Micro optimize count_1s_max_15() for 32 bit system

[stockfish] / src / bitboard.h

diff --git a/src/bitboard.h b/src/bitboard.h
index b193786cb8e28eae2096cbcb9095b14dd006cf5c..1140664838b47bfb875a4a4841b7c07f90c1e7e3 100644 (file)
--- a/src/bitboard.h
+++ b/src/bitboard.h
@@ -236,6 +236,19 @@ inline Bitboard in_front_bb(Color c, Square s) {
 }
 
 
+/// behind_bb() takes a color and a rank or square as input, and returns a
+/// bitboard representing all the squares on all ranks behind of the rank
+/// (or square), from the given color's point of view.
+
+inline Bitboard behind_bb(Color c, Rank r) {
+  return InFrontBB[opposite_color(c)][r];
+}
+
+inline Bitboard behind_bb(Color c, Square s) {
+  return in_front_bb(opposite_color(c), square_rank(s));
+}
+
+
 /// ray_bb() gives a bitboard representing all squares along the ray in a
 /// given direction from a given square.
 
@@ -384,11 +397,10 @@ inline int count_1s(Bitboard b) {
   unsigned w = unsigned(b >> 32), v = unsigned(b);
   v = v - ((v >> 1) & 0x55555555);
   w = w - ((w >> 1) & 0x55555555);
+  v += w;
   v = (v & 0x33333333) + ((v >> 2) & 0x33333333);
-  w = (w & 0x33333333) + ((w >> 2) & 0x33333333);
   v = (v + (v >> 4)) & 0x0F0F0F0F;
-  w = (w + (w >> 4)) & 0x0F0F0F0F;
-  v = ((v+w) * 0x01010101) >> 24; // mul is fast on amd procs
+  v = (v * 0x01010101) >> 24; // mul is fast on amd procs
   return int(v);
 }
 
@@ -396,9 +408,9 @@ inline int count_1s_max_15(Bitboard b) {
   unsigned w = unsigned(b >> 32), v = unsigned(b);
   v = v - ((v >> 1) & 0x55555555);
   w = w - ((w >> 1) & 0x55555555);
+  v += w;
   v = (v & 0x33333333) + ((v >> 2) & 0x33333333);
-  w = (w & 0x33333333) + ((w >> 2) & 0x33333333);
-  v = ((v+w) * 0x11111111) >> 28;
+  v = (v * 0x11111111) >> 28;
   return int(v);
 }