Renamed table to index and reorganized benchmarks.

cxxmph/Makefile.am

@@ -1,28 +1,28 @@
 AM_CXXFLAGS='-std=c++0x'
 TESTS = $(check_PROGRAMS)
-check_PROGRAMS = mph_map_test mph_table_test trigraph_test
+check_PROGRAMS = mph_map_test mph_index_test trigraph_test
-noinst_PROGRAMS = bm_numbers bm_urls
+noinst_PROGRAMS = bm_index bm_map
 bin_PROGRAMS = cxxmph
 lib_LTLIBRARIES = libcxxmph.la
-libcxxmph_la_SOURCES = MurmurHash2.h trigragh.h trigraph.cc mph_table.h mph_table.cc seeded_hash.h stringpiece.h benchmark.h benchmark.cc
+libcxxmph_la_SOURCES = MurmurHash2.h trigragh.h trigraph.cc mph_index.h mph_index.cc seeded_hash.h stringpiece.h benchmark.h benchmark.cc
 libcxxmph_la_LDFLAGS = -version-info 0:0:0
 cxxmph_includedir = $(includedir)/cxxmph/
-cxxmph_include_HEADERS = mph_map.h mph_table.h MurmurHash2.h trigraph.h seeded_hash.h stringpiece.h
+cxxmph_include_HEADERS = mph_map.h mph_index.h MurmurHash2.h trigraph.h seeded_hash.h stringpiece.h
 
 mph_map_test_LDADD   = libcxxmph.la
 mph_map_test_SOURCES = mph_map_test.cc
 
-mph_table_test_LDADD   = libcxxmph.la
+mph_index_test_LDADD   = libcxxmph.la
-mph_table_test_SOURCES = mph_table_test.cc
+mph_index_test_SOURCES = mph_index_test.cc
 
 trigraph_test_LDADD   = libcxxmph.la
 trigraph_test_SOURCES = trigraph_test.cc
 
-bm_numbers_LDADD = libcxxmph.la
+bm_index_LDADD = libcxxmph.la
-bm_numbers_SOURCES = bm_numbers.cc
+bm_index_SOURCES = bm_common.cc bm_index.cc 
 
-bm_urls_LDADD = libcxxmph.la
+bm_map_LDADD = libcxxmph.la
-bm_urls_SOURCES = bm_urls.cc
+bm_map_SOURCES = bm_common.cc bm_map.cc
 
 cxxmph_LDADD   = libcxxmph.la
 cxxmph_SOURCES = cxxmph.cc

cxxmph/benchmark.cc

@@ -1,7 +1,9 @@
 #include "benchmark.h"
 
+#include <cerrno>
 #include <cstring>
 #include <cstdio>
+#include <sys/time.h>
 #include <sys/resource.h>
 
 #include <iostream>
@@ -50,6 +52,16 @@ struct rusage getrusage_or_die() {
   return rs;
 }
 
+struct timeval gettimeofday_or_die() {
+  struct timeval tv;
+  int ret = gettimeofday(&tv, NULL); 
+  if (ret != 0) {
+    cerr << "gettimeofday failed: " << strerror(errno) << endl;
+    exit(-1);
+  }
+  return tv;
+}
+
 #ifdef HAVE_CXA_DEMANGLE
 string demangle(const string& name) {
   char buf[1024];
@@ -79,25 +91,33 @@ namespace cxxmph {
 }
 
 /* static */ void Benchmark::RunAll() {
-  for (auto it = g_benchmarks.begin(); it != g_benchmarks.end(); ++it) {
+  for (int i = 0; i < g_benchmarks.size(); ++i) {
-    (*it)->MeasureRun();
+    Benchmark* bm = g_benchmarks[i];
-    delete *it;
+    bm->SetUp();
+    bm->MeasureRun();
+    bm->TearDown(); 
+    delete bm;
   }
 }
 
 void Benchmark::MeasureRun() {
+  struct timeval walltime_begin = gettimeofday_or_die();
   struct rusage begin = getrusage_or_die();
-  Run(iters_);
+  Run();
   struct rusage end = getrusage_or_die();
+  struct timeval walltime_end = gettimeofday_or_die();
 
   struct timeval utime;
   timeval_subtract(&utime, &end.ru_utime, &begin.ru_utime);
   struct timeval stime;
   timeval_subtract(&stime, &end.ru_stime, &begin.ru_stime);
+  struct timeval wtime;
+  timeval_subtract(&wtime, &walltime_end, &walltime_begin);
 
   printf("Benchmark: %s\n", name().c_str());
-  printf("User time used  : %ld.%06ld\n", utime.tv_sec, utime.tv_usec);
+  printf("CPU User time  : %ld.%06ld\n", utime.tv_sec, utime.tv_usec);
-  printf("System time used: %ld.%06ld\n", stime.tv_sec, stime.tv_usec);
+  printf("CPU System time: %ld.%06ld\n", stime.tv_sec, stime.tv_usec);
+  printf("Wall clock time: %ld.%06ld\n", wtime.tv_sec, wtime.tv_usec);
   printf("\n");
 }
 

cxxmph/benchmark.h

@@ -8,9 +8,9 @@ namespace cxxmph {
 
 class Benchmark {
  public:
-  Benchmark(int iters = 1) : iters_(iters) { }
+  Benchmark() {}
-  virtual void Run(int iters) = 0;
+  virtual ~Benchmark() {}
-  virtual ~Benchmark() { }
+
   const std::string& name() { return name_; }
   void set_name(const std::string& name) { name_ = name; }
 
