From 0e5af195f6c54dbf5a24a1ec12ed2d0bd02f5b7f Mon Sep 17 00:00:00 2001 From: Jim Laskey Date: Fri, 27 Oct 2006 16:16:16 +0000 Subject: [PATCH] Breakout folding hash set from SelectionDAGCSEMap. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31215 91177308-0d34-0410-b5e6-96231b3b80d8 --- include/llvm/ADT/FoldingSet.h | 281 +++++++++++++++++++++++++++++++++ lib/Support/FoldingSet.cpp | 282 ++++++++++++++++++++++++++++++++++ 2 files changed, 563 insertions(+) create mode 100644 include/llvm/ADT/FoldingSet.h create mode 100644 lib/Support/FoldingSet.cpp diff --git a/include/llvm/ADT/FoldingSet.h b/include/llvm/ADT/FoldingSet.h new file mode 100644 index 00000000000..c808d3bebb5 --- /dev/null +++ b/include/llvm/ADT/FoldingSet.h @@ -0,0 +1,281 @@ +//===-- llvm/ADT/FoldingSet.h - Uniquing Hash Set ---------------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file was developed by James M. Laskey and is distributed under +// the University of Illinois Open Source License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file defines a hash set that can be used to remove duplication of nodes +// in a graph. This code was originally created by Chris Lattner for use with +// SelectionDAGCSEMap, but was isolated to provide use across the llvm code set. +// +//===----------------------------------------------------------------------===// + +#ifndef LLVM_ADT_FOLDINGSET_H +#define LLVM_ADT_FOLDINGSET_H + +#include "llvm/ADT/SmallVector.h" + +namespace llvm { + +/// This folding set used for two purposes: +/// 1. Given information about a node we want to create, look up the unique +/// instance of the node in the set. If the node already exists, return +/// it, otherwise return the bucket it should be inserted into. +/// 2. Given a node that has already been created, remove it from the set. +/// +/// This class is implemented as a single-link chained hash table, where the +/// "buckets" are actually the nodes themselves (the next pointer is in the +/// node). The last node points back to the bucket to simplified node removal. +/// +/// Any node that is to be included in the folding set must be a subclass of +/// FoldingSetNode. The node class must also define a Profile method used to +/// establish the unique bits of data for the node. The Profile method is +/// passed a FoldingSetNodeID object which is used to gather the bits. Just +/// call one of the Add* functions defined in the FoldingSetImpl::NodeID class. +/// +/// Eg. +/// class MyNode : public FoldingSetNode { +/// private: +/// std::string Name; +/// unsigned Value; +/// public: +/// MyNode(const char *N, unsigned V) : Name(N), Value(V) {} +/// ... +/// void Profile(FoldingSetNodeID &ID) { +/// ID.AddString(Name); +/// ID.AddInteger(Value); +/// } +/// ... +/// }; +/// +/// To define the folding set itself use the FoldingSet template; +/// +/// Eg. +/// FoldingSet MyFoldingSet; +/// +/// Four public methods are available to manipulate the folding set; +/// +/// 1) If you have an existing node that you want add to the set but unsure +/// that the node might already exist then call; +/// +/// MyNode *M = MyFoldingSet.GetOrInsertNode(N); +/// +/// If The result is equal to the input then the node has been inserted. +/// Otherwise, the result is the node existing in the folding set, and the +/// input can be discarded (use the result instead.) +/// +/// 2) If you are ready to construct a node but want to check if it already +/// exists, then call FindNodeOrInsertPos with a FoldingSetNodeID of the bits to +/// check; +/// +/// FoldingSetNodeID ID; +/// ID.AddString(Name); +/// ID.AddInteger(Value); +/// void *InsertPoint; +/// +/// MyNode *M = MyFoldingSet.FindNodeOrInsertPos(ID, InsertPoint); +/// +/// If found then M with be non-NULL, else InsertPoint will point to where it +/// should be inserted using InsertNode. +/// +/// 3) If you get a NULL result from FindNodeOrInsertPos then you can ass a new +/// node with FindNodeOrInsertPos; +/// +/// InsertNode(N, InsertPoint); +/// +/// 