From: Marco Costalba <mcostalba@gmail.com>
Date: Tue, 24 Nov 2009 12:39:47 +0000 (+0100)
Subject: Speed up sorting of non-captures
X-Git-Url: https://git.sesse.net/?p=stockfish;a=commitdiff_plain;h=55b5b03273e9740b63663ee84a17788c5424eec3

Speed up sorting of non-captures

Becasue we have a lot of zero scores (around 30% of moves)
it is a good idea to do a couple a presorting loops across
the move list and shuffle the moves a bit so that with a
small effort we end up with 3 groups of moves: positives
scores, zero scores and negative scores.

We have two advantages

1) We don't need to sort zero scores

2) Sort two small groups is faster then sort a single big one

Speed up is of about 2%

Because equal scored moves could be reordered in a different way
this is not a "no functional change" although I have verified
the output list is always correctly sorted.

Signed-off-by: Marco Costalba <mcostalba@gmail.com>
---

diff --git a/src/move.h b/src/move.h
index d70bb81e..9702d9fb 100644
--- a/src/move.h
+++ b/src/move.h
@@ -65,9 +65,9 @@ struct MoveStack {
 // Note that operator< is set up such that sorting will be in descending order
 inline bool operator<(const MoveStack& f, const MoveStack& s) { return s.score < f.score; }
 
-// Our stable insertion sort in range [firstMove, lastMove), platform independent
+// An helper insertion sort implementation
 template<typename T>
-inline void sort_moves(T* firstMove, T* lastMove)
+inline void insertion_sort(T* firstMove, T* lastMove)
 {
     T value;
     T *cur, *p, *d;
@@ -86,6 +86,59 @@ inline void sort_moves(T* firstMove, T* lastMove)
         }
 }
 
+// Our dedicated sort in range [firstMove, lastMove), it is well
+// tuned for non-captures where we have a lot of zero scored moves.
+template<typename T>
+inline void sort_moves(T* firstMove, T* lastMove)
+{
+    T tmp;
+    T *p, *d;
+
+    d = lastMove;
+    p = firstMove - 1;
+
+    d->score = -1; // right guard
+
+    // Split positives vs non-positives
+    do {
+        while ((++p)->score > 0);
+
+        if (p != d)
+        {
+            while (--d != p && d->score <= 0);
+
+            tmp = *p;
+            *p = *d;
+            *d = tmp;
+        }
+
+    } while (p != d);
+
+    // Sort positives
+    insertion_sort<T>(firstMove, p);
+
+    d = lastMove;
+    p--;
+
+    // Split zero vs negatives
+    do {
+        while ((++p)->score == 0);
+
+        if (p != d)
+        {
+            while (--d != p && d->score < 0);
+
+            tmp = *p;
+            *p = *d;
+            *d = tmp;
+        }
+
+    } while (p != d);
+
+    // Sort negatives
+    insertion_sort<T>(p, lastMove);
+}
+
 // Picks up the best move in range [curMove, lastMove), one per cycle.
 // It is faster then sorting all the moves in advance when moves are few,
 // as normally are the possible captures. Note that is not a stable alghoritm.