/*
Stockfish, a UCI chess playing engine derived from Glaurung 2.1
Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
- Copyright (C) 2008-2009 Marco Costalba
+ Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad
+ Copyright (C) 2015-2016 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
Stockfish is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
-
Stockfish is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
-
-#if !defined(BITCOUNT_H_INCLUDED)
+#ifndef 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 "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
-
+#include <cassert>
-// Select type of intrinsic bit count instruction to use
+#include "types.h"
-#if defined(_MSC_VER) && defined(_WIN64) // Microsoft compiler
+enum BitCountType {
+ CNT_64,
+ CNT_64_MAX15,
+ CNT_32,
+ CNT_32_MAX15,
+ CNT_HW_POPCNT
+};
-#include <intrin.h>
+/// Determine at compile time the best popcount<> specialization according to
+/// whether the platform is 32 or 64 bit, the maximum number of non-zero
+/// bits to count and if the hardware popcnt instruction is available.
+const BitCountType Full = HasPopCnt ? CNT_HW_POPCNT : Is64Bit ? CNT_64 : CNT_32;
+const BitCountType Max15 = HasPopCnt ? CNT_HW_POPCNT : Is64Bit ? CNT_64_MAX15 : CNT_32_MAX15;
-inline bool cpu_has_popcnt() {
- int CPUInfo[4] = {-1};
- __cpuid(CPUInfo, 0x00000001);
- 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
-
-#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)
-#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 BITSCAN_INTRINSIC(idx, x) sw_count_1s(x) // dummy
-
-#endif
+/// popcount() counts the number of non-zero bits in a bitboard
+template<BitCountType> inline int popcount(Bitboard);
-
-/// 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;
+template<>
+inline int popcount<CNT_64>(Bitboard b) {
+ b -= (b >> 1) & 0x5555555555555555ULL;
+ b = ((b >> 2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
+ b = ((b >> 4) + b) & 0x0F0F0F0F0F0F0F0FULL;
+ return (b * 0x0101010101010101ULL) >> 56;
}
-inline int sw_count_1s_max_15(Bitboard b) {
- return count_1s(b);
+template<>
+inline int popcount<CNT_64_MAX15>(Bitboard b) {
+ b -= (b >> 1) & 0x5555555555555555ULL;
+ b = ((b >> 2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
+ return (b * 0x1111111111111111ULL) >> 60;
}
-#elif defined(BITCOUNT_SWAR_32)
-
-inline int sw_count_1s(Bitboard b) {
+template<>
+inline int popcount<CNT_32>(Bitboard b) {
unsigned w = unsigned(b >> 32), v = unsigned(b);
- v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
- w -= (w >> 1) & 0x55555555;
- v = ((v >> 2) & 0x33333333) + (v & 0x33333333); // 0-4 in 4 bits
- w = ((w >> 2) & 0x33333333) + (w & 0x33333333);
- v = ((v >> 4) + v) & 0x0F0F0F0F; // 0-8 in 8 bits
- v += (((w >> 4) + w) & 0x0F0F0F0F); // 0-16 in 8 bits
- v *= 0x01010101; // mul is fast on amd procs
- return int(v >> 24);
+ v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
+ w -= (w >> 1) & 0x55555555;
+ v = ((v >> 2) & 0x33333333) + (v & 0x33333333); // 0-4 in 4 bits
+ w = ((w >> 2) & 0x33333333) + (w & 0x33333333);
+ v = ((v >> 4) + v + (w >> 4) + w) & 0x0F0F0F0F;
+ return (v * 0x01010101) >> 24;
}
-inline int sw_count_1s_max_15(Bitboard b) {
+template<>
+inline int popcount<CNT_32_MAX15>(Bitboard b) {
unsigned w = unsigned(b >> 32), v = unsigned(b);
- v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
- w -= (w >> 1) & 0x55555555;
- v = ((v >> 2) & 0x33333333) + (v & 0x33333333); // 0-4 in 4 bits
- w = ((w >> 2) & 0x33333333) + (w & 0x33333333);
- v += w; // 0-8 in 4 bits
- v *= 0x11111111;
- 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);
+ v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
+ w -= (w >> 1) & 0x55555555;
+ v = ((v >> 2) & 0x33333333) + (v & 0x33333333); // 0-4 in 4 bits
+ w = ((w >> 2) & 0x33333333) + (w & 0x33333333);
+ return ((v + w) * 0x11111111) >> 28;
}
-#endif // BITCOUNT
+template<>
+inline int popcount<CNT_HW_POPCNT>(Bitboard b) {
+#ifndef USE_POPCNT
-/// count_1s() counts the number of nonzero bits in a bitboard.
-/// If template parameter is true an intrinsic is called, otherwise
-/// we fallback on a software implementation.
+ assert(false);
+ return b != 0; // Avoid 'b not used' warning
-template<bool UseIntrinsic>
-inline int count_1s(Bitboard b) {
+#elif defined(_MSC_VER) && defined(__INTEL_COMPILER)
- return UseIntrinsic ? POPCNT_INTRINSIC(b) : sw_count_1s(b);
-}
+ return _mm_popcnt_u64(b);
-template<bool UseIntrinsic>
-inline int count_1s_max_15(Bitboard b) {
+#elif defined(_MSC_VER)
- return UseIntrinsic ? POPCNT_INTRINSIC(b) : sw_count_1s_max_15(b);
-}
+ return (int)__popcnt64(b);
+#else // Assumed gcc or compatible compiler
-// Global variable 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
-const bool CpuHasPOPCNT = cpu_has_popcnt();
-#endif
+ return __builtin_popcountll(b);
-
-// 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)
-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<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);
}
-#endif // !defined(BITCOUNT_H_INCLUDED)
+#endif // #ifndef BITCOUNT_H_INCLUDED
projects / stockfish / blobdiff