@@ -18,10 +18,11 @@ class Benchmark {
   static void RunAll();
 
  protected:
-  int iters() { return iters_; }
+  virtual bool SetUp() {}; 
+  virtual void Run() = 0;
+  virtual bool TearDown() {};
 
  private:
-  int iters_;
   std::string name_;
   void MeasureRun();
 };

cxxmph/bm_map.cc

@@ -0,0 +1,52 @@
+#include <string>
+#include <unordered_map>
+
+#include "bm_common.h"
+#include "mph_map.h"
+
+using cxxmph::mph_map;
+using std::string;
+using std::unordered_map;
+
+namespace cxxmph {
+
+template <class MapType>
+class BM_MapCreate : public UrlsBenchmark {
+ public:
+  virtual void Run() {
+    MapType mymap;
+    for (auto it = urls_.begin(); it != urls_.end(); ++it) {
+      mymap[*it] = *it;
+    }
+  }
+};
+
+template <class MapType>
+class BM_MapSearch : public SearchUrlsBenchmark {
+ public:
+  virtual void Run() {
+    for (auto it = random_.begin(); it != random_.end(); ++it) {
+      auto value = mymap[*it];
+    }
+  }
+ protected:
+  virtual void SetUp() {
+    for (auto it = urls_.begin(); it != urls_.end(); ++it) {
+      mymap_[*it] = *it;
+    }
+    mymap_.resize(mymap.size());
+  }
+  MapType mymap_;
+};
+
+}  // namespace cxxmph
+
+using namespace cxxmph;
+
+int main(int argc, char** argv) {
+  Benchmark::Register(new BM_MapCreate<mph_map<string, string>>("URLS100k"));
+  Benchmark::Register(new BM_MapCreate<unordered_map<string, string>>("URLS100k"));
+  Benchmark::Register(new BM_MapSearch<mph_map<string, string>>("URLS100k", 1000 * 1000));
+  Benchmark::Register(new BM_MapSearch<unordered_map<string, string>>("URLS100k", 1000 * 1000));
+  Benchmark::RunAll();
+}

cxxmph/bm_numbers.cc

@@ -1,52 +0,0 @@
-#include <set>
-#include <vector>
-
-#include "benchmark.h"
-#include "mph_table.h"
-
-using std::set;
-using std::vector;
-
-namespace cxxmph {
-class BM_NumbersCreate : public Benchmark {
- public:
-  BM_NumbersCreate(int iters = 1) : Benchmark(iters) {
-    set<int> unique;
-    while (unique.size() < 1000 * 1000) {
-      int v = random();
-      if (unique.find(v) == unique.end()) {
-        unique.insert(v);
-        random_unique_.push_back(v);
-      }
-    }
-  }
- protected:
-  virtual void Run(int iters) {
-    SimpleMPHTable<int> table;
-    table.Reset(random_unique_.begin(), random_unique_.end());
-  }
-  std::vector<int> random_unique_;
-};
-
-class BM_NumbersFind : public BM_NumbersCreate {
- public:
-  BM_NumbersFind(int iters) : BM_NumbersCreate(iters) { table_.Reset(random_unique_.begin(), random_unique_.end()); } 
-  virtual void Run(int iters) {
-    for (int i = 0; i < iters * 100; ++i) {
-      int pos = random() % random_unique_.size();;
-      int h = table_.index(pos);
-    }
-  }
- private:
-   SimpleMPHTable<int> table_;
-};
-
-}  // namespace cxxmph
-
-using namespace cxxmph;
-
-int main(int argc, char** argv) {
-  Benchmark::Register(new BM_NumbersCreate());
-  Benchmark::Register(new BM_NumbersFind(1000 * 1000));
-  Benchmark::RunAll();
-}