4) Finally, if you want to remove a node from the folding set call; +/// +/// bool WasRemoved = RemoveNode(N); +/// +/// The result indicates whether the node did exist in the folding set. + + +//===----------------------------------------------------------------------===// +/// FoldingSetImpl - Implements the folding set functionality. The main +/// structure is an array of buckets. Each bucket is indexed by the hash of +/// the nodes it contains. The bucket itself points to the nodes contained +/// in the bucket via a singly linked list. The last node in the list points +/// back to the bucket to facilitate node removal. +/// +class FoldingSetImpl { +private: + // Buckets - Array of bucket chains. + void **Buckets; + + // NumBuckets - Length of the Buckets array. Always a power of 2. + unsigned NumBuckets; + + // NumNodes - Number of nodes in the folding set. Growth occurs when NumNodes + // is greater than twice teh number of buckets. + unsigned NumNodes; + +public: + FoldingSetImpl(); + ~FoldingSetImpl(); + + // Forward declaration. + class Node; + + //===--------------------------------------------------------------------===// + /// NodeID - This class is used to gather all the unique data bits of a + /// node. When all the bits are gathered this class is used to produce a + /// hash value for the node. + /// + class NodeID { + /// Bits - Vector of all the data bits that make the node unique. + /// Use a SmallVector to avoid a heap allocation in the common case. + SmallVector Bits; + + public: + NodeID() {} + + /// getRawData - Return the ith entry in the Bits data. + /// + unsigned getRawData(unsigned i) const { + return Bits[i]; + } + + /// Add* - Add various data types to Bit data. + /// + void AddPointer(const void *Ptr); + void AddInteger(signed I); + void AddInteger(unsigned I); + void AddInteger(uint64_t I); + void AddFloat(float F); + void AddDouble(double D); + void AddString(const std::string &String); + + /// ComputeHash - Compute a strong hash value for this NodeID, used to + /// lookup the node in the FoldingSetImpl. + unsigned ComputeHash() const; + + /// operator== - Used to compare two nodes to each other. + /// + bool operator==(const NodeID &RHS) const; + }; + + //===--------------------------------------------------------------------===// + /// Node - This class is used to maintain the singly linked bucket list in + /// a folding set. + /// + class Node { + private: + // nextInBucket - next linek in the bucket list. + void *nextInBucket; + + public: + + Node() : nextInBucket(0) {} + + // Accessors + void *getNextInBucket() const { return nextInBucket; } + void SetNextInBucket(void *N) { nextInBucket = N; } + }; + + /// RemoveNode - Remove a node from the folding set, returning true if one + /// was removed or false if the node was not in the folding set. + bool RemoveNode(Node *N); + + /// GetOrInsertNode - If there is an existing simple Node exactly + /// equal to the specified node, return it. Otherwise, insert 'N' and return + /// it instead. + Node *GetOrInsertNode(Node *N); + + /// FindNodeOrInsertPos - Look up the node specified by ID. If it exists, + /// return it. If not, return the insertion token that will make insertion + /// faster. + Node *FindNodeOrInsertPos(const NodeID &ID, void *&InsertPos); + + /// InsertNode - Insert the specified node into the folding set, knowing that + /// it is not already in the folding set. InsertPos must be obtained from + /// FindNodeOrInsertPos. + void InsertNode(Node *N, void *InsertPos); + + private: + /// GetNextPtr - In order to save space, each bucket is a + /// singly-linked-list. In order to make deletion more efficient, we make + /// the list circular, so we can delete a node without computing its hash. + /// The problem with this is that the start of the hash buckets are not + /// Nodes. If NextInBucketPtr is a bucket pointer, this method returns null + /// : use GetBucketPtr when this happens. + Node *GetNextPtr(void *NextInBucketPtr); + + /// GetNextPtr - This is just like the previous GetNextPtr implementation, + /// but allows a bucket array to be specified. + Node *GetNextPtr(void *NextInBucketPtr, void **Buckets, unsigned NumBuck); + + /// GetBucketPtr - Provides a casting of a bucket pointer for isNode + /// testing. + void **GetBucketPtr(void *NextInBucketPtr); + + /// GetBucketFor - Hash the specified node ID and return the hash bucket for + /// the specified ID. + void **GetBucketFor(const NodeID &ID) const; + + /// GrowHashTable - Double the size of the hash table and rehash everything. + /// + void GrowHashTable(); + + protected: + + /// GetNodeProfile - Instantiations of the FoldingSet template implement + /// this function to gather data bits for teh given node. + virtual void GetNodeProfile(NodeID &ID, Node *N) = 0; + }; + + // Convenence types to hide the implementation of the folding set. + typedef FoldingSetImpl::Node FoldingSetNode; + typedef FoldingSetImpl::NodeID FoldingSetNodeID; + + //===--------------------------------------------------------------------===// + /// FoldingSet - This template class is used to instantiate a specialized + /// implementation of the folding set to the node class T. T must be a + /// subclass of FoldingSetNode and implement a Profile function. + /// + template class FoldingSet : public FoldingSetImpl { + private: + /// GetNodeProfile - Each instantiatation of the FoldingSet + virtual void GetNodeProfile(NodeID &ID, Node *N) { + T *TN = static_cast(N); + TN->Profile(ID); + } + + public: + /// RemoveNode - Remove a node from the folding set, returning true if one + /// was removed or false if the node was not in the folding set. + bool RemoveNode(T *N) { + return FoldingSetImpl::RemoveNode(static_cast(N)); + } + + /// GetOrInsertNode - If there is an existing simple Node exactly + /// equal to the specified node, return it. Otherwise, insert 'N' and + /// return it instead. + T *GetOrInsertNode(Node *N) { + return static_cast(FoldingSetImpl::GetOrInsertNode( + static_cast(N))); + } + + /// FindNodeOrInsertPos - Look up the node specified by ID. If it exists, + /// return it. If not, return the insertion token that will make insertion + /// faster. + T *FindNodeOrInsertPos(const FoldingSetNodeID &ID, void *&InsertPos) { + return static_cast(FoldingSetImpl::FindNodeOrInsertPos(ID, + InsertPos)); + } + + /// InsertNode - Insert the specified node into the folding set, knowing + /// that it is not already in the folding set. InsertPos must be obtained + /// from FindNodeOrInsertPos. + void InsertNode(T *N, void *InsertPos) { + FoldingSetImpl::InsertNode(static_cast(N), InsertPos); + } + }; + +}; // End of namespace llvm. + + +#endif + diff --git a/lib/Support/FoldingSet.cpp b/lib/Support/FoldingSet.cpp new file mode 100644 index 00000000000..6c1b13f3b29 --- /dev/null +++ b/lib/Support/FoldingSet.cpp @@ -0,0 +1,282 @@ +//===-- Support/FoldingSet.cpp - Uniquing Hash Set --------------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file was developed by James M. Laskey and is distributed under +// the University of Illinois Open Source License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file implements a hash set that can be used to remove duplication of +// nodes in a graph. This code was originally created by Chris Lattner for use +// with SelectionDAGCSEMap, but was isolated to provide use across the llvm code +// set. +// +//===----------------------------------------------------------------------===// + +#include "llvm/ADT/FoldingSet.h" + +#include "llvm/ADT/MathExtras.h" + +using namespace llvm; + +//===----------------------------------------------------------------------===// +// FoldingSetImpl::NodeID Implementation + +/// Add* - Add various data types to Bit data. +/// +void FoldingSetImpl::NodeID::AddPointer(const void *Ptr) { + // Note: this adds pointers to the hash using sizes and endianness that + // depend on the host. It doesn't matter