1 //===-- llvm/Support/MathExtras.h - Useful math functions -------*- C++ -*-===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file contains some functions that are useful for math stuff.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #ifndef LLVM_SUPPORT_MATHEXTRAS_H
14 #define LLVM_SUPPORT_MATHEXTRAS_H
15 
16 #include "llvm/Support/Compiler.h"
17 #include "llvm/Support/SwapByteOrder.h"
18 #include <algorithm>
19 #include <cassert>
20 #include <climits>
21 #include <cstring>
22 #include <limits>
23 #include <type_traits>
24 
25 #ifdef __ANDROID_NDK__
26 #include <android/api-level.h>
27 #endif
28 
29 #ifdef _MSC_VER
30 // Declare these intrinsics manually rather including intrin.h. It's very
31 // expensive, and MathExtras.h is popular.
32 // #include <intrin.h>
33 extern "C" {
34 unsigned char _BitScanForward(unsigned long *_Index, unsigned long _Mask);
35 unsigned char _BitScanForward64(unsigned long *_Index, unsigned __int64 _Mask);
36 unsigned char _BitScanReverse(unsigned long *_Index, unsigned long _Mask);
37 unsigned char _BitScanReverse64(unsigned long *_Index, unsigned __int64 _Mask);
38 }
39 #endif
40 
41 namespace llvm {
42 
43 /// The behavior an operation has on an input of 0.
44 enum ZeroBehavior {
45   /// The returned value is undefined.
46   ZB_Undefined,
47   /// The returned value is numeric_limits<T>::max()
48   ZB_Max,
49   /// The returned value is numeric_limits<T>::digits
50   ZB_Width
51 };
52 
53 /// Mathematical constants.
54 namespace numbers {
55 // TODO: Track C++20 std::numbers.
56 // TODO: Favor using the hexadecimal FP constants (requires C++17).
57 constexpr double e          = 2.7182818284590452354, // (0x1.5bf0a8b145749P+1) https://oeis.org/A001113
58                  egamma     = .57721566490153286061, // (0x1.2788cfc6fb619P-1) https://oeis.org/A001620
59                  ln2        = .69314718055994530942, // (0x1.62e42fefa39efP-1) https://oeis.org/A002162
60                  ln10       = 2.3025850929940456840, // (0x1.24bb1bbb55516P+1) https://oeis.org/A002392
61                  log2e      = 1.4426950408889634074, // (0x1.71547652b82feP+0)
62                  log10e     = .43429448190325182765, // (0x1.bcb7b1526e50eP-2)
63                  pi         = 3.1415926535897932385, // (0x1.921fb54442d18P+1) https://oeis.org/A000796
64                  inv_pi     = .31830988618379067154, // (0x1.45f306bc9c883P-2) https://oeis.org/A049541
65                  sqrtpi     = 1.7724538509055160273, // (0x1.c5bf891b4ef6bP+0) https://oeis.org/A002161
66                  inv_sqrtpi = .56418958354775628695, // (0x1.20dd750429b6dP-1) https://oeis.org/A087197
67                  sqrt2      = 1.4142135623730950488, // (0x1.6a09e667f3bcdP+0) https://oeis.org/A00219
68                  inv_sqrt2  = .70710678118654752440, // (0x1.6a09e667f3bcdP-1)
69                  sqrt3      = 1.7320508075688772935, // (0x1.bb67ae8584caaP+0) https://oeis.org/A002194
70                  inv_sqrt3  = .57735026918962576451, // (0x1.279a74590331cP-1)
71                  phi        = 1.6180339887498948482; // (0x1.9e3779b97f4a8P+0) https://oeis.org/A001622
72 constexpr float ef          = 2.71828183F, // (0x1.5bf0a8P+1) https://oeis.org/A001113
73                 egammaf     = .577215665F, // (0x1.2788d0P-1) https://oeis.org/A001620
74                 ln2f        = .693147181F, // (0x1.62e430P-1) https://oeis.org/A002162
75                 ln10f       = 2.30258509F, // (0x1.26bb1cP+1) https://oeis.org/A002392