cxxmph/bm_urls.cc

@@ -1,70 +0,0 @@
-#include <fstream>
-#include <iostream>
-#include <set>
-#include <string>
-#include <vector>
-#include <unordered_map>
-
-#include "benchmark.h"
-#include "mph_map.h"
-
-using std::ifstream;
-using std::set;
-using std::string;
-using std::vector;
-
-namespace cxxmph {
-
-class BM_UrlsCreate : public Benchmark {
- public:
-  BM_UrlsCreate(int iters = 1) : Benchmark(iters) {
-    ReadUrls();
-  }
- protected:
-  virtual void Run(int iters) {
-    BuildTable();
-  }
-  void BuildTable() {
-    for (auto it = urls_.begin(); it != urls_.end(); ++it) {
-      table_[*it] = it - urls_.begin();
-    }
-    table_.pack();
-  }
-  void ReadUrls() {
-    vector<string> urls;
-    std::ifstream f("URLS100k");
-    string buffer;
-    while(std::getline(f, buffer)) urls.push_back(buffer);
-    set<string> unique(urls.begin(), urls.end());
-    if (unique.size() != urls.size()) {
-      cerr << "Input file has repeated keys." << endl;
-      exit(-1);
-    }
-    urls_.swap(urls);
-  }
-  vector<string> urls_;
-  cxxmph::mph_map<string, int> table_;
-};
-   
-class BM_UrlsFind : public BM_UrlsCreate {
- public:
-  BM_UrlsFind(int iters = 1) : BM_UrlsCreate(iters) { ReadUrls(); BuildTable(); }
- protected:
-  virtual void Run(int iters) {
-    for (int i = 0; i < iters * 100; ++i) {
-      int pos = random() % urls_.size();;
-      int h = table_[urls_[pos]];
-      assert(h == pos);
-    }
-  }
-};
-
-}  // namespace cxxmph
-
-using namespace cxxmph;
-
-int main(int argc, char** argv) {
-  Benchmark::Register(new BM_UrlsCreate());
-  Benchmark::Register(new BM_UrlsFind(1000 * 1000));
-  Benchmark::RunAll();
-}

cxxmph/mph_index.cc

@@ -5,7 +5,7 @@
 using std::cerr;
 using std::endl;
 
-#include "mph_table.h"
+#include "mph_index.h"
 
 using std::vector;
 
@@ -13,7 +13,7 @@ namespace {
 
 static const uint8_t kUnassigned = 3;
 // table used for looking up the number of assigned vertices to a 8-bit integer
-static uint8_t kBdzLookupTable[] =
+static uint8_t kBdzLookupIndex[] =
 {
 4, 4, 4, 3, 4, 4, 4, 3, 4, 4, 4, 3, 3, 3, 3, 2,
 4, 4, 4, 3, 4, 4, 4, 3, 4, 4, 4, 3, 3, 3, 3, 2,
@@ -37,13 +37,13 @@ static uint8_t kBdzLookupTable[] =
 
 namespace cxxmph {
 
-const uint8_t MPHTable::valuemask[] = { 0xfc, 0xf3, 0xcf, 0x3f};
+const uint8_t MPHIndex::valuemask[] = { 0xfc, 0xf3, 0xcf, 0x3f};
 
-MPHTable::~MPHTable() {
+MPHIndex::~MPHIndex() {
   clear();
 }
 
-void MPHTable::clear() {
+void MPHIndex::clear() {
   if (!deserialized_) delete [] g_;
   g_ = NULL;
   g_size_ = 0;
@@ -53,7 +53,7 @@ void MPHTable::clear() {
   // TODO(davi) implement me
 }
 
-bool MPHTable::GenerateQueue(
+bool MPHIndex::GenerateQueue(
     TriGraph* graph, vector<uint32_t>* queue_output) {
   uint32_t queue_head = 0, queue_tail = 0;
   uint32_t nedges = m_;
@@ -109,7 +109,7 @@ bool MPHTable::GenerateQueue(
   return cycles == 0;
 }
 
-void MPHTable::Assigning(
+void MPHIndex::Assigning(
     const vector<TriGraph::Edge>& edges, const vector<uint32_t>& queue) {
   uint32_t current_edge = 0;
   vector<bool> marked_vertices(n_ + 1);
@@ -164,7 +164,7 @@ void MPHTable::Assigning(
   g_ = g;
 }
 
-void MPHTable::Ranking() {
+void MPHIndex::Ranking() {
   uint32_t nbytes_total = static_cast<uint32_t>(ceil(n_ / 4.0));
   uint32_t size = k_ >> 2U;
   ranktable_size_ = static_cast<uint32_t>(
@@ -179,7 +179,7 @@ void MPHTable::Ranking() {
   while (1) {
     if (i == ranktable_size_) break;
     uint32_t nbytes = size < nbytes_total ? size : nbytes_total;
-    for (uint32_t j = 0; j < nbytes; ++j) count += kBdzLookupTable[g_[offset + j]];
+    for (uint32_t j = 0; j < nbytes; ++j) count += kBdzLookupIndex[g_[offset + j]];
     ranktable[i] = count;
     offset += nbytes;
     nbytes_total -= size;
@@ -188,13 +188,13 @@ void MPHTable::Ranking() {
   ranktable_ = ranktable;
 }
 
-uint32_t MPHTable::Rank(uint32_t vertex) const {
+uint32_t MPHIndex::Rank(uint32_t vertex) const {
   uint32_t index = vertex >> b_;
   uint32_t base_rank = ranktable_[index];
   uint32_t beg_idx_v = index << b_;
   uint32_t beg_idx_b = beg_idx_v >> 2;
   uint32_t end_idx_b = vertex >> 2;
-  while (beg_idx_b < end_idx_b) base_rank += kBdzLookupTable[g_[beg_idx_b++]];
+  while (beg_idx_b < end_idx_b) base_rank += kBdzLookupIndex[g_[beg_idx_b++]];
   beg_idx_v = beg_idx_b << 2;
   // cerr << "beg_idx_v: " << beg_idx_v << endl;
   // cerr << "base rank: " << base_rank << endl;
@@ -213,21 +213,21 @@ uint32_t MPHTable::Rank(uint32_t vertex) const {
   return base_rank;
 }
 
