1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
|
///
/// @file popcnt.hpp
/// @brief Functions to count the number of 1 bits inside
/// a 64-bit variable.
///
/// Copyright (C) 2024 Kim Walisch, <kim.walisch@gmail.com>
///
/// This file is distributed under the BSD License. See the COPYING
/// file in the top level directory.
///
#ifndef POPCNT_HPP
#define POPCNT_HPP
#include <macros.hpp>
#include <stdint.h>
#if defined(ENABLE_MULTIARCH_x86_POPCNT)
#include <cpu_supports_popcnt.hpp>
#endif
// GCC & Clang
#if defined(__GNUC__) || \
__has_builtin(__builtin_popcountl)
// CPUID is only enabled on x86 and x86-64 CPUs
// if the user compiles without -mpopcnt.
#if defined(ENABLE_MULTIARCH_x86_POPCNT)
#if defined(__x86_64__)
namespace {
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
NOINLINE uint64_t popcnt64_bitwise_noinline(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
// On my AMD EPYC 7642 CPU using GCC 12 this runtime
// check incurs an overall overhead of about 1%.
if_likely(cpu_supports_popcnt)
{
__asm__("popcnt %1, %0" : "=r"(x) : "r"(x));
return x;
}
else
return popcnt64_bitwise_noinline(x);
}
} // namespace
#elif defined(__i386__)
namespace {
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
NOINLINE uint64_t popcnt64_bitwise_noinline(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
if_likely(cpu_supports_popcnt)
{
uint32_t x0 = uint32_t(x);
uint32_t x1 = uint32_t(x >> 32);
__asm__("popcnt %1, %0" : "=r"(x0) : "r"(x0));
__asm__("popcnt %1, %0" : "=r"(x1) : "r"(x1));
return x0 + x1;
}
else
return popcnt64_bitwise_noinline(x);
}
} // namespace
#endif // i386
#else // GCC & Clang (no CPUID, not x86)
namespace {
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
#if __cplusplus >= 201703L
if constexpr(sizeof(int) >= sizeof(uint64_t))
return (uint64_t) __builtin_popcount(x);
else if constexpr(sizeof(long) >= sizeof(uint64_t))
return (uint64_t) __builtin_popcountl(x);
else if constexpr(sizeof(long long) >= sizeof(uint64_t))
return (uint64_t) __builtin_popcountll(x);
#else
return (uint64_t) __builtin_popcountll(x);
#endif
}
} // namespace
#endif // GCC & Clang
#elif defined(_MSC_VER) && \
defined(_M_X64) && \
__has_include(<intrin.h>)
#include <intrin.h>
namespace {
#if defined(__POPCNT__) || \
defined(__AVX__)
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
return __popcnt64(x);
}
#elif defined(ENABLE_MULTIARCH_x86_POPCNT)
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
NOINLINE uint64_t popcnt64_bitwise_noinline(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
if_likely(cpu_supports_popcnt)
return __popcnt64(x);
else
return popcnt64_bitwise_noinline(x);
}
#else
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
#endif
} // namespace
#elif defined(_MSC_VER) && \
defined(_M_IX86) && \
__has_include(<intrin.h>)
#include <intrin.h>
namespace {
#if defined(__POPCNT__) || \
defined(__AVX__)
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
return __popcnt(uint32_t(x)) +
__popcnt(uint32_t(x >> 32));
}
#elif defined(ENABLE_MULTIARCH_x86_POPCNT)
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
NOINLINE uint64_t popcnt64_bitwise_noinline(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
if_likely(cpu_supports_popcnt)
return __popcnt(uint32_t(x)) +
__popcnt(uint32_t(x >> 32));
else
return popcnt64_bitwise_noinline(x);
}
#else
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
#endif
} // namespace
#elif __cplusplus >= 202002L && \
__has_include(<bit>)
#include <bit>
namespace {
/// We only use the C++ standard library as a fallback if there
/// are no compiler intrinsics available for POPCNT.
/// Compiler intrinsics often generate faster assembly.
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
return std::popcount(x);
}
} // namespace
#else
namespace {
/// This uses fewer arithmetic operations than any other known
/// implementation on machines with fast multiplication.
/// It uses 12 arithmetic operations, one of which is a multiply.
/// http://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
///
ALWAYS_INLINE uint64_t popcnt64(uint64_t x)
{
uint64_t m1 = 0x5555555555555555ull;
uint64_t m2 = 0x3333333333333333ull;
uint64_t m4 = 0x0F0F0F0F0F0F0F0Full;
uint64_t h01 = 0x0101010101010101ull;
x -= (x >> 1) & m1;
x = (x & m2) + ((x >> 2) & m2);
x = (x + (x >> 4)) & m4;
return (x * h01) >> 56;
}
} // namespace
#endif
#endif // POPCNT_HPP
|
