Upon changing the number of threads, make sure all threads are bound

The heuristic to avoid thread binding if less than 8 threads are requested resulted in the first 7 threads not being bound.
The branch was verified to yield a roughly 13% speedup by @CoffeeOne on the appropriate hardware and OS, and an earlier version of this patch tested well on his machine:

http://tests.stockfishchess.org/tests/view/5a3693480ebc590ccbb8be5a
ELO: 9.24 +-4.6 (95%) LOS: 100.0%
Total: 5000 W: 634 L: 501 D: 3865

To make sure all threads (including mainThread) are bound as soon as the total number exceeds 7, recreate all threads on a change of thread number.
To do this, unify Threads::init, Threads::exit and Threads::set are unified in a single Threads::set function that goes through the needed steps.
The code includes several suggestions from @joergoster.

Fixes issue #1312 

No functional change

src/main.cpp

@@ -45,11 +45,11 @@ int main(int argc, char* argv[]) {
   Pawns::init();
   Tablebases::init(Options["SyzygyPath"]);
   TT.resize(Options["Hash"]);
-  Threads.init(Options["Threads"]);
+  Threads.set(Options["Threads"]);
   Search::clear(); // After threads are up
 
   UCI::loop(argc, argv);
 
-  Threads.exit();
+  Threads.set(0);
   return 0;
 }

src/misc.cpp

@@ -293,14 +293,6 @@ int get_group(size_t idx) {
 
 void bindThisThread(size_t idx) {
 
-  // If OS already scheduled us on a different group than 0 then don't overwrite
-  // the choice, eventually we are one of many one-threaded processes running on
-  // some Windows NUMA hardware, for instance in fishtest. To make it simple,
-  // just check if running threads are below a threshold, in this case all this
-  // NUMA machinery is not needed.
-  if (Threads.size() < 8)
-      return;
-
   // Use only local variables to be thread-safe
   int group = get_group(idx);
 

src/search.cpp

@@ -173,13 +173,7 @@ void Search::clear() {
 
   Time.availableNodes = 0;
   TT.clear();
-
-  for (Thread* th : Threads)
-      th->clear();
-
-  Threads.main()->callsCnt = 0;
-  Threads.main()->previousScore = VALUE_INFINITE;
-  Threads.main()->previousTimeReduction = 1;
+  Threads.clear();
 }
 
 

src/thread.cpp

@@ -24,6 +24,7 @@
 #include "movegen.h"
 #include "search.h"
 #include "thread.h"
+#include "uci.h"
 #include "syzygy/tbprobe.h"
 
 ThreadPool Threads; // Global object
@@ -35,7 +36,6 @@ ThreadPool Threads; // Global object
 Thread::Thread(size_t n) : idx(n), stdThread(&Thread::idle_loop, this) {
 
   wait_for_search_finished();
-  clear(); // Zero-init histories (based on std::array)
 }
 
 
@@ -92,7 +92,13 @@ void Thread::wait_for_search_finished() {
 
 void Thread::idle_loop() {
 
-  WinProcGroup::bindThisThread(idx);
+  // If OS already scheduled us on a different group than 0 then don't overwrite
+  // the choice, eventually we are one of many one-threaded processes running on
+  // some Windows NUMA hardware, for instance in fishtest. To make it simple,
+  // just check if running threads are below a threshold, in this case all this
+  // NUMA machinery is not needed.
+  if (Options["Threads"] >= 8)
+      WinProcGroup::bindThisThread(idx);
 
   while (true)
   {
@@ -110,41 +116,39 @@ void Thread::idle_loop() {
   }
 }
 
-
-/// ThreadPool::init() creates and launches the threads that will go
-/// immediately to sleep in idle_loop. We cannot use the constructor because
-/// Threads is a static object and we need a fully initialized engine at
-/// this point due to allocation of Endgames in the Thread constructor.
-
-void ThreadPool::init(size_t requested) {
-
-  push_back(new MainThread(0));
-  set(requested);
-}
-
-
-/// ThreadPool::exit() terminates threads before the program exits. Cannot be
-/// done in the destructor because threads must be terminated before deleting
-/// any static object, so before main() returns.
-
-void ThreadPool::exit() {
-
-  main()->wait_for_search_finished();
-  set(0);
-}
-
-
-/// ThreadPool::set() creates/destroys threads to match the requested number
+/// ThreadPool::set() creates/destroys threads to match the requested number.
+/// Created and launced threads wil go immediately to sleep in idle_loop.
+/// Upon resizing, threads are recreated to allow for binding if necessary.
 
 void ThreadPool::set(size_t requested) {
 
-  while (size() < requested)
-      push_back(new Thread(size()));
+  if (size() > 0) { // destroy any existing thread(s)
+      main()->wait_for_search_finished();
 
-  while (size() > requested)
-      delete back(), pop_back();
+      while (size() > 0)
+          delete back(), pop_back();
+  }
+
+  if (requested > 0) { // create new thread(s)
+      push_back(new MainThread(0));
+
+      while (size() < requested)
+          push_back(new Thread(size()));
+      clear();
+  }
 }
 
+/// ThreadPool::clear() sets threadPool data to initial values.
+
+void ThreadPool::clear() {
+
+  for (Thread* th : *this)
+      th->clear();
+
+  main()->callsCnt = 0;
+  main()->previousScore = VALUE_INFINITE;
+  main()->previousTimeReduction = 1;
+}
 
 /// ThreadPool::start_thinking() wakes up main thread waiting in idle_loop() and
 /// returns immediately. Main thread will wake up other threads and start the search.
@@ -181,7 +185,7 @@ void ThreadPool::start_thinking(Position& pos, StateListPtr& states,
   // is shared by threads but is accessed in read-only mode.
   StateInfo tmp = setupStates->back();
 
-  for (Thread* th : Threads)
+  for (Thread* th : *this)
   {
       th->nodes = th->tbHits = 0;
       th->rootDepth = th->completedDepth = DEPTH_ZERO;

src/thread.h

@@ -96,9 +96,8 @@ struct MainThread : public Thread {
 
 struct ThreadPool : public std::vector<Thread*> {
 
-  void init(size_t); // No constructor and destructor, threads rely on globals that should
-  void exit();       // be initialized and valid during the whole thread lifetime.
   void start_thinking(Position&, StateListPtr&, const Search::LimitsType&, bool = false);
+  void clear();
   void set(size_t);
 
   MainThread* main()        const { return static_cast<MainThread*>(front()); }