-uint32_t MPHTable::serialize_bytes_needed() const {
+uint32_t MPHIndex::serialize_bytes_needed() const {
-  return sizeof(MPHTable) + g_size_ + ranktable_size_*sizeof(uint32_t);
+  return sizeof(MPHIndex) + g_size_ + ranktable_size_*sizeof(uint32_t);
 }
-void MPHTable::serialize(char* memory) const {
+void MPHIndex::serialize(char* memory) const {
-  memcpy(memory, this, sizeof(MPHTable));
+  memcpy(memory, this, sizeof(MPHIndex));
-  memcpy(memory + sizeof(MPHTable), g_, g_size_);
+  memcpy(memory + sizeof(MPHIndex), g_, g_size_);
-  memcpy(memory + sizeof(MPHTable) + g_size_,
+  memcpy(memory + sizeof(MPHIndex) + g_size_,
       ranktable_, ranktable_size_*sizeof(uint32_t));
 }
 
-bool MPHTable::deserialize(const char* serialized_memory) {
+bool MPHIndex::deserialize(const char* serialized_memory) {
-  memcpy(this, serialized_memory, sizeof(MPHTable)); 
+  memcpy(this, serialized_memory, sizeof(MPHIndex)); 
-  g_ = reinterpret_cast<const uint8_t*>(serialized_memory + sizeof(MPHTable));
+  g_ = reinterpret_cast<const uint8_t*>(serialized_memory + sizeof(MPHIndex));
   ranktable_ = reinterpret_cast<const uint32_t*>(
-      serialized_memory + sizeof(MPHTable) + g_size_);
+      serialized_memory + sizeof(MPHIndex) + g_size_);
   deserialized_ = true;
   return true;
 }