76                 log2ef      = 1.44269504F, // (0x1.715476P+0)
77                 log10ef     = .434294482F, // (0x1.bcb7b2P-2)
78                 pif         = 3.14159265F, // (0x1.921fb6P+1) https://oeis.org/A000796
79                 inv_pif     = .318309886F, // (0x1.45f306P-2) https://oeis.org/A049541
80                 sqrtpif     = 1.77245385F, // (0x1.c5bf8aP+0) https://oeis.org/A002161
81                 inv_sqrtpif = .564189584F, // (0x1.20dd76P-1) https://oeis.org/A087197
82                 sqrt2f      = 1.41421356F, // (0x1.6a09e6P+0) https://oeis.org/A002193
83                 inv_sqrt2f  = .707106781F, // (0x1.6a09e6P-1)
84                 sqrt3f      = 1.73205081F, // (0x1.bb67aeP+0) https://oeis.org/A002194
85                 inv_sqrt3f  = .577350269F, // (0x1.279a74P-1)
86                 phif        = 1.61803399F; // (0x1.9e377aP+0) https://oeis.org/A001622
87 } // namespace numbers
88 
89 namespace detail {
90 template <typename T, std::size_t SizeOfT> struct TrailingZerosCounter {
91   static unsigned count(T Val, ZeroBehavior) {
92     if (!Val)
93       return std::numeric_limits<T>::digits;
94     if (Val & 0x1)
95       return 0;
96 
97     // Bisection method.
98     unsigned ZeroBits = 0;
99     T Shift = std::numeric_limits<T>::digits >> 1;
100     T Mask = std::numeric_limits<T>::max() >> Shift;
101     while (Shift) {
102       if ((Val & Mask) == 0) {
103         Val >>= Shift;
104         ZeroBits |= Shift;
105       }
106       Shift >>= 1;
107       Mask >>= Shift;
108     }
109     return ZeroBits;
110   }
111 };
112 
113 #if defined(__GNUC__) || defined(_MSC_VER)
114 template <typename T> struct TrailingZerosCounter<T, 4> {
115   static unsigned count(T Val, ZeroBehavior ZB) {
116     if (ZB != ZB_Undefined && Val == 0)
117       return 32;
118 
119 #if __has_builtin(__builtin_ctz) || defined(__GNUC__)
120     return __builtin_ctz(Val);
121 #elif defined(_MSC_VER)
122     unsigned long Index;
123     _BitScanForward(&Index, Val);
124     return Index;
125 #endif
126   }
127 };
128 
129 #if !defined(_MSC_VER) || defined(_M_X64)
130 template <typename T> struct TrailingZerosCounter<T, 8> {
131   static unsigned count(T Val, ZeroBehavior ZB) {
132     if (ZB != ZB_Undefined && Val == 0)
133       return 64;
134 
135 #if __has_builtin(__builtin_ctzll) || defined(__GNUC__)
136     return __builtin_ctzll(Val);
137 #elif defined(_MSC_VER)
138     unsigned long Index;
139     _BitScanForward64(&Index, Val);
140     return Index;
141 #endif
142   }
143 };
144 #endif
145 #endif
146 } // namespace detail
147 
148 /// Count number of 0's from the least significant bit to the most
149 ///   stopping at the first 1.
150 ///
151 /// Only unsigned integral types are allowed.
152 ///
153 /// \param ZB the behavior on an input of 0. Only ZB_Width and ZB_Undefined are
154 ///   valid arguments.
155 template <typename T>
156 unsigned countTrailingZeros(T Val, ZeroBehavior ZB = ZB_Width) {
157   static_assert(std::numeric_limits<T>::is_integer &&
158                     !std::numeric_limits<T>::is_signed,
159                 "Only unsigned integral types are allowed.");
160   return llvm::detail::TrailingZerosCounter<T, sizeof(T)>::count(Val, ZB);
161 }
162 
163 namespace detail {
164 template <typename T, std::size_t SizeOfT> struct LeadingZerosCounter {
165   static unsigned count(T Val, ZeroBehavior) {
166     if (!Val)
167       return std::numeric_limits<T>::digits;
168 
169     // Bisection method.
