made insertion and constant visitation (but *not* erasure) optionally latch-free

include/boost/unordered/detail/foa/concurrent_table.hpp

@@ -83,6 +83,8 @@ public:
   cache_aligned_array(const cache_aligned_array&)=delete;
   cache_aligned_array& operator=(const cache_aligned_array&)=delete;
 
+  constexpr std::size_t size()const noexcept{return N;}
+
   T& operator[](std::size_t pos)noexcept{return *data(pos);}
 
 private:
@@ -149,14 +151,12 @@ template<typename Mutex>
 class lock_guard
 {
 public:
-  lock_guard(Mutex& m_)noexcept:m(m_){m.lock();}
-  ~lock_guard()noexcept{m.unlock();}
-
-  /* not used but VS in pre-C++17 mode needs to see it for RVO */
-  lock_guard(const lock_guard&);
+  lock_guard(Mutex& m)noexcept:pm(&m){pm->lock();}
+  lock_guard(lock_guard&& x)noexcept:pm(x.pm){x.pm=nullptr;}
+  ~lock_guard()noexcept{if(pm)pm->unlock();}
 
 private:
-  Mutex &m;
+  Mutex* pm;
 };
 
 /* inspired by boost/multi_index/detail/scoped_bilock.hpp */
@@ -970,8 +970,30 @@ private:
   using exclusive_lock_guard=reentrancy_checked<lock_guard<multimutex_type>>;
   using exclusive_bilock_guard=
     reentrancy_bichecked<scoped_bilock<multimutex_type>>;
+
+#if defined(BOOST_UNORDERED_LATCH_FREE)
+  struct group_shared_lock_guard
+  {
+    group_shared_lock_guard(std::atomic_int& cs_):cs{cs_}{cs.store(1);}
+    ~group_shared_lock_guard(){cs.store(0);}
+
+    std::atomic_int& cs;
+  };
+  struct group_exclusive_lock_guard
+  {
+    group_exclusive_lock_guard(
+      std::atomic_int& cs_,group_access::exclusive_lock_guard&& lck_):
+      cs{cs_},lck{std::move(lck_)}{cs.store(1);}
+    ~group_exclusive_lock_guard(){cs.store(0);}
+
+    std::atomic_int&                   cs;
+    group_access::exclusive_lock_guard lck;
+  };
+#else
   using group_shared_lock_guard=typename group_access::shared_lock_guard;
   using group_exclusive_lock_guard=typename group_access::exclusive_lock_guard;
+#endif
+
   using group_insert_counter_type=typename group_access::insert_counter_type;
 
   concurrent_table(const concurrent_table& x,exclusive_lock_guard):
@@ -985,9 +1007,15 @@ private:
     concurrent_table&& x,const Allocator& al_,exclusive_lock_guard):
     super{std::move(x),al_}{}
 
+  static inline std::size_t thread_id()
+  {
+    thread_local auto id=(++thread_counter);
+    return id;
+  }
+
   inline shared_lock_guard shared_access()const
   {
-    thread_local auto id=(++thread_counter)%mutexes.size();
+    thread_local auto id=thread_id()%mutexes.size();
 
     return shared_lock_guard{this,mutexes[id]};
   }
@@ -1018,13 +1046,24 @@ private:
 
   inline group_shared_lock_guard access(group_shared,std::size_t pos)const
   {
+#if defined(BOOST_UNORDERED_LATCH_FREE)
+    return {csections[thread_id()%csections.size()]};
+#else
     return this->arrays.group_accesses()[pos].shared_access();
+#endif
   }
 
   inline group_exclusive_lock_guard access(
     group_exclusive,std::size_t pos)const
   {
+#if defined(BOOST_UNORDERED_LATCH_FREE)
+    return {
+      csections[thread_id()%csections.size()],
+      this->arrays.group_accesses()[pos].exclusive_access()
+    };
+#else
     return this->arrays.group_accesses()[pos].exclusive_access();
+#endif
   }
 
   inline group_insert_counter_type& insert_counter(std::size_t pos)const
@@ -1436,6 +1475,55 @@ private:
     bool        commit_=false;
   };
 
+#if defined(BOOST_UNORDERED_LATCH_FREE)
+  template<typename GroupAccessMode,typename F,typename... Args>
+  BOOST_FORCEINLINE int
+  unprotected_norehash_emplace_or_visit(
+    GroupAccessMode access_mode,F&& f,Args&&... args)
+  {
+    const auto &k=this->key_from(std::forward<Args>(args)...);
+    auto        hash=this->hash_for(k);
+    auto        pos0=this->position_for(hash);
+
+    for(;;){
+    startover:
+      boost::uint32_t counter=insert_counter(pos0);
+      if(unprotected_visit(
+        access_mode,k,pos0,hash,std::forward<F>(f)))return 0;
+
+      reserve_size rsize(*this);
+      if(BOOST_LIKELY(rsize.succeeded())){
+        for(prober pb(pos0);;pb.next(this->arrays.groups_size_mask)){
+          auto pos=pb.get();
+          auto pg=this->arrays.groups()+pos;
+          auto mask=pg->match_available();
+          if(BOOST_LIKELY(mask!=0)){
+            auto n=unchecked_countr_zero(mask);
+            unsigned char expected=0;
+            if(!reinterpret_cast<std::atomic<unsigned char>*>(pg)[n].
+               compare_exchange_weak(expected,1)){
+              /* slot wasn't empty */
+              goto startover;
+            }
+            wait_for_csections(); // WHY BEFORE THE FOLLOWING?
+            if(BOOST_UNLIKELY(insert_counter(pos0)++!=counter)){
+              /* other thread inserted from pos0, need to start over */
+              pg->reset(n);
+              goto startover;
+            }
+            auto p=this->arrays.elements()+pos*N+n;
+            this->construct_element(p,std::forward<Args>(args)...);
+            pg->set(n,hash);
+            rsize.commit();
+            return 1;
+          }
+          pg->mark_overflow(hash);
+        }
+      }
+      else return -1;
+    }
+  }
+#else
   template<typename GroupAccessMode,typename F,typename... Args>
   BOOST_FORCEINLINE int
   unprotected_norehash_emplace_or_visit(
@@ -1477,6 +1565,7 @@ private:
       else return -1;
     }
   }
+#endif
 
   void rehash_if_full()
   {
@@ -1724,6 +1813,18 @@ private:
 
   static std::atomic<std::size_t> thread_counter;
   mutable multimutex_type         mutexes;
+
+#if defined(BOOST_UNORDERED_LATCH_FREE)
+  mutable cache_aligned_array<
+    std::atomic_int,128>          csections;
+
+  void wait_for_csections()const
+  {
+    for(std::size_t i=0;i<csections.size();++i){
+      while(csections[i].load(std::memory_order_acquire)){}
+    }
+  }
+#endif
 };
 
 template<typename T,typename H,typename P,typename A>