however, because hashing on + // pointer values in inherently unstable. Nothing should depend on the + // ordering of nodes in the folding set. + intptr_t PtrI = (intptr_t)Ptr; + Bits.push_back(unsigned(PtrI)); + if (sizeof(intptr_t) > sizeof(unsigned)) + Bits.push_back(unsigned(uint64_t(PtrI) >> 32)); +} +void FoldingSetImpl::NodeID::AddInteger(signed I) { + Bits.push_back(I); +} +void FoldingSetImpl::NodeID::AddInteger(unsigned I) { + Bits.push_back(I); +} +void FoldingSetImpl::NodeID::AddInteger(uint64_t I) { + Bits.push_back(unsigned(I)); + Bits.push_back(unsigned(I >> 32)); +} +void FoldingSetImpl::NodeID::AddFloat(float F) { + Bits.push_back(FloatToBits(F)); +} +void FoldingSetImpl::NodeID::AddDouble(double D) { + Bits.push_back(DoubleToBits(D)); +} +void FoldingSetImpl::NodeID::AddString(const std::string &String) { + // Note: An assumption is made here that strings are composed of one byte + // chars. + unsigned Size = String.size(); + unsigned Units = Size / sizeof(unsigned); + const unsigned *Base = (const unsigned *)String.data(); + Bits.insert(Bits.end(), Base, Base + Units); + if (Size & 3) { + unsigned V = 0; + for (unsigned i = Units * sizeof(unsigned); i < Size; ++i) + V = (V << 8) | String[i]; + Bits.push_back(V); + } +} + +/// ComputeHash - Compute a strong hash value for this NodeID, used to +/// lookup the node in the FoldingSetImpl. +unsigned FoldingSetImpl::NodeID::ComputeHash() const { + // This is adapted from SuperFastHash by Paul Hsieh. + unsigned Hash = Bits.size(); + for (const unsigned *BP = &Bits[0], *E = BP+Bits.size(); BP != E; ++BP) { + unsigned Data = *BP; + Hash += Data & 0xFFFF; + unsigned Tmp = ((Data >> 16) << 11) ^ Hash; + Hash = (Hash << 16) ^ Tmp; + Hash += Hash >> 11; + } + + // Force "avalanching" of final 127 bits. + Hash ^= Hash << 3; + Hash += Hash >> 5; + Hash ^= Hash << 4; + Hash += Hash >> 17; + Hash ^= Hash << 25; + Hash += Hash >> 6; + return Hash; +} + +/// operator== - Used to compare two nodes to each other. +/// +bool FoldingSetImpl::NodeID::operator==(const FoldingSetImpl::NodeID &RHS)const{ + if (Bits.size() != RHS.Bits.size()) return false; + return memcmp(&Bits[0], &RHS.Bits[0], Bits.size()*sizeof(Bits[0])) == 0; +} + + +//===----------------------------------------------------------------------===// +// FoldingSetImpl Implementation + +FoldingSetImpl::FoldingSetImpl() : NumNodes(0) { + NumBuckets = 64; + Buckets = new void*[NumBuckets]; + memset(Buckets, 0, NumBuckets*sizeof(void*)); +} +FoldingSetImpl::~FoldingSetImpl() { + delete [] Buckets; +} + +/// GetNextPtr - In order to save space, each bucket is a +/// singly-linked-list. In order to make deletion more efficient, we make +/// the list circular, so we can delete a node without computing its hash. +/// The problem with this is that the start of the hash buckets are not +/// Nodes. If NextInBucketPtr is a bucket pointer, this method returns null +/// : use GetBucketPtr when this happens. +FoldingSetImpl::Node *FoldingSetImpl::GetNextPtr(void *NextInBucketPtr) { + if (NextInBucketPtr >= Buckets && NextInBucketPtr < Buckets+NumBuckets) + return 0; + return static_cast(NextInBucketPtr); +} + +/// GetNextPtr - This is just like the previous GetNextPtr implementation, +/// but allows a bucket array to be specified. +FoldingSetImpl::Node *FoldingSetImpl::GetNextPtr(void *NextInBucketPtr, + void **Bucks, + unsigned NumBuck) { + if (NextInBucketPtr >= Bucks && NextInBucketPtr < Bucks+NumBuck) + return 0; + return static_cast(NextInBucketPtr); +} + +/// GetBucketPtr - Provides a casting of a bucket pointer for isNode +/// testing. +void **FoldingSetImpl::GetBucketPtr(void *NextInBucketPtr) { + return static_cast(NextInBucketPtr); +} + +/// GetBucketFor - Hash the specified node ID and return the hash bucket for +/// the specified ID. +void **FoldingSetImpl::GetBucketFor(const NodeID &ID) const { + // NumBuckets is always a power of 2. + unsigned BucketNum = ID.ComputeHash() & (NumBuckets-1); + return