170     unsigned ZeroBits = 0;
171     for (T Shift = std::numeric_limits<T>::digits >> 1; Shift; Shift >>= 1) {
172       T Tmp = Val >> Shift;
173       if (Tmp)
174         Val = Tmp;
175       else
176         ZeroBits |= Shift;
177     }
178     return ZeroBits;
179   }
180 };
181 
182 #if defined(__GNUC__) || defined(_MSC_VER)
183 template <typename T> struct LeadingZerosCounter<T, 4> {
184   static unsigned count(T Val, ZeroBehavior ZB) {
185     if (ZB != ZB_Undefined && Val == 0)
186       return 32;
187 
188 #if __has_builtin(__builtin_clz) || defined(__GNUC__)
189     return __builtin_clz(Val);
190 #elif defined(_MSC_VER)
191     unsigned long Index;
192     _BitScanReverse(&Index, Val);
193     return Index ^ 31;
194 #endif
195   }
196 };
197 
198 #if !defined(_MSC_VER) || defined(_M_X64)
199 template <typename T> struct LeadingZerosCounter<T, 8> {
200   static unsigned count(T Val, ZeroBehavior ZB) {
201     if (ZB != ZB_Undefined && Val == 0)
202       return 64;
203 
204 #if __has_builtin(__builtin_clzll) || defined(__GNUC__)
205     return __builtin_clzll(Val);
206 #elif defined(_MSC_VER)
207     unsigned long Index;
208     _BitScanReverse64(&Index, Val);
209     return Index ^ 63;
210 #endif
211   }
212 };
213 #endif
214 #endif
215 } // namespace detail
216 
217 /// Count number of 0's from the most significant bit to the least
218 ///   stopping at the first 1.
219 ///
220 /// Only unsigned integral types are allowed.
221 ///
222 /// \param ZB the behavior on an input of 0. Only ZB_Width and ZB_Undefined are
223 ///   valid arguments.
224 template <typename T>
225 unsigned countLeadingZeros(T Val, ZeroBehavior ZB = ZB_Width) {
226   static_assert(std::numeric_limits<T>::is_integer &&
227                     !std::numeric_limits<T>::is_signed,
228                 "Only unsigned integral types are allowed.");
229   return llvm::detail::LeadingZerosCounter<T, sizeof(T)>::count(Val, ZB);
230 }
231 
232 /// Get the index of the first set bit starting from the least
233 ///   significant bit.
234 ///
235 /// Only unsigned integral types are allowed.
236 ///
237 /// \param ZB the behavior on an input of 0. Only ZB_Max and ZB_Undefined are
238 ///   valid arguments.
239 template <typename T> T findFirstSet(T Val, ZeroBehavior ZB = ZB_Max) {
240   if (ZB == ZB_Max && Val == 0)
241     return std::numeric_limits<T>::max();
242 
243   return countTrailingZeros(Val, ZB_Undefined);
244 }
245 
246 /// Create a bitmask with the N right-most bits set to 1, and all other
247 /// bits set to 0.  Only unsigned types are allowed.
248 template <typename T> T maskTrailingOnes(unsigned N) {
249   static_assert(std::is_unsigned<T>::value, "Invalid type!");
250   const unsigned Bits = CHAR_BIT * sizeof(T);
251   assert(N <= Bits && "Invalid bit index");
252   return N == 0 ? 0 : (T(-1) >> (Bits - N));
253 }
254 
255 /// Create a bitmask with the N left-most bits set to 1, and all other
256 /// bits set to 0.  Only unsigned types are allowed.
257 template <typename T> T maskLeadingOnes(unsigned N) {
258   return ~maskTrailingOnes<T>(CHAR_BIT * sizeof(T) - N);
259 }
260 
261 /// Create a bitmask with the N right-most bits set to 0, and all other
262 /// bits set to 1.  Only unsigned types are allowed.
263 template <typename T> T maskTrailingZeros(unsigned N) {
264   return maskLeadingOnes<T>(CHAR_BIT * sizeof(T) - N);
265 }
266 
267 /// Create a bitmask with the N left-most bits set to 0, and all other
268 /// bits set to 1.  Only unsigned types are allowed.
