2 Stockfish, a UCI chess playing engine derived from Glaurung 2.1
3 Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
4 Copyright (C) 2008-2009 Marco Costalba
6 Stockfish is free software: you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation, either version 3 of the License, or
9 (at your option) any later version.
12 Stockfish is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>.
22 #if !defined(BITCOUNT_H_INCLUDED)
23 #define BITCOUNT_H_INCLUDED
25 // To enable POPCNT support uncomment USE_POPCNT define. For PGO compile on a Core i7
26 // you may want to collect profile data first with USE_POPCNT disabled and then, in a
27 // second profiling session, with USE_POPCNT enabled so to exercise both paths. Don't
28 // forget to leave USE_POPCNT enabled for the final optimized compile though ;-)
35 // Select type of intrinsic bit count instruction to use
37 #if defined(__INTEL_COMPILER) && defined(IS_64BIT) && defined(USE_POPCNT) // Intel compiler
39 #include <nmmintrin.h>
41 inline bool cpu_has_popcnt() {
43 int CPUInfo[4] = {-1};
44 __cpuid(CPUInfo, 0x00000001);
45 return (CPUInfo[2] >> 23) & 1;
48 // Define a dummy template to workaround a compile error if _mm_popcnt_u64() is not defined.
50 // If _mm_popcnt_u64() is defined in <nmmintrin.h> it will be choosen first due to
51 // C++ overload rules that always prefer a function to a template with the same name.
52 // If not, we avoid a compile error and because cpu_has_popcnt() should return false,
53 // our templetized _mm_popcnt_u64() is never called anyway.
54 template<typename T> inline unsigned _mm_popcnt_u64(T) { return 0; } // Is never called
56 #define POPCNT_INTRINSIC(x) _mm_popcnt_u64(x)
58 #elif defined(_MSC_VER) && defined(IS_64BIT) && defined(USE_POPCNT) // Microsoft compiler
62 inline bool cpu_has_popcnt() {
64 int CPUInfo[4] = {-1};
65 __cpuid(CPUInfo, 0x00000001);
66 return (CPUInfo[2] >> 23) & 1;
69 // See comment of _mm_popcnt_u64<>() few lines above for an explanation.
70 template<typename T> inline unsigned __popcnt64(T) { return 0; } // Is never called
72 #define POPCNT_INTRINSIC(x) __popcnt64(x)
74 #else // Safe fallback for unsupported compilers or when USE_POPCNT is disabled
76 inline bool cpu_has_popcnt() { return false; }
78 #define POPCNT_INTRINSIC(x) 0
80 #endif // cpu_has_popcnt() and POPCNT_INTRINSIC() definitions
83 /// Software implementation of bit count functions
87 inline int count_1s(Bitboard b) {
88 b -= ((b>>1) & 0x5555555555555555ULL);
89 b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
90 b = ((b>>4) + b) & 0x0F0F0F0F0F0F0F0FULL;
91 b *= 0x0101010101010101ULL;
95 inline int count_1s_max_15(Bitboard b) {
96 b -= (b>>1) & 0x5555555555555555ULL;
97 b = ((b>>2) & 0x3333333333333333ULL) + (b & 0x3333333333333333ULL);
98 b *= 0x1111111111111111ULL;
102 #else // if !defined(IS_64BIT)
104 inline int count_1s(Bitboard b) {
105 unsigned w = unsigned(b >> 32), v = unsigned(b);
106 v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
107 w -= (w >> 1) & 0x55555555;
108 v = ((v >> 2) & 0x33333333) + (v & 0x33333333); // 0-4 in 4 bits
109 w = ((w >> 2) & 0x33333333) + (w & 0x33333333);
110 v = ((v >> 4) + v) & 0x0F0F0F0F; // 0-8 in 8 bits
111 v += (((w >> 4) + w) & 0x0F0F0F0F); // 0-16 in 8 bits
112 v *= 0x01010101; // mul is fast on amd procs
116 inline int count_1s_max_15(Bitboard b) {
117 unsigned w = unsigned(b >> 32), v = unsigned(b);
118 v -= (v >> 1) & 0x55555555; // 0-2 in 2 bits
119 w -= (w >> 1) & 0x55555555;
120 v = ((v >> 2) & 0x33333333) + (v & 0x33333333); // 0-4 in 4 bits
121 w = ((w >> 2) & 0x33333333) + (w & 0x33333333);
122 v += w; // 0-8 in 4 bits
130 /// count_1s() counts the number of nonzero bits in a bitboard.
131 /// If template parameter is true an intrinsic is called, otherwise
132 /// we fallback on a software implementation.
134 template<bool UseIntrinsic>
135 inline int count_1s(Bitboard b) {
137 return UseIntrinsic ? POPCNT_INTRINSIC(b) : count_1s(b);
140 template<bool UseIntrinsic>
141 inline int count_1s_max_15(Bitboard b) {
143 return UseIntrinsic ? POPCNT_INTRINSIC(b) : count_1s_max_15(b);
147 // Global constant initialized at startup that is set to true if
148 // CPU on which application runs supports POPCNT intrinsic. Unless
149 // USE_POPCNT is not defined.
150 const bool CpuHasPOPCNT = cpu_has_popcnt();
153 // Global constant used to print info about the use of 64 optimized
154 // functions to verify that a 64 bit compile has been correctly built.
155 #if defined(IS_64BIT)
156 const bool CpuHas64BitPath = true;
158 const bool CpuHas64BitPath = false;
161 #endif // !defined(BITCOUNT_H_INCLUDED)