cxxmph/mph_index.h

@@ -0,0 +1,173 @@
+#ifndef __CXXMPH_MPH_INDEX_H__
+#define __CXXMPH_MPH_INDEX_H__
+
+// Minimal perfect hash abstraction implementing the BDZ algorithm
+
+#include <stdint.h>
+
+#include <cassert>
+#include <cmath>
+#include <unordered_map>  // for std::hash
+#include <vector>
+
+#include <iostream>
+
+using std::cerr;
+using std::endl;
+
+#include "seeded_hash.h"
+#include "trigraph.h"
+
+namespace cxxmph {
+
+class MPHIndex {
+ public:
+  MPHIndex(double c = 1.23, uint8_t b = 7) :
+      c_(c), b_(b), m_(0), n_(0), k_(0), r_(0),
+      g_(NULL), g_size_(0), ranktable_(NULL), ranktable_size_(0),
+      deserialized_(false) { }
+  ~MPHIndex();
+
+  template <class SeededHashFcn, class ForwardIterator>
+  bool Reset(ForwardIterator begin, ForwardIterator end);
+  template <class SeededHashFcn, class Key>  // must agree with Reset
+  uint32_t index(const Key& x) const;
+  uint32_t size() const { return m_; }
+  void clear();
+
+  // Serialization machinery for mmap usage. 
+  // Serialized tables are not guaranteed to work across versions or different
+  // endianness (although they could easily be made to be).
+  uint32_t serialize_bytes_needed() const;
+  void serialize(char *memory) const;
+  bool deserialize(const char* serialized_memory);
+
+ private:
+  template <class SeededHashFcn, class ForwardIterator>
+  bool Mapping(ForwardIterator begin, ForwardIterator end,
+               std::vector<TriGraph::Edge>* edges,
+               std::vector<uint32_t>* queue);
+  bool GenerateQueue(TriGraph* graph, std::vector<uint32_t>* queue);
+  void Assigning(const std::vector<TriGraph::Edge>& edges,
+                 const std::vector<uint32_t>& queue);
+  void Ranking();
+  uint32_t Rank(uint32_t vertex) const;
+
+  // Algorithm parameters
+  double c_;  // Number of bits per key (? is it right)
+  uint8_t b_;  // Number of bits of the kth index in the ranktable
+
+  // Values used during generation
+  uint32_t m_;  // edges count
+  uint32_t n_;  // vertex count
+  uint32_t k_;  // kth index in ranktable, $k = log_2(n=3r)\varepsilon$
+
+  // Values used during search
+
+  // Partition vertex count, derived from c parameter.
+  uint32_t r_;
+  // The array containing the minimal perfect hash function graph. Do not use
+  // c++ vector to make mmap based backing easier.
+  const uint8_t* g_;
+  uint32_t g_size_;
+  // The table used for the rank step of the minimal perfect hash function
+  const uint32_t* ranktable_;
+  uint32_t ranktable_size_;
+  // The selected hash seed triplet for finding the edges in the minimal
+  // perfect hash function graph.
+  uint32_t hash_seed_[3];
+
+  bool deserialized_;
+
+  static const uint8_t valuemask[];
+  static void set_2bit_value(uint8_t *d, uint32_t i, uint8_t v) {
+    d[(i >> 2)] &= ((v << ((i & 3) << 1)) | valuemask[i & 3]);
+  }
+  static uint32_t get_2bit_value(const uint8_t* d, uint32_t i) {
+    return (d[(i >> 2)] >> (((i & 3) << 1)) & 3);
+  }
+
+  
+};
+
+// Template method needs to go in the header file.
+template <class SeededHashFcn, class ForwardIterator>
+bool MPHIndex::Reset(ForwardIterator begin, ForwardIterator end) {
+  m_ = end - begin;
+  r_ = static_cast<uint32_t>(ceil((c_*m_)/3));
+  if ((r_ % 2) == 0) r_ += 1;
+  n_ = 3*r_;
+  k_ = 1U << b_;
+
+  // cerr << "m " << m_ << " n " << n_ << " r " << r_ << endl;
+
+  int iterations = 10;
+  std::vector<TriGraph::Edge> edges;
+  std::vector<uint32_t> queue;
+  while (1) {
+    // cerr << "Iterations missing: " << iterations << endl;
+    for (int i = 0; i < 3; ++i) hash_seed_[i] = random() % m_;
+    // for (int i = 0; i < 3; ++i) hash_seed_[i] = random() + i;
+    if (Mapping<SeededHashFcn>(begin, end, &edges, &queue)) break;
+    else --iterations;
+    if (iterations == 0) break;
+  }
+  if (iterations == 0) return false;
+  Assigning(edges, queue);
+  std::vector<TriGraph::Edge>().swap(edges);
+  Ranking();
+  deserialized_ = false;
+  return true;
+}
+
+template <class SeededHashFcn, class ForwardIterator>
+bool MPHIndex::Mapping(
+    ForwardIterator begin, ForwardIterator end,
+    std::vector<TriGraph::Edge>* edges, std::vector<uint32_t>* queue) {
+  TriGraph graph(n_, m_);
+  for (ForwardIterator it = begin; it != end; ++it) { 
+    uint32_t h[3];
+    for (int i = 0; i < 3; ++i) h[i] = SeededHashFcn()(*it, hash_seed_[i]);
+    uint32_t v0 = h[0] % r_;
+    uint32_t v1 = h[1] % r_ + r_;
+    uint32_t v2 = h[2] % r_ + (r_ << 1);
+    // cerr << "Key: " << *it << " edge " <<  it - begin << " (" << v0 << "," << v1 << "," << v2 << ")" << endl;
+    graph.AddEdge(TriGraph::Edge(v0, v1, v2));
+  }
+  if (GenerateQueue(&graph, queue)) {
+     graph.ExtractEdgesAndClear(edges);
+     return true;
+  }
+  return false;
+}
+
+template <class SeededHashFcn, class Key>
+uint32_t MPHIndex::index(const Key& key) const {
+  uint32_t h[3];
+  for (int i = 0; i < 3; ++i) h[i] = SeededHashFcn()(key, hash_seed_[i]);
+  h[0] = h[0] % r_;
+  h[1] = h[1] % r_ + r_;
+  h[2] = h[2] % r_ + (r_ << 1);
+  assert(g_size_);
+  // cerr << "g_.size() " << g_size_ << " h0 >> 2 " << (h[0] >> 2) << endl;
+  assert((h[0] >> 2) <g_size_);
+  assert((h[1] >> 2) <g_size_);
+  assert((h[2] >> 2) <g_size_);
+  uint32_t vertex = h[(get_2bit_value(g_, h[0]) + get_2bit_value(g_, h[1]) + get_2bit_value(g_, h[2])) % 3];
+  // cerr << "Search found vertex " << vertex << endl;
+  return Rank(vertex);
+}
+
+template <class Key, class HashFcn = typename seeded_hash<std::hash<Key> >::hash_function>
+class SimpleMPHIndex : public MPHIndex {
+ public:
+  template <class ForwardIterator>
+  bool Reset(ForwardIterator begin, ForwardIterator end) {
+    return MPHIndex::Reset<HashFcn>(begin, end);
+  }
+  uint32_t index(const Key& key) { return MPHIndex::index<HashFcn>(key); }
+};
+
+}  // namespace cxxmph
+
+#endif // __CXXMPH_MPH_INDEX_H__

cxxmph/mph_index_test.cc

@@ -3,11 +3,11 @@
 #include <string>
 #include <vector>
 
-#include "mph_table.h"
+#include "mph_index.h"
 
 using std::string;
 using std::vector;
-using cxxmph::SimpleMPHTable;
+using cxxmph::SimpleMPHIndex;
 
 int main(int argc, char** argv) {
 
@@ -23,20 +23,20 @@ int main(int argc, char** argv) {
   keys.push_back("diogo");
   keys.push_back("algume");
 
-  SimpleMPHTable<string> mph_table;
+  SimpleMPHIndex<string> mph_index;
-  assert(mph_table.Reset(keys.begin(), keys.end()));
+  assert(mph_index.Reset(keys.begin(), keys.end()));
   vector<int> ids;
   for (vector<int>::size_type i = 0; i < keys.size(); ++i) {
-     ids.push_back(mph_table.index(keys[i]));
+     ids.push_back(mph_index.index(keys[i]));
      cerr << " " << *(ids.end() - 1);
   }
   cerr << endl;
   sort(ids.begin(), ids.end());
   for (vector<int>::size_type i = 0; i < ids.size(); ++i) assert(ids[i] == static_cast<vector<int>::value_type>(i));
 