269 template <typename T> T maskLeadingZeros(unsigned N) {
270   return maskTrailingOnes<T>(CHAR_BIT * sizeof(T) - N);
271 }
272 
273 /// Get the index of the last set bit starting from the least
274 ///   significant bit.
275 ///
276 /// Only unsigned integral types are allowed.
277 ///
278 /// \param ZB the behavior on an input of 0. Only ZB_Max and ZB_Undefined are
279 ///   valid arguments.
280 template <typename T> T findLastSet(T Val, ZeroBehavior ZB = ZB_Max) {
281   if (ZB == ZB_Max && Val == 0)
282     return std::numeric_limits<T>::max();
283 
284   // Use ^ instead of - because both gcc and llvm can remove the associated ^
285   // in the __builtin_clz intrinsic on x86.
286   return countLeadingZeros(Val, ZB_Undefined) ^
287          (std::numeric_limits<T>::digits - 1);
288 }
289 
290 /// Macro compressed bit reversal table for 256 bits.
291 ///
292 /// http://graphics.stanford.edu/~seander/bithacks.html#BitReverseTable
293 static const unsigned char BitReverseTable256[256] = {
294 #define R2(n) n, n + 2 * 64, n + 1 * 64, n + 3 * 64
295 #define R4(n) R2(n), R2(n + 2 * 16), R2(n + 1 * 16), R2(n + 3 * 16)
296 #define R6(n) R4(n), R4(n + 2 * 4), R4(n + 1 * 4), R4(n + 3 * 4)
297   R6(0), R6(2), R6(1), R6(3)
298 #undef R2
299 #undef R4
300 #undef R6
301 };
302 
303 /// Reverse the bits in \p Val.
304 template <typename T>
305 T reverseBits(T Val) {
306   unsigned char in[sizeof(Val)];
307   unsigned char out[sizeof(Val)];
308   std::memcpy(in, &Val, sizeof(Val));
309   for (unsigned i = 0; i < sizeof(Val); ++i)
310     out[(sizeof(Val) - i) - 1] = BitReverseTable256[in[i]];
311   std::memcpy(&Val, out, sizeof(Val));
312   return Val;
313 }
314 
315 // NOTE: The following support functions use the _32/_64 extensions instead of
316 // type overloading so that signed and unsigned integers can be used without
317 // ambiguity.
318 
319 /// Return the high 32 bits of a 64 bit value.
320 constexpr inline uint32_t Hi_32(uint64_t Value) {
321   return static_cast<uint32_t>(Value >> 32);
322 }
323 
324 /// Return the low 32 bits of a 64 bit value.
325 constexpr inline uint32_t Lo_32(uint64_t Value) {
326   return static_cast<uint32_t>(Value);
327 }
328 
329 /// Make a 64-bit integer from a high / low pair of 32-bit integers.
330 constexpr inline uint64_t Make_64(uint32_t High, uint32_t Low) {
331   return ((uint64_t)High << 32) | (uint64_t)Low;
332 }
333 
334 /// Checks if an integer fits into the given bit width.
335 template <unsigned N> constexpr inline bool isInt(int64_t x) {
336   return N >= 64 || (-(INT64_C(1)<<(N-1)) <= x && x < (INT64_C(1)<<(N-1)));
337 }
338 // Template specializations to get better code for common cases.
339 template <> constexpr inline bool isInt<8>(int64_t x) {
340   return static_cast<int8_t>(x) == x;
341 }
342 template <> constexpr inline bool isInt<16>(int64_t x) {
343   return static_cast<int16_t>(x) == x;
344 }
345 template <> constexpr inline bool isInt<32>(int64_t x) {
346   return static_cast<int32_t>(x) == x;
347 }
348 
349 /// Checks if a signed integer is an N bit number shifted left by S.
350 template <unsigned N, unsigned S>
351 constexpr inline bool isShiftedInt(int64_t x) {
352   static_assert(
353       N > 0, "isShiftedInt<0> doesn't make sense (refers to a 0-bit number.");
354   static_assert(N + S <= 64, "isShiftedInt<N, S> with N + S > 64 is too wide.");
355   return isInt<N + S>(x) && (x % (UINT64_C(1) << S) == 0);
356 }
357 
358 /// Checks if an unsigned integer fits into the given bit width.
