1 //===- llvm/ADT/SmallPtrSet.h - 'Normally small' pointer set ----*- C++ -*-===//
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 defines the SmallPtrSet class.  See the doxygen comment for
11 // SmallPtrSetImplBase for more details on the algorithm used.
12 //
13 //===----------------------------------------------------------------------===//
14 
15 #ifndef LLVM_ADT_SMALLPTRSET_H
16 #define LLVM_ADT_SMALLPTRSET_H
17 
18 #include "llvm/Support/Compiler.h"
19 #include "llvm/Support/DataTypes.h"
20 #include "llvm/Support/PointerLikeTypeTraits.h"
21 #include <cassert>
22 #include <cstddef>
23 #include <cstring>
24 #include <iterator>
25 #include <utility>
26 
27 namespace llvm {
28 
29 class SmallPtrSetIteratorImpl;
30 
31 /// SmallPtrSetImplBase - This is the common code shared among all the
32 /// SmallPtrSet<>'s, which is almost everything.  SmallPtrSet has two modes, one
33 /// for small and one for large sets.
34 ///
35 /// Small sets use an array of pointers allocated in the SmallPtrSet object,
36 /// which is treated as a simple array of pointers.  When a pointer is added to
37 /// the set, the array is scanned to see if the element already exists, if not
38 /// the element is 'pushed back' onto the array.  If we run out of space in the
39 /// array, we grow into the 'large set' case.  SmallSet should be used when the
40 /// sets are often small.  In this case, no memory allocation is used, and only
41 /// light-weight and cache-efficient scanning is used.
42 ///
43 /// Large sets use a classic exponentially-probed hash table.  Empty buckets are
44 /// represented with an illegal pointer value (-1) to allow null pointers to be
45 /// inserted.  Tombstones are represented with another illegal pointer value
46 /// (-2), to allow deletion.  The hash table is resized when the table is 3/4 or
47 /// more.  When this happens, the table is doubled in size.
48 ///
49 class SmallPtrSetImplBase {
50   friend class SmallPtrSetIteratorImpl;
51 
52 protected:
53   /// SmallArray - Points to a fixed size set of buckets, used in 'small mode'.
54   const void **SmallArray;
55   /// CurArray - This is the current set of buckets.  If equal to SmallArray,
56   /// then the set is in 'small mode'.
57   const void **CurArray;
58   /// CurArraySize - The allocated size of CurArray, always a power of two.
59   unsigned CurArraySize;
60 
61   // If small, this is # elts allocated consecutively
62   unsigned NumElements;
63   unsigned NumTombstones;
64 
65   // Helpers to copy and move construct a SmallPtrSet.
66   SmallPtrSetImplBase(const void **SmallStorage, const SmallPtrSetImplBase &that);
67   SmallPtrSetImplBase(const void **SmallStorage, unsigned SmallSize,
68                   SmallPtrSetImplBase &&that);
SmallPtrSetImplBase(const void ** SmallStorage,unsigned SmallSize)69   explicit SmallPtrSetImplBase(const void **SmallStorage, unsigned SmallSize) :
70     SmallArray(SmallStorage), CurArray(SmallStorage), CurArraySize(SmallSize) {
71     assert(SmallSize && (SmallSize & (SmallSize-1)) == 0 &&
72            "Initial size must be a power of two!");
73     clear();
74   }
75   ~SmallPtrSetImplBase();
76 
77 public:
78   typedef unsigned size_type;
empty()79   bool LLVM_ATTRIBUTE_UNUSED_RESULT empty() const { return size() == 0; }
size()80   size_type size() const { return NumElements; }
81 
clear()82   void clear() {
83     // If the capacity of the array is huge, and the # elements used is small,
84     // shrink the array.
85     if (!isSmall() && NumElements*4 < CurArraySize && CurArraySize > 32)
86       return shrink_and_clear();
87 
88     // Fill the array with empty markers.
89     memset(CurArray, -1, CurArraySize*sizeof(void*));
90     NumElements = 0;
91     NumTombstones = 0;
92   }
93 
94 protected:
getTombstoneMarker()95   static void *getTombstoneMarker() { return reinterpret_cast<void*>(-2); }
getEmptyMarker()96   static void *getEmptyMarker() {
97     // Note that -1 is chosen to make clear() efficiently implementable with
98     // memset and because it's not a valid pointer value.
99     return reinterpret_cast<void*>(-1);
100   }
101 
102   /// insert_imp - This returns true if the pointer was new to the set, false if
103   /// it was already in the set.  This is hidden from the client so that the
104   /// derived class can check that the right type of pointer is passed in.
105   std::pair<const void *const *, bool> insert_imp(const void *Ptr);
106 
107   /// erase_imp - If the set contains the specified pointer, remove it and
108   /// return true, otherwise return false.  This is hidden from the client so
109   /// that the derived class can check that the right type of pointer is passed
110   /// in.
111   bool erase_imp(const void * Ptr);
112 
count_imp(const void * Ptr)113   bool count_imp(const void * Ptr) const {
114     if (isSmall()) {
115       // Linear search for the item.