-  char* serialized = new char[mph_table.serialize_bytes_needed()];
+  char* serialized = new char[mph_index.serialize_bytes_needed()];
-  mph_table.serialize(serialized);
+  mph_index.serialize(serialized);
-  SimpleMPHTable<string> other_mph_table;
+  SimpleMPHIndex<string> other_mph_index;
-  other_mph_table.deserialize(serialized);
+  other_mph_index.deserialize(serialized);
 }
   

cxxmph/mph_map.h

@@ -4,7 +4,7 @@
 #include <utility>  // for std::pair
 
 #include "MurmurHash2.h"
-#include "mph_table.h"
+#include "mph_index.h"
 
 namespace cxxmph {
 
@@ -70,7 +70,7 @@ class mph_map {
 
    void rehash();
    std::vector<value_type> values_;
-   SimpleMPHTable<Key, typename seeded_hash<HashFcn>::hash_function> table_;
+   SimpleMPHIndex<Key, typename seeded_hash<HashFcn>::hash_function> index_;
    // TODO(davi) optimize slack to no hold a copy of the key
    typedef typename std::unordered_map<Key, uint32_t, HashFcn, EqualKey, Alloc> slack_type;
    slack_type slack_;
@@ -93,8 +93,8 @@ MPH_MAP_METHOD_DECL(insert_return_type, insert)(const value_type& x) {
   if (it != end()) return std::make_pair(it, false);
   values_.push_back(x);
   slack_.insert(std::make_pair(x.first, values_.size() - 1));
-  if (slack_.size() == table_.size() ||
+  if (slack_.size() == index_.size() ||
-      (slack_.size() >= 256 && table_.size() == 0)) {
+      (slack_.size() >= 256 && index_.size() == 0)) {
      rehash();
   }
   it = find(x.first);
@@ -104,14 +104,14 @@ MPH_MAP_METHOD_DECL(insert_return_type, insert)(const value_type& x) {
 MPH_MAP_METHOD_DECL(void_type, rehash)() {
   if (values_.empty()) return;
   slack_type().swap(slack_);
-  bool success = table_.Reset(
+  bool success = index_.Reset(
       make_iterator_first(values_.begin()),
       make_iterator_first(values_.end()));
   assert(success);
   std::vector<value_type> new_values(values_.size());
   for (const_iterator it = values_.begin(), end = values_.end();
        it != end; ++it) {
-    size_type id = table_.index(it->first);
+    size_type id = index_.index(it->first);
     assert(id < new_values.size());
     new_values[id] = *it;
   }
@@ -127,7 +127,7 @@ MPH_MAP_METHOD_DECL(bool_type, empty)() const { return values_.empty(); }
 MPH_MAP_METHOD_DECL(void_type, clear)() {
   values_.clear();
   slack_.clear();
-  table_.clear();
+  index_.clear();
 }
 
 MPH_MAP_METHOD_DECL(void_type, erase)(iterator pos) {
@@ -145,8 +145,8 @@ MPH_MAP_METHOD_DECL(const_iterator, find)(const key_type& k) const {
     typename slack_type::const_iterator it = slack_.find(k);
     if (it != slack_.end()) return values_.begin() + it->second;
   }
-  if (table_.size() == 0) return end();
+  if (index_.size() == 0) return end();
-  size_type id = table_.index(k);
+  size_type id = index_.index(k);
   if (key_equal()(values_[id].first, k)) {
     return values_.begin() + id;
   }
@@ -157,8 +157,8 @@ MPH_MAP_METHOD_DECL(iterator, find)(const key_type& k) {
     typename slack_type::const_iterator it = slack_.find(k);
     if (it != slack_.end()) return values_.begin() + it->second;
   }
-  if (table_.size() == 0) return end();
+  if (index_.size() == 0) return end();
-  size_type id = table_.index(k);
+  size_type id = index_.index(k);
   if (key_equal()(values_[id].first, k)) {
     return values_.begin() + id;
   }

cxxmph/mph_table.h

@@ -1,173 +1,16 @@
-#ifndef __CXXMPH_MPH_TABLE_H__
+#include "mph_index.h"
-#define __CXXMPH_MPH_TABLE_H__
 