359 ///
360 /// This is written as two functions rather than as simply
361 ///
362 ///   return N >= 64 || X < (UINT64_C(1) << N);
363 ///
364 /// to keep MSVC from (incorrectly) warning on isUInt<64> that we're shifting
365 /// left too many places.
366 template <unsigned N>
367 constexpr inline typename std::enable_if<(N < 64), bool>::type
368 isUInt(uint64_t X) {
369   static_assert(N > 0, "isUInt<0> doesn't make sense");
370   return X < (UINT64_C(1) << (N));
371 }
372 template <unsigned N>
373 constexpr inline typename std::enable_if<N >= 64, bool>::type
374 isUInt(uint64_t X) {
375   return true;
376 }
377 
378 // Template specializations to get better code for common cases.
379 template <> constexpr inline bool isUInt<8>(uint64_t x) {
380   return static_cast<uint8_t>(x) == x;
381 }
382 template <> constexpr inline bool isUInt<16>(uint64_t x) {
383   return static_cast<uint16_t>(x) == x;
384 }
385 template <> constexpr inline bool isUInt<32>(uint64_t x) {
386   return static_cast<uint32_t>(x) == x;
387 }
388 
389 /// Checks if a unsigned integer is an N bit number shifted left by S.
390 template <unsigned N, unsigned S>
391 constexpr inline bool isShiftedUInt(uint64_t x) {
392   static_assert(
393       N > 0, "isShiftedUInt<0> doesn't make sense (refers to a 0-bit number)");
394   static_assert(N + S <= 64,
395                 "isShiftedUInt<N, S> with N + S > 64 is too wide.");
396   // Per the two static_asserts above, S must be strictly less than 64.  So
397   // 1 << S is not undefined behavior.
398   return isUInt<N + S>(x) && (x % (UINT64_C(1) << S) == 0);
399 }
400 
401 /// Gets the maximum value for a N-bit unsigned integer.
402 inline uint64_t maxUIntN(uint64_t N) {
403   assert(N > 0 && N <= 64 && "integer width out of range");
404 
405   // uint64_t(1) << 64 is undefined behavior, so we can't do
406   //   (uint64_t(1) << N) - 1
407   // without checking first that N != 64.  But this works and doesn't have a
408   // branch.
409   return UINT64_MAX >> (64 - N);
410 }
411 
412 /// Gets the minimum value for a N-bit signed integer.
413 inline int64_t minIntN(int64_t N) {
414   assert(N > 0 && N <= 64 && "integer width out of range");
415 
416   return -(UINT64_C(1)<<(N-1));
417 }
418 
419 /// Gets the maximum value for a N-bit signed integer.
420 inline int64_t maxIntN(int64_t N) {
421   assert(N > 0 && N <= 64 && "integer width out of range");
422 
423   // This relies on two's complement wraparound when N == 64, so we convert to
424   // int64_t only at the very end to avoid UB.
425   return (UINT64_C(1) << (N - 1)) - 1;
426 }
427 
428 /// Checks if an unsigned integer fits into the given (dynamic) bit width.
429 inline bool isUIntN(unsigned N, uint64_t x) {
430   return N >= 64 || x <= maxUIntN(N);
431 }
432 
433 /// Checks if an signed integer fits into the given (dynamic) bit width.
434 inline bool isIntN(unsigned N, int64_t x) {
435   return N >= 64 || (minIntN(N) <= x && x <= maxIntN(N));
436 }
437 
438 /// Return true if the argument is a non-empty sequence of ones starting at the
439 /// least significant bit with the remainder zero (32 bit version).
440 /// Ex. isMask_32(0x0000FFFFU) == true.
441 constexpr inline bool isMask_32(uint32_t Value) {
442   return Value && ((Value + 1) & Value) == 0;
443 }
444 
445 /// Return true if the argument is a non-empty sequence of ones starting at the
446 /// least significant bit with the remainder zero (64 bit version).
