Use compiler name lookup to simplify code

[stockfish] / src / bitcount.h

diff --git a/src/bitcount.h b/src/bitcount.h
index c6e969a82411d4b37dc1633e512ca5873d4a2912..871247abcf4be72b16df1def8d04f97a6e6d8f52 100644 (file)
--- a/src/bitcount.h
+++ b/src/bitcount.h
@@ -66,6 +66,7 @@ inline bool cpu_has_popcnt() {
 }
 
 #define POPCNT_INTRINSIC(x) __popcnt64(x)
+#define BITSCAN_INTRINSIC(idx, x) _BitScanForward64(idx, x)
 
 #elif defined(__INTEL_COMPILER) && (defined(__x86_64) || defined(_M_X64)) // Intel compiler
 
@@ -79,12 +80,14 @@ inline bool cpu_has_popcnt() {
 }
 
 #define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x)
+#define BITSCAN_INTRINSIC(idx, x) _BitScanForward64(idx, x)
 
 #else // Safe fallback for unsupported compilers
 
 inline bool cpu_has_popcnt() { return false; }
 
-#define POPCNT_INTRINSIC(x) sw_count_1s(x)
+#define POPCNT_INTRINSIC(x) count_1s(x)
+#define BITSCAN_INTRINSIC(idx, x) count_1s(x) // dummy
 
 #endif
 
@@ -93,19 +96,19 @@ inline bool cpu_has_popcnt() { return false; }
 
 #if defined(BITCOUNT_LOOP)
 
-inline int sw_count_1s(Bitboard b) {
+inline int count_1s(Bitboard b) {
   int r;
   for(r = 0; b; r++, b &= b - 1);
   return r;
 }
 
-inline int sw_count_1s_max_15(Bitboard b) {
+inline int count_1s_max_15(Bitboard b) {
   return count_1s(b);
 }
 
 #elif defined(BITCOUNT_SWAR_32)
 
-inline int sw_count_1s(Bitboard b) {
+inline int count_1s(Bitboard b) {
   unsigned w = unsigned(b >> 32), v = unsigned(b);
   v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
   w -= (w >> 1) & 0x55555555;
@@ -117,7 +120,7 @@ inline int sw_count_1s(Bitboard b) {
   return int(v >> 24);
 }
 
-inline int sw_count_1s_max_15(Bitboard b) {
+inline int count_1s_max_15(Bitboard b) {
   unsigned w = unsigned(b >> 32), v = unsigned(b);
   v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
   w -= (w >> 1) & 0x55555555;
@@ -130,7 +133,7 @@ inline int sw_count_1s_max_15(Bitboard b) {
 
 #elif defined(BITCOUNT_SWAR_64)
 
-inline int sw_count_1s(Bitboard b) {
+inline int count_1s(Bitboard b) {
   b -= ((b>>1) & 0x5555555555555555ULL);
   b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
   b = ((b>>4) + b) & 0x0F0F0F0F0F0F0F0FULL;
@@ -138,7 +141,7 @@ inline int sw_count_1s(Bitboard b) {
   return int(b >> 56);
 }
 
-inline int sw_count_1s_max_15(Bitboard b) {
+inline int count_1s_max_15(Bitboard b) {
   b -= (b>>1) & 0x5555555555555555ULL;
   b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
   b *= 0x1111111111111111ULL;
@@ -155,13 +158,34 @@ inline int sw_count_1s_max_15(Bitboard b) {
 template<bool UseIntrinsic>
 inline int count_1s(Bitboard b) {
 
-  return UseIntrinsic ? POPCNT_INTRINSIC(b) : sw_count_1s(b);
+  return UseIntrinsic ? POPCNT_INTRINSIC(b) : count_1s(b);
 }
 
 template<bool UseIntrinsic>
 inline int count_1s_max_15(Bitboard b) {
 
-  return UseIntrinsic ? POPCNT_INTRINSIC(b) : sw_count_1s_max_15(b);
+  return UseIntrinsic ? POPCNT_INTRINSIC(b) : count_1s_max_15(b);
+}
+
+
+/// pop_1st_bit() finds and clears the least significant nonzero bit in a
+/// nonzero bitboard. If template parameter is true an intrinsic is called,
+/// otherwise we fallback on a software implementation.
+
+template<bool UseIntrinsic>
+inline Square pop_1st_bit(Bitboard *b) {
+
+  return pop_1st_bit(b);
+}
+
+template<>
+inline Square pop_1st_bit<true>(Bitboard *b) {
+
+  unsigned long idx;
+  Bitboard bb = *b;
+  BITSCAN_INTRINSIC(&idx, bb);
+  *b &= (bb - 1);
+  return Square(idx);
 }