From: Marco Costalba <mcostalba@gmail.com>
Date: Mon, 23 Feb 2015 18:22:37 +0000 (+0100)
Subject: Improve spinlock implementation
X-Git-Url: https://git.sesse.net/?p=stockfish;a=commitdiff_plain;h=d3d26a94b3e501459a610b8d730394125c9afd45

Improve spinlock implementation

Calling lock.test_and_set() in a tight loop creates expensive
memory synchronizations among processors and penalize other
running threads. So syncronize only only once at the beginning
with fetch_sub() and then loop on a simple load() that puts much
less pressure on the system.

Reported about 2-3% speed up on various systems.

Patch by Ronald de Man.

No functional change.
---

diff --git a/src/thread.h b/src/thread.h
index f8e65941..ea645b45 100644
--- a/src/thread.h
+++ b/src/thread.h
@@ -39,16 +39,20 @@ const size_t MAX_THREADS = 128;
 const size_t MAX_SPLITPOINTS_PER_THREAD = 8;
 const size_t MAX_SLAVES_PER_SPLITPOINT = 4;
 
-/// Spinlock class wraps low level atomic operations to provide spin lock functionality
+
+/// Spinlock class wraps low level atomic operations to provide a spin lock
 
 class Spinlock {
 
-  std::atomic_flag lock;
+  std::atomic_int lock;
 
 public:
-  Spinlock() { std::atomic_flag_clear(&lock); }
-  void acquire() { while (lock.test_and_set(std::memory_order_acquire)) {} }
-  void release() { lock.clear(std::memory_order_release); }
+  Spinlock() { lock = 1; } // Init here to workaround a bug with MSVC 2013
+  void acquire() {
+    while (lock.fetch_sub(1, std::memory_order_acquire) != 1)
+        while (lock.load(std::memory_order_relaxed) <= 0) {}
+  }
+  void release() { lock.store(1, std::memory_order_release); }
 };