447 constexpr inline bool isMask_64(uint64_t Value) {
448   return Value && ((Value + 1) & Value) == 0;
449 }
450 
451 /// Return true if the argument contains a non-empty sequence of ones with the
452 /// remainder zero (32 bit version.) Ex. isShiftedMask_32(0x0000FF00U) == true.
453 constexpr inline bool isShiftedMask_32(uint32_t Value) {
454   return Value && isMask_32((Value - 1) | Value);
455 }
456 
457 /// Return true if the argument contains a non-empty sequence of ones with the
458 /// remainder zero (64 bit version.)
459 constexpr inline bool isShiftedMask_64(uint64_t Value) {
460   return Value && isMask_64((Value - 1) | Value);
461 }
462 
463 /// Return true if the argument is a power of two > 0.
464 /// Ex. isPowerOf2_32(0x00100000U) == true (32 bit edition.)
465 constexpr inline bool isPowerOf2_32(uint32_t Value) {
466   return Value && !(Value & (Value - 1));
467 }
468 
469 /// Return true if the argument is a power of two > 0 (64 bit edition.)
470 constexpr inline bool isPowerOf2_64(uint64_t Value) {
471   return Value && !(Value & (Value - 1));
472 }
473 
474 /// Return a byte-swapped representation of the 16-bit argument.
475 inline uint16_t ByteSwap_16(uint16_t Value) {
476   return sys::SwapByteOrder_16(Value);
477 }
478 
479 /// Return a byte-swapped representation of the 32-bit argument.
480 inline uint32_t ByteSwap_32(uint32_t Value) {
481   return sys::SwapByteOrder_32(Value);
482 }
483 
484 /// Return a byte-swapped representation of the 64-bit argument.
485 inline uint64_t ByteSwap_64(uint64_t Value) {
486   return sys::SwapByteOrder_64(Value);
487 }
488 
489 /// Count the number of ones from the most significant bit to the first
490 /// zero bit.
491 ///
492 /// Ex. countLeadingOnes(0xFF0FFF00) == 8.
493 /// Only unsigned integral types are allowed.
494 ///
495 /// \param ZB the behavior on an input of all ones. Only ZB_Width and
496 /// ZB_Undefined are valid arguments.
497 template <typename T>
498 unsigned countLeadingOnes(T Value, ZeroBehavior ZB = ZB_Width) {
499   static_assert(std::numeric_limits<T>::is_integer &&
500                     !std::numeric_limits<T>::is_signed,
501                 "Only unsigned integral types are allowed.");
502   return countLeadingZeros<T>(~Value, ZB);
503 }
504 
505 /// Count the number of ones from the least significant bit to the first
506 /// zero bit.
507 ///
508 /// Ex. countTrailingOnes(0x00FF00FF) == 8.
509 /// Only unsigned integral types are allowed.
510 ///
511 /// \param ZB the behavior on an input of all ones. Only ZB_Width and
512 /// ZB_Undefined are valid arguments.
513 template <typename T>
514 unsigned countTrailingOnes(T Value, ZeroBehavior ZB = ZB_Width) {
515   static_assert(std::numeric_limits<T>::is_integer &&
516                     !std::numeric_limits<T>::is_signed,
517                 "Only unsigned integral types are allowed.");
518   return countTrailingZeros<T>(~Value, ZB);
519 }
520 
521 namespace detail {
522 template <typename T, std::size_t SizeOfT> struct PopulationCounter {
523   static unsigned count(T Value) {
524     // Generic version, forward to 32 bits.
525     static_assert(SizeOfT <= 4, "Not implemented!");
526 #if defined(__GNUC__)
527     return __builtin_popcount(Value);
528 #else
529     uint32_t v = Value;
530     v = v - ((v >> 1) & 0x55555555);
531     v = (v & 0x33333333) + ((v >> 2) & 0x33333333);
532     return ((v + (v >> 4) & 0xF0F0F0F) * 0x1010101) >> 24;
533 #endif
534   }
535 };
536 
537 template <typename T> struct PopulationCounter<T, 8> {
538   static unsigned count(T Value) {
539 #if defined(__GNUC__)
540     return __builtin_popcountll(Value);
541 #else
542     uint64_t v = Value;
543     v = v - ((v >> 1) & 0x5555555555555555ULL);
544     v = (v & 0x3333333333333333ULL) + ((v >> 2) & 0x3333333333333333ULL);
545     v = (v + (v >> 4)) & 0x0F0F0F0F0F0F0F0FULL;
546     return unsigned((uint64_t)(v * 0x0101010101010101ULL) >> 56);
547 #endif
548   }
549 };
550 } // namespace detail
551 
552 /// Count the number of set bits in a value.
