-// Select type of software bit count function to use
-
-#if !defined(AUTO_CONFIGURATION) || defined(IS_64BIT)
-
-//#define USE_COMPACT_ROOK_ATTACKS
-//#define USE_32BIT_ATTACKS
-#define USE_FOLDED_BITSCAN
-
-#define BITCOUNT_SWAR_64
-//#define BITCOUNT_SWAR_32
-//#define BITCOUNT_LOOP
-
-#else
-
-#define USE_32BIT_ATTACKS
-#define USE_FOLDED_BITSCAN
-#define BITCOUNT_SWAR_32
-
-#endif
-
-
-// Select type of intrinsic bit count instruction to use
-
-#if defined(_MSC_VER) && defined(_WIN64) // Microsoft compiler
-
-#include <intrin.h>
-
-inline bool cpu_has_popcnt() {
-
- int CPUInfo[4] = {-1};
- __cpuid(CPUInfo, 0x00000001);
- return (CPUInfo[2] >> 23) & 1;
-}
-
-#define POPCNT_INTRINSIC(x) __popcnt64(x)
-
-#elif defined(__INTEL_COMPILER) && (defined(__x86_64) || defined(_M_X64)) // Intel compiler
-
-#include <nmmintrin.h>
-
-inline bool cpu_has_popcnt() {
-
- int CPUInfo[4] = {-1};
- __cpuid(CPUInfo, 0x00000001);
- return (CPUInfo[2] >> 23) & 1;
-}
-
-#define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x)
-
-#else // Safe fallback for unsupported compilers
-
-inline bool cpu_has_popcnt() { return false; }
-
-#define POPCNT_INTRINSIC(x) sw_count_1s(x)
-
-#endif
-
-
-/// Software implementation of bit count functions
-
-#if defined(BITCOUNT_LOOP)
-
-inline int sw_count_1s(Bitboard b) {
- int r;
- for(r = 0; b; r++, b &= b - 1);
- return r;
+/// count_1s() counts the number of nonzero bits in a bitboard.
+/// We have different optimized versions according if platform
+/// is 32 or 64 bits, and to the maximum number of nonzero bits.
+/// We also support hardware popcnt instruction. See Readme.txt
+/// on how to pgo compile with popcnt support.
+template<BitCountType> inline int count_1s(Bitboard);
+
+template<>
+inline int count_1s<CNT64>(Bitboard b) {
+ b -= ((b>>1) & 0x5555555555555555ULL);
+ b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
+ b = ((b>>4) + b) & 0x0F0F0F0F0F0F0F0FULL;
+ b *= 0x0101010101010101ULL;
+ return int(b >> 56);