116       for (const void *const *APtr = SmallArray,
117                       *const *E = SmallArray+NumElements; APtr != E; ++APtr)
118         if (*APtr == Ptr)
119           return true;
120       return false;
121     }
122 
123     // Big set case.
124     return *FindBucketFor(Ptr) == Ptr;
125   }
126 
127 private:
isSmall()128   bool isSmall() const { return CurArray == SmallArray; }
129 
130   const void * const *FindBucketFor(const void *Ptr) const;
131   void shrink_and_clear();
132 
133   /// Grow - Allocate a larger backing store for the buckets and move it over.
134   void Grow(unsigned NewSize);
135 
136   void operator=(const SmallPtrSetImplBase &RHS) = delete;
137 
138 protected:
139   /// swap - Swaps the elements of two sets.
140   /// Note: This method assumes that both sets have the same small size.
141   void swap(SmallPtrSetImplBase &RHS);
142 
143   void CopyFrom(const SmallPtrSetImplBase &RHS);
144   void MoveFrom(unsigned SmallSize, SmallPtrSetImplBase &&RHS);
145 };
146 
147 /// SmallPtrSetIteratorImpl - This is the common base class shared between all
148 /// instances of SmallPtrSetIterator.
149 class SmallPtrSetIteratorImpl {
150 protected:
151   const void *const *Bucket;
152   const void *const *End;
153 
154 public:
SmallPtrSetIteratorImpl(const void * const * BP,const void * const * E)155   explicit SmallPtrSetIteratorImpl(const void *const *BP, const void*const *E)
156     : Bucket(BP), End(E) {
157       AdvanceIfNotValid();
158   }
159 
160   bool operator==(const SmallPtrSetIteratorImpl &RHS) const {
161     return Bucket == RHS.Bucket;
162   }
163   bool operator!=(const SmallPtrSetIteratorImpl &RHS) const {
164     return Bucket != RHS.Bucket;
165   }
166 
167 protected:
168   /// AdvanceIfNotValid - If the current bucket isn't valid, advance to a bucket
169   /// that is.   This is guaranteed to stop because the end() bucket is marked
170   /// valid.
AdvanceIfNotValid()171   void AdvanceIfNotValid() {
172     assert(Bucket <= End);
173     while (Bucket != End &&
174            (*Bucket == SmallPtrSetImplBase::getEmptyMarker() ||
175             *Bucket == SmallPtrSetImplBase::getTombstoneMarker()))
176       ++Bucket;
177   }
178 };
179 
180 /// SmallPtrSetIterator - This implements a const_iterator for SmallPtrSet.
181 template<typename PtrTy>
182 class SmallPtrSetIterator : public SmallPtrSetIteratorImpl {
183   typedef PointerLikeTypeTraits<PtrTy> PtrTraits;
184 
185 public:
186   typedef PtrTy                     value_type;
187   typedef PtrTy                     reference;
188   typedef PtrTy                     pointer;
189   typedef std::ptrdiff_t            difference_type;
190   typedef std::forward_iterator_tag iterator_category;
191 
SmallPtrSetIterator(const void * const * BP,const void * const * E)192   explicit SmallPtrSetIterator(const void *const *BP, const void *const *E)
193     : SmallPtrSetIteratorImpl(BP, E) {}
194 
195   // Most methods provided by baseclass.
196 
197   const PtrTy operator*() const {
198     assert(Bucket < End);
199     return PtrTraits::getFromVoidPointer(const_cast<void*>(*Bucket));
200   }
201 
202   inline SmallPtrSetIterator& operator++() {          // Preincrement
203     ++Bucket;
204     AdvanceIfNotValid();
205     return *this;
206   }
207 
208   SmallPtrSetIterator operator++(int) {        // Postincrement
209     SmallPtrSetIterator tmp = *this; ++*this; return tmp;
210   }
211 };
212 
213 /// RoundUpToPowerOfTwo - This is a helper template that rounds N up to the next
214 /// power of two (which means N itself if N is already a power of two).
215 template<unsigned N>
216 struct RoundUpToPowerOfTwo;
217 
218 /// RoundUpToPowerOfTwoH - If N is not a power of two, increase it.  This is a
219 /// helper template used to implement RoundUpToPowerOfTwo.
220 template<unsigned N, bool isPowerTwo>
221 struct RoundUpToPowerOfTwoH {
222   enum { Val = N };
223 };
224 template<unsigned N>
225 struct RoundUpToPowerOfTwoH<N, false> {
226   enum {
227     // We could just use NextVal = N+1, but this converges faster.  N|(N-1) sets
228     // the right-most zero bits to one all at once, e.g. 0b0011000 -> 0b0011111.
229     Val = RoundUpToPowerOfTwo<(N|(N-1)) + 1>::Val
230   };
231 };
232 
233 template<unsigned N>
234 struct RoundUpToPowerOfTwo {
235   enum { Val = RoundUpToPowerOfTwoH<N, (N&(N-1)) == 0>::Val };
236 };
237 
238 /// \brief A templated base class for \c SmallPtrSet which provides the
239 /// typesafe interface that is common across all small sizes.