-// Minimal perfect hash abstraction implementing the BDZ algorithm
+// String to string map working on mmap'ed memory
-
-#include <stdint.h>
-
-#include <cassert>
-#include <cmath>
-#include <unordered_map>  // for std::hash
-#include <vector>
-
-#include <iostream>
-
-using std::cerr;
-using std::endl;
-
-#include "seeded_hash.h"
-#include "trigraph.h"
-
-namespace cxxmph {
 
 class MPHTable {
  public:
-  MPHTable(double c = 1.23, uint8_t b = 7) :
+  typedef StringPiece key_type;
-      c_(c), b_(b), m_(0), n_(0), k_(0), r_(0),
+  typedef StringPiece data_type;
-      g_(NULL), g_size_(0), ranktable_(NULL), ranktable_size_(0),
+  typedef std::pair<StringPiece, StringPiece> value_type;
-      deserialized_(false) { }
-  ~MPHTable();
-
-  template <class SeededHashFcn, class ForwardIterator>
-  bool Reset(ForwardIterator begin, ForwardIterator end);
-  template <class SeededHashFcn, class Key>  // must agree with Reset
-  uint32_t index(const Key& x) const;
-  uint32_t size() const { return m_; }
-  void clear();
-
-  // Serialization machinery for mmap usage. 
-  // Serialized tables are not guaranteed to work across versions or different
-  // endianness (although they could easily be made to be).
-  uint32_t serialize_bytes_needed() const;
-  void serialize(char *memory) const;
-  bool deserialize(const char* serialized_memory);
-
- private:
-  template <class SeededHashFcn, class ForwardIterator>
-  bool Mapping(ForwardIterator begin, ForwardIterator end,
-               std::vector<TriGraph::Edge>* edges,
-               std::vector<uint32_t>* queue);
-  bool GenerateQueue(TriGraph* graph, std::vector<uint32_t>* queue);
-  void Assigning(const std::vector<TriGraph::Edge>& edges,
-                 const std::vector<uint32_t>& queue);
-  void Ranking();
-  uint32_t Rank(uint32_t vertex) const;
-
-  // Algorithm parameters
-  double c_;  // Number of bits per key (? is it right)
-  uint8_t b_;  // Number of bits of the kth index in the ranktable
-
-  // Values used during generation
-  uint32_t m_;  // edges count
-  uint32_t n_;  // vertex count
-  uint32_t k_;  // kth index in ranktable, $k = log_2(n=3r)\varepsilon$
-
-  // Values used during search
-
-  // Partition vertex count, derived from c parameter.
-  uint32_t r_;
-  // The array containing the minimal perfect hash function graph. Do not use
-  // c++ vector to make mmap based backing easier.
-  const uint8_t* g_;
-  uint32_t g_size_;
-  // The table used for the rank step of the minimal perfect hash function
-  const uint32_t* ranktable_;
-  uint32_t ranktable_size_;
-  // The selected hash seed triplet for finding the edges in the minimal
-  // perfect hash function graph.
-  uint32_t hash_seed_[3];
-
-  bool deserialized_;
-
-  static const uint8_t valuemask[];
-  static void set_2bit_value(uint8_t *d, uint32_t i, uint8_t v) {
-    d[(i >> 2)] &= ((v << ((i & 3) << 1)) | valuemask[i & 3]);
-  }
-  static uint32_t get_2bit_value(const uint8_t* d, uint32_t i) {
-    return (d[(i >> 2)] >> (((i & 3) << 1)) & 3);
-  }
-
-  
-};
-
-// Template method needs to go in the header file.
-template <class SeededHashFcn, class ForwardIterator>
-bool MPHTable::Reset(ForwardIterator begin, ForwardIterator end) {
-  m_ = end - begin;
-  r_ = static_cast<uint32_t>(ceil((c_*m_)/3));
-  if ((r_ % 2) == 0) r_ += 1;
-  n_ = 3*r_;
-  k_ = 1U << b_;
-
-  // cerr << "m " << m_ << " n " << n_ << " r " << r_ << endl;
-
-  int iterations = 10;
-  std::vector<TriGraph::Edge> edges;
-  std::vector<uint32_t> queue;
-  while (1) {
-    // cerr << "Iterations missing: " << iterations << endl;
-    for (int i = 0; i < 3; ++i) hash_seed_[i] = random() % m_;
-    // for (int i = 0; i < 3; ++i) hash_seed_[i] = random() + i;
-    if (Mapping<SeededHashFcn>(begin, end, &edges, &queue)) break;
-    else --iterations;
-    if (iterations == 0) break;
-  }
-  if (iterations == 0) return false;
-  Assigning(edges, queue);
-  std::vector<TriGraph::Edge>().swap(edges);
-  Ranking();
-  deserialized_ = false;
-  return true;
-}
-
-template <class SeededHashFcn, class ForwardIterator>
-bool MPHTable::Mapping(
-    ForwardIterator begin, ForwardIterator end,
-    std::vector<TriGraph::Edge>* edges, std::vector<uint32_t>* queue) {
-  TriGraph graph(n_, m_);
-  for (ForwardIterator it = begin; it != end; ++it) { 
-    uint32_t h[3];
-    for (int i = 0; i < 3; ++i) h[i] = SeededHashFcn()(*it, hash_seed_[i]);
-    uint32_t v0 = h[0] % r_;
-    uint32_t v1 = h[1] % r_ + r_;
-    uint32_t v2 = h[2] % r_ + (r_ << 1);
-    // cerr << "Key: " << *it << " edge " <<  it - begin << " (" << v0 << "," << v1 << "," << v2 << ")" << endl;
-    graph.AddEdge(TriGraph::Edge(v0, v1, v2));
-  }
-  if (GenerateQueue(&graph, queue)) {
-     graph.ExtractEdgesAndClear(edges);
-     return true;
-  }
-  return false;
-}
-
-template <class SeededHashFcn, class Key>
-uint32_t MPHTable::index(const Key& key) const {
-  uint32_t h[3];
-  for (int i = 0; i < 3; ++i) h[i] = SeededHashFcn()(key, hash_seed_[i]);
-  h[0] = h[0] % r_;
-  h[1] = h[1] % r_ + r_;
-  h[2] = h[2] % r_ + (r_ << 1);
-  assert(g_size_);
-  // cerr << "g_.size() " << g_size_ << " h0 >> 2 " << (h[0] >> 2) << endl;
-  assert((h[0] >> 2) <g_size_);
-  assert((h[1] >> 2) <g_size_);
-  assert((h[2] >> 2) <g_size_);
-  uint32_t vertex = h[(get_2bit_value(g_, h[0]) + get_2bit_value(g_, h[1]) + get_2bit_value(g_, h[2])) % 3];
-  // cerr << "Search found vertex " << vertex << endl;
-  return Rank(vertex);
-}
-
-template <class Key, class HashFcn = typename seeded_hash<std::hash<Key> >::hash_function>
-class SimpleMPHTable : public MPHTable {
- public:
   template <class ForwardIterator>
-  bool Reset(ForwardIterator begin, ForwardIterator end) {
+  bool Reset(ForwardIterator begin, ForwardIterator end);
-    return MPHTable::Reset<HashFcn>(begin, end);
+ private:
-  }
+  char* data_;
-  uint32_t index(const Key& key) { return MPHTable::index<HashFcn>(key); }
+  vector<uint64_t> offsets_;
+  MPHIndex index_;
 };
-
-}  // namespace cxxmph
-
-#endif // __CXXMPH_MPH_TABLE_H__