Buckets+BucketNum; +} + +/// GrowHashTable - Double the size of the hash table and rehash everything. +/// +void FoldingSetImpl::GrowHashTable() { + void **OldBuckets = Buckets; + unsigned OldNumBuckets = NumBuckets; + NumBuckets <<= 1; + + // Reset the node count to zero: we're going to reinsert everything. + NumNodes = 0; + + // Clear out new buckets. + Buckets = new void*[NumBuckets]; + memset(Buckets, 0, NumBuckets*sizeof(void*)); + + // Walk the old buckets, rehashing nodes into their new place. + for (unsigned i = 0; i != OldNumBuckets; ++i) { + void *Probe = OldBuckets[i]; + if (!Probe) continue; + while (Node *NodeInBucket = GetNextPtr(Probe, OldBuckets, OldNumBuckets)){ + // Figure out the next link, remove NodeInBucket from the old link. + Probe = NodeInBucket->getNextInBucket(); + NodeInBucket->SetNextInBucket(0); + + // Insert the node into the new bucket, after recomputing the hash. + NodeID ID; + GetNodeProfile(ID, NodeInBucket); + InsertNode(NodeInBucket, GetBucketFor(ID)); + } + } + + delete[] OldBuckets; +} + +/// FindNodeOrInsertPos - Look up the node specified by ID. If it exists, +/// return it. If not, return the insertion token that will make insertion +/// faster. +FoldingSetImpl::Node *FoldingSetImpl::FindNodeOrInsertPos(const NodeID &ID, + void *&InsertPos) { + void **Bucket = GetBucketFor(ID); + void *Probe = *Bucket; + + InsertPos = 0; + + while (Node *NodeInBucket = GetNextPtr(Probe)) { + NodeID OtherID; + GetNodeProfile(OtherID, NodeInBucket); + if (OtherID == ID) + return NodeInBucket; + + Probe = NodeInBucket->getNextInBucket(); + } + + // Didn't find the node, return null with the bucket as the InsertPos. + InsertPos = Bucket; + return 0; +} + +/// InsertNode - Insert the specified node into the folding set, knowing that it +/// is not already in the map. InsertPos must be obtained from +/// FindNodeOrInsertPos. +void FoldingSetImpl::InsertNode(Node *N, void *InsertPos) { + ++NumNodes; + // Do we need to grow the hashtable? + if (NumNodes > NumBuckets*2) { + GrowHashTable(); + NodeID ID; + GetNodeProfile(ID, N); + InsertPos = GetBucketFor(ID); + } + + /// The insert position is actually a bucket pointer. + void **Bucket = static_cast(InsertPos); + + void *Next = *Bucket; + + // If this is the first insertion into this bucket, its next pointer will be + // null. Pretend as if it pointed to itself. + if (Next == 0) + Next = Bucket; + + // Set the nodes next pointer, and make the bucket point to the node. + N->SetNextInBucket(Next); + *Bucket = N; +} + +/// RemoveNode - Remove a node from the folding set, returning true if one was +/// removed or false if the node was not in the folding set. +bool FoldingSetImpl::RemoveNode(Node *N) { + // Because each bucket is a circular list, we don't need to compute N's hash + // to remove it. Chase around the list until we find the node (or bucket) + // which points to N. + void *Ptr = N->getNextInBucket(); + if (Ptr == 0) return false; // Not in folding set. + + --NumNodes; + + void *NodeNextPtr = Ptr; + N->SetNextInBucket(0); + while (true) { + if (Node *NodeInBucket = GetNextPtr(Ptr)) { + // Advance pointer. + Ptr = NodeInBucket->getNextInBucket(); + + // We found a node that points to N, change it to point to N's next node, + // removing N from the list. + if (Ptr == N) { + NodeInBucket->SetNextInBucket(NodeNextPtr); + return true; + } + } else { + void **Bucket = GetBucketPtr(Ptr); + Ptr = *Bucket; + + // If we found that the bucket points to N, update the bucket to point to + // whatever is next. + if (Ptr == N) { + *Bucket = NodeNextPtr; + return true; + } + } + } +} + +/// GetOrInsertNode - If there is an existing simple Node exactly +/// equal to the specified node, return it. Otherwise, insert 'N' and it +/// instead. +FoldingSetImpl::Node *FoldingSetImpl::GetOrInsertNode(FoldingSetImpl::Node *N) { + NodeID ID; + GetNodeProfile(ID, N); + void *IP; + if (Node *E = FindNodeOrInsertPos(ID, IP)) + return E; + InsertNode(N, IP); + return N; +}