo] == FullHashValue)) {
+ // If the full hash value matches, check deeply for a match. The common
+ // case here is that we are only looking at the buckets (for item info
+ // being non-null and for the full hash value) not at the items. This
+ // is important for cache locality.
+
+ // Do the comparison like this because Name isn't necessarily
+ // null-terminated!
+ char *ItemStr = (char*)BucketItem+ItemSize;
+ if (Name == StringRef(ItemStr, BucketItem->getKeyLength())) {
+ // We found a match!
+ return BucketNo;
+ }
+ }
+
+ // Okay, we didn't find the item. Probe to the next bucket.
+ BucketNo = (BucketNo+ProbeAmt) & (HTSize-1);
+
+ // Use quadratic probing, it has fewer clumping artifacts than linear
+ // probing and has good cache behavior in the common case.
+ ++ProbeAmt;
+ }
+}
+
+
+/// FindKey - Look up the bucket that contains the specified key. If it exists
+/// in the map, return the bucket number of the key. Otherwise return -1.
+/// This does not modify the map.
+int StringMapImpl::FindKey(StringRef Key) const {
+ unsigned HTSize = NumBuckets;
+ if (HTSize == 0) return -1; // Really empty table?
+ unsigned FullHashValue = HashString(Key);
+ unsigned BucketNo = FullHashValue & (HTSize-1);
+ unsigned *HashTable = (unsigned *)(TheTable + NumBuckets + 1);
+
+ unsigned ProbeAmt = 1;
+ while (1) {
+ StringMapEntryBase *BucketItem = TheTable[BucketNo];
+ // If we found an empty bucket, this key isn't in the table yet, return.
+ if (LLVM_LIKELY(!BucketItem))
+ return -1;
+
+ if (BucketItem == getTombstoneVal()) {
+ // Ignore tombstones.
+ } else if (LLVM_LIKELY(HashTable[BucketNo] == FullHashValue)) {
+ // If the full hash value matches, check deeply for a match. The common
+ // case here is that we are only looking at the buckets (for item info
+ // being non-null and for the full hash value) not at the items. This
+ // is important for cache locality.
+
// Do the comparison like this because NameStart isn't necessarily
// null-terminated!
char *ItemStr = (char*)BucketItem+ItemSize;
- unsigned ItemStrLen = BucketItem->getKeyLength();
- if (unsigned(NameEnd-NameStart) == ItemStrLen &&
- memcmp(ItemStr, NameStart, ItemStrLen) == 0) {
+ if (Key == StringRef(ItemStr, BucketItem->getKeyLength())) {
// We found a match!
return BucketNo;
}
}
}
+/// RemoveKey - Remove the specified StringMapEntry from the table, but do not
+/// delete it. This aborts if the value isn't in the table.
+void StringMapImpl::RemoveKey(StringMapEntryBase *V) {
+ const char *VStr = (char*)V + ItemSize;
+ StringMapEntryBase *V2 = RemoveKey(StringRef(VStr, V->getKeyLength()));
+ (void)V2;
+ assert(V == V2 && "Didn't find key?");
+}
+
+/// RemoveKey - Remove the StringMapEntry for the specified key from the
+/// table, returning it. If the key is not in the table, this returns null.
+StringMapEntryBase *StringMapImpl::RemoveKey(StringRef Key) {
+ int Bucket = FindKey(Key);
+ if (Bucket == -1) return nullptr;
+
+ StringMapEntryBase *Result = TheTable[Bucket];
+ TheTable[Bucket] = getTombstoneVal();
+ --NumItems;
+ ++NumTombstones;
+ assert(NumItems + NumTombstones <= NumBuckets);
+
+ return Result;
+}
+
+
+
/// RehashTable - Grow the table, redistributing values into the buckets with
/// the appropriate mod-of-hashtable-size.
-void StringMapImpl::RehashTable() {
- unsigned NewSize = NumBuckets*2;
- ItemBucket *NewTableArray = new ItemBucket[NewSize]();
- memset(NewTableArray, 0, NewSize*sizeof(ItemBucket));
-
+unsigned StringMapImpl::RehashTable(unsigned BucketNo) {
+ unsigned NewSize;
+ unsigned *HashTable = (unsigned *)(TheTable + NumBuckets + 1);
+
+ // If the hash table is now more than 3/4 full, or if fewer than 1/8 of
+ // the buckets are empty (meaning that many are filled with tombstones),
+ // grow/rehash the table.
+ if (NumItems*4 > NumBuckets*3) {
+ NewSize = NumBuckets*2;
+ } else if (NumBuckets-(NumItems+NumTombstones) <= NumBuckets/8) {
+ NewSize = NumBuckets;
+ } else {
+ return BucketNo;
+ }
+
+ unsigned NewBucketNo = BucketNo;
+ // Allocate one extra bucket which will always be non-empty. This allows the
+ // iterators to stop at end.
+ StringMapEntryBase **NewTableArray =
+ (StringMapEntryBase **)calloc(NewSize+1, sizeof(StringMapEntryBase *) +
+ sizeof(unsigned));
+ unsigned *NewHashArray = (unsigned *)(NewTableArray + NewSize + 1);
+ NewTableArray[NewSize] = (StringMapEntryBase*)2;
+
// Rehash all the items into their new buckets. Luckily :) we already have
// the hash values available, so we don't have to rehash any strings.
- for (ItemBucket *IB = TheTable, *E = TheTable+NumBuckets; IB != E; ++IB) {
- if (IB->Item) {
+ for (unsigned I = 0, E = NumBuckets; I != E; ++I) {
+ StringMapEntryBase *Bucket = TheTable[I];
+ if (Bucket && Bucket != getTombstoneVal()) {
// Fast case, bucket available.
- unsigned FullHash = IB->FullHashValue;
+ unsigned FullHash = HashTable[I];
unsigned NewBucket = FullHash & (NewSize-1);
- if (NewTableArray[NewBucket].Item == 0) {
- NewTableArray[FullHash & (NewSize-1)].Item = IB->Item;
- NewTableArray[FullHash & (NewSize-1)].FullHashValue = FullHash;
+ if (!NewTableArray[NewBucket]) {
+ NewTableArray[FullHash & (NewSize-1)] = Bucket;
+ NewHashArray[FullHash & (NewSize-1)] = FullHash;
+ if (I == BucketNo)
+ NewBucketNo = NewBucket;
continue;
}
+ // Otherwise probe for a spot.
unsigned ProbeSize = 1;
do {
NewBucket = (NewBucket + ProbeSize++) & (NewSize-1);
- } while (NewTableArray[NewBucket].Item);
+ } while (NewTableArray[NewBucket]);
// Finally found a slot. Fill it in.
- NewTableArray[NewBucket].Item = IB->Item;
- NewTableArray[NewBucket].FullHashValue = FullHash;
+ NewTableArray[NewBucket] = Bucket;
+ NewHashArray[NewBucket] = FullHash;
+ if (I == BucketNo)
+ NewBucketNo = NewBucket;
}
}
- delete[] TheTable;
+ free(TheTable);
TheTable = NewTableArray;
NumBuckets = NewSize;
-}
-
-
-/// VisitEntries - This method walks through all of the items,
-/// invoking Visitor.Visit for each of them.
-void StringMapImpl::VisitEntries(const StringMapVisitor &Visitor) const {
- for (ItemBucket *IB = TheTable, *E = TheTable+NumBuckets; IB != E; ++IB) {
- if (StringMapEntryBase *Id = IB->Item)
- Visitor.Visit((char*)Id + ItemSize, Id);
- }
+ NumTombstones = 0;
+ return NewBucketNo;
}