src/bm_numbers.c

@@ -1,6 +1,10 @@
 #include <stdlib.h>
 #include <string.h>
 
+#include <ext/hash_set>
+using __gnu_cxx::hash_set;
+static const char cxx_name = "__gnu_cxx::hash_set";
+
 #include "bitbool.h"
 #include "cmph.h"
 #include "cmph_benchmark.h"
@@ -71,8 +75,8 @@ void bm_search(CMPH_ALGO algo, int iters) {
   cmph_t* mphf = NULL; 
 
   
-  snprintf(mphf_name, 128, "%s:%u", cmph_names[algo], iters);
+  snprintf(mphf_name, 128, "%s:%u", cxx_name, iters);
-  mphf = lsmap_search(g_created_mphf, mphf_name);
+  mphf = (cmph_t*)lsmap_search(g_created_mphf, mphf_name);
 
   cmph_uint32* count = (cmph_uint32*)malloc(sizeof(cmph_uint32)*iters);  
   cmph_uint32* hash_count = (cmph_uint32*)malloc(sizeof(cmph_uint32)*iters);  
@@ -102,6 +106,49 @@ DECLARE_ALGO(CMPH_BRZ);
 DECLARE_ALGO(CMPH_FCH);
 DECLARE_ALGO(CMPH_BDZ);
 
+void bm_create_ext_hash_set(int iters) {
+  cmph_uint32 i = 0;
+
+  if (iters > g_numbers_len) {
+    fprintf(stderr, "No input with proper size.");
+    exit(-1);
+  }
+
+  hash_set<cmph_uint32>* ext_hash_set = new hash_set<cmph_uint32>;
+  for (i = 0; i < iters; ++i) {
+    ext_hash_set->insert(g_numbers[i]);
+  }
+  lsmap_append(g_created_mphf, cxx_name, ext_hash_set);
+}
+
+void bm_search_ext_hash_set(int iters) {
+  cmph_uint32 i = 0;
+
+  if (iters > g_numbers_len) {
+    fprintf(stderr, "No input with proper size.");
+    exit(-1);
+  }
+
+  snprintf(mphf_name, 128, "%s:%u", hash_count, iters);
+  mphf = (__gnu_cxx::hash_set*)lsmap_search(g_created_mphf, mphf_name);
+
+  cmph_uint32* count = (cmph_uint32*)malloc(sizeof(cmph_uint32)*iters);  
+  cmph_uint32* hash_count = (cmph_uint32*)malloc(sizeof(cmph_uint32)*iters);  
+
+  for (i = 0; i < iters * 100; ++i) {
+    cmph_uint32 pos = random() % iters;
+    const char* buf = (const char*)(g_numbers + pos);
+    cmph_uint32 h = cmph_search(mphf, buf, sizeof(cmph_uint32));
+    ++count[pos];
+    ++hash_count[h];
+  }
+
+  // Verify correctness later.
+  lsmap_append(g_expected_probes, create_lsmap_key(algo, iters), count);
+  lsmap_append(g_mphf_probes, create_lsmap_key(algo, iters), hash_count);
+}
+}
+
 int main(int argc, char** argv) {
   g_numbers_len = 1000 * 1000;
   g_numbers = random_numbers_vector_new(g_numbers_len);