X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Fbitcount.h;h=43895180c1e9ae7566e16e3f35f29b93e00e18eb;hp=41a1446e01b700b7d6fa08b9425b1a199a2a9122;hb=92bada1a32c7ebd5f5b4438bd85bb003d20bd046;hpb=f90f810ac4de5ea2f5582ca05a9354c33971a953 diff --git a/src/bitcount.h b/src/bitcount.h index 41a1446e..43895180 100644 --- a/src/bitcount.h +++ b/src/bitcount.h @@ -22,41 +22,23 @@ #if !defined(BITCOUNT_H_INCLUDED) #define BITCOUNT_H_INCLUDED -// To disable POPCNT support uncomment following line. You should do it only -// in PGO compiling to exercise the default fallback path. Don't forget to -// re-comment the line for the final optimized compile though ;-) -//#define DISABLE_POPCNT_SUPPORT +#include "types.h" +// Select type of intrinsic bit count instruction to use, see +// README.txt on how to pgo compile with POPCNT support. -#include "bitboard.h" - - -// 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 +#if defined(__INTEL_COMPILER) && defined(USE_POPCNT) // Intel compiler +inline bool cpu_has_popcnt() { -// Select type of intrinsic bit count instruction to use + int CPUInfo[4] = {-1}; + __cpuid(CPUInfo, 0x00000001); + return (CPUInfo[2] >> 23) & 1; +} -#if defined(_MSC_VER) && defined(_WIN64) // Microsoft compiler +#define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x) -#include +#elif defined(_MSC_VER) && defined(USE_POPCNT) // Microsoft compiler inline bool cpu_has_popcnt() { @@ -65,50 +47,53 @@ inline bool cpu_has_popcnt() { return (CPUInfo[2] >> 23) & 1; } -#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 +#define POPCNT_INTRINSIC(x) (int)__popcnt64(x) -#include +#elif defined(__GNUC__) && defined(USE_POPCNT) // Gcc compiler inline bool cpu_has_popcnt() { - int CPUInfo[4] = {-1}; - __cpuid(CPUInfo, 0x00000001); - return (CPUInfo[2] >> 23) & 1; + unsigned int eax, ebx, ecx, edx; + __cpuid(1, &eax, &ebx, &ecx, &edx); + return (ecx >> 23) & 1; } -#define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x) -#define BITSCAN_INTRINSIC(idx, x) _BitScanForward64(idx, x) +#define POPCNT_INTRINSIC(x) ({ \ + unsigned long __ret; \ + __asm__("popcnt %1, %0" : "=r" (__ret) : "r" (x)); \ + __ret; }) -#else // Safe fallback for unsupported compilers +#else // Safe fallback for unsupported compilers or when USE_POPCNT is disabled inline bool cpu_has_popcnt() { return false; } -#define POPCNT_INTRINSIC(x) sw_count_1s(x) -#define BITSCAN_INTRINSIC(idx, x) sw_count_1s(x) // dummy +#define POPCNT_INTRINSIC(x) 0 -#endif +#endif // cpu_has_popcnt() and POPCNT_INTRINSIC() definitions /// Software implementation of bit count functions -#if defined(BITCOUNT_LOOP) +#if defined(IS_64BIT) -inline int sw_count_1s(Bitboard b) { - int r; - for(r = 0; b; r++, b &= b - 1); - return r; +inline int count_1s(Bitboard b) { + b -= ((b>>1) & 0x5555555555555555ULL); + b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL); + b = ((b>>4) + b) & 0x0F0F0F0F0F0F0F0FULL; + b *= 0x0101010101010101ULL; + return int(b >> 56); } -inline int sw_count_1s_max_15(Bitboard b) { - return count_1s(b); +inline int count_1s_max_15(Bitboard b) { + b -= (b>>1) & 0x5555555555555555ULL; + b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL); + b *= 0x1111111111111111ULL; + return int(b >> 60); } -#elif defined(BITCOUNT_SWAR_32) +#else // if !defined(IS_64BIT) -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; @@ -120,7 +105,7 @@ 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; @@ -131,23 +116,6 @@ inline int sw_count_1s_max_15(Bitboard b) { return int(v >> 28); } -#elif defined(BITCOUNT_SWAR_64) - -inline int sw_count_1s(Bitboard b) { - b -= ((b>>1) & 0x5555555555555555ULL); - b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL); - b = ((b>>4) + b) & 0x0F0F0F0F0F0F0F0FULL; - b *= 0x0101010101010101ULL; - return int(b >> 56); -} - -inline int sw_count_1s_max_15(Bitboard b) { - b -= (b>>1) & 0x5555555555555555ULL; - b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL); - b *= 0x1111111111111111ULL; - return int(b >> 60); -} - #endif // BITCOUNT @@ -158,53 +126,28 @@ inline int sw_count_1s_max_15(Bitboard b) { template inline int count_1s(Bitboard b) { - return UseIntrinsic ? POPCNT_INTRINSIC(b) : sw_count_1s(b); + return UseIntrinsic ? POPCNT_INTRINSIC(b) : count_1s(b); } template 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); } -// Global variable initialized at startup that is set to true if +// Global constant initialized at startup that is set to true if // CPU on which application runs supports POPCNT intrinsic. Unless -// DISABLE_POPCNT_SUPPORT is defined. -#if defined(DISABLE_POPCNT_SUPPORT) -const bool CpuHasPOPCNT = false; -#else +// USE_POPCNT is not defined. const bool CpuHasPOPCNT = cpu_has_popcnt(); -#endif -// Global variable used to print info about the use of 64 optimized -// functions to verify that a 64bit compile has been correctly built. -#if defined(BITCOUNT_SWAR_64) +// Global constant used to print info about the use of 64 optimized +// functions to verify that a 64 bit compile has been correctly built. +#if defined(IS_64BIT) const bool CpuHas64BitPath = true; #else const bool CpuHas64BitPath = false; #endif - -/// 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 -inline Square pop_1st_bit(Bitboard *b) { - - return pop_1st_bit(b); -} - -template<> -inline Square pop_1st_bit(Bitboard *b) { - - unsigned long idx; - Bitboard bb = *b; - BITSCAN_INTRINSIC(&idx, bb); - *b &= (bb - 1); - return Square(idx); -} - #endif // !defined(BITCOUNT_H_INCLUDED)