240 ///
241 /// This is particularly useful for passing around between interface boundaries
242 /// to avoid encoding a particular small size in the interface boundary.
243 template <typename PtrType>
244 class SmallPtrSetImpl : public SmallPtrSetImplBase {
245   typedef PointerLikeTypeTraits<PtrType> PtrTraits;
246 
247   SmallPtrSetImpl(const SmallPtrSetImpl &) = delete;
248 
249 protected:
250   // Constructors that forward to the base.
251   SmallPtrSetImpl(const void **SmallStorage, const SmallPtrSetImpl &that)
252       : SmallPtrSetImplBase(SmallStorage, that) {}
253   SmallPtrSetImpl(const void **SmallStorage, unsigned SmallSize,
254                   SmallPtrSetImpl &&that)
255       : SmallPtrSetImplBase(SmallStorage, SmallSize, std::move(that)) {}
256   explicit SmallPtrSetImpl(const void **SmallStorage, unsigned SmallSize)
257       : SmallPtrSetImplBase(SmallStorage, SmallSize) {}
258 
259 public:
260   typedef SmallPtrSetIterator<PtrType> iterator;
261   typedef SmallPtrSetIterator<PtrType> const_iterator;
262 
263   /// Inserts Ptr if and only if there is no element in the container equal to
264   /// Ptr. The bool component of the returned pair is true if and only if the
265   /// insertion takes place, and the iterator component of the pair points to
266   /// the element equal to Ptr.
267   std::pair<iterator, bool> insert(PtrType Ptr) {
268     auto p = insert_imp(PtrTraits::getAsVoidPointer(Ptr));
269     return std::make_pair(iterator(p.first, CurArray + CurArraySize), p.second);
270   }
271 
272   /// erase - If the set contains the specified pointer, remove it and return
273   /// true, otherwise return false.
274   bool erase(PtrType Ptr) {
275     return erase_imp(PtrTraits::getAsVoidPointer(Ptr));
276   }
277 
278   /// count - Return 1 if the specified pointer is in the set, 0 otherwise.
279   size_type count(PtrType Ptr) const {
280     return count_imp(PtrTraits::getAsVoidPointer(Ptr)) ? 1 : 0;
281   }
282 
283   template <typename IterT>
284   void insert(IterT I, IterT E) {
285     for (; I != E; ++I)
286       insert(*I);
287   }
288 
289   inline iterator begin() const {
290     return iterator(CurArray, CurArray+CurArraySize);
291   }
292   inline iterator end() const {
293     return iterator(CurArray+CurArraySize, CurArray+CurArraySize);
294   }
295 };
296 
297 /// SmallPtrSet - This class implements a set which is optimized for holding
298 /// SmallSize or less elements.  This internally rounds up SmallSize to the next
299 /// power of two if it is not already a power of two.  See the comments above
300 /// SmallPtrSetImplBase for details of the algorithm.
301 template<class PtrType, unsigned SmallSize>
302 class SmallPtrSet : public SmallPtrSetImpl<PtrType> {
303   typedef SmallPtrSetImpl<PtrType> BaseT;
304 
305   // Make sure that SmallSize is a power of two, round up if not.
306   enum { SmallSizePowTwo = RoundUpToPowerOfTwo<SmallSize>::Val };
307   /// SmallStorage - Fixed size storage used in 'small mode'.
308   const void *SmallStorage[SmallSizePowTwo];
309 
310 public:
311   SmallPtrSet() : BaseT(SmallStorage, SmallSizePowTwo) {}
312   SmallPtrSet(const SmallPtrSet &that) : BaseT(SmallStorage, that) {}
313   SmallPtrSet(SmallPtrSet &&that)
314       : BaseT(SmallStorage, SmallSizePowTwo, std::move(that)) {}
315 
316   template<typename It>
317   SmallPtrSet(It I, It E) : BaseT(SmallStorage, SmallSizePowTwo) {
318     this->insert(I, E);
319   }
320 
321   SmallPtrSet<PtrType, SmallSize> &
322   operator=(const SmallPtrSet<PtrType, SmallSize> &RHS) {
323     if (&RHS != this)
324       this->CopyFrom(RHS);
325     return *this;
326   }
327 
328   SmallPtrSet<PtrType, SmallSize>&
329   operator=(SmallPtrSet<PtrType, SmallSize> &&RHS) {
330     if (&RHS != this)
331       this->MoveFrom(SmallSizePowTwo, std::move(RHS));
332     return *this;
333   }
334 
335   /// swap - Swaps the elements of two sets.
336   void swap(SmallPtrSet<PtrType, SmallSize> &RHS) {
337     SmallPtrSetImplBase::swap(RHS);
338   }
339 };
340 }
341 
342 namespace std {
343   /// Implement std::swap in terms of SmallPtrSet swap.
344   template<class T, unsigned N>
345   inline void swap(llvm::SmallPtrSet<T, N> &LHS, llvm::SmallPtrSet<T, N> &RHS) {
346     LHS.swap(RHS);
347   }
348 }
349 
350 #endif
351