553 /// Ex. countPopulation(0xF000F000) = 8
554 /// Returns 0 if the word is zero.
555 template <typename T>
556 inline unsigned countPopulation(T Value) {
557   static_assert(std::numeric_limits<T>::is_integer &&
558                     !std::numeric_limits<T>::is_signed,
559                 "Only unsigned integral types are allowed.");
560   return detail::PopulationCounter<T, sizeof(T)>::count(Value);
561 }
562 
563 /// Compile time Log2.
564 /// Valid only for positive powers of two.
565 template <size_t kValue> constexpr inline size_t CTLog2() {
566   static_assert(kValue > 0 && llvm::isPowerOf2_64(kValue),
567                 "Value is not a valid power of 2");
568   return 1 + CTLog2<kValue / 2>();
569 }
570 
571 template <> constexpr inline size_t CTLog2<1>() { return 0; }
572 
573 /// Return the log base 2 of the specified value.
574 inline double Log2(double Value) {
575 #if defined(__ANDROID_API__) && __ANDROID_API__ < 18
576   return __builtin_log(Value) / __builtin_log(2.0);
577 #else
578   return log2(Value);
579 #endif
580 }
581 
582 /// Return the floor log base 2 of the specified value, -1 if the value is zero.
583 /// (32 bit edition.)
584 /// Ex. Log2_32(32) == 5, Log2_32(1) == 0, Log2_32(0) == -1, Log2_32(6) == 2
585 inline unsigned Log2_32(uint32_t Value) {
586   return 31 - countLeadingZeros(Value);
587 }
588 
589 /// Return the floor log base 2 of the specified value, -1 if the value is zero.
590 /// (64 bit edition.)
591 inline unsigned Log2_64(uint64_t Value) {
592   return 63 - countLeadingZeros(Value);
593 }
594 
595 /// Return the ceil log base 2 of the specified value, 32 if the value is zero.
596 /// (32 bit edition).
597 /// Ex. Log2_32_Ceil(32) == 5, Log2_32_Ceil(1) == 0, Log2_32_Ceil(6) == 3
598 inline unsigned Log2_32_Ceil(uint32_t Value) {
599   return 32 - countLeadingZeros(Value - 1);
600 }
601 
602 /// Return the ceil log base 2 of the specified value, 64 if the value is zero.
603 /// (64 bit edition.)
604 inline unsigned Log2_64_Ceil(uint64_t Value) {
605   return 64 - countLeadingZeros(Value - 1);
606 }
607 
608 /// Return the greatest common divisor of the values using Euclid's algorithm.
609 template <typename T>
610 inline T greatestCommonDivisor(T A, T B) {
611   while (B) {
612     T Tmp = B;
613     B = A % B;
614     A = Tmp;
615   }
616   return A;
617 }
618 
619 inline uint64_t GreatestCommonDivisor64(uint64_t A, uint64_t B) {
620   return greatestCommonDivisor<uint64_t>(A, B);
621 }
622 
623 /// This function takes a 64-bit integer and returns the bit equivalent double.
624 inline double BitsToDouble(uint64_t Bits) {
625   double D;
626   static_assert(sizeof(uint64_t) == sizeof(double), "Unexpected type sizes");
627   memcpy(&D, &Bits, sizeof(Bits));
628   return D;
629 }
630 
631 /// This function takes a 32-bit integer and returns the bit equivalent float.
632 inline float BitsToFloat(uint32_t Bits) {
633   float F;
634   static_assert(sizeof(uint32_t) == sizeof(float), "Unexpected type sizes");
635   memcpy(&F, &Bits, sizeof(Bits));
636   return F;
637 }
638 
639 /// This function takes a double and returns the bit equivalent 64-bit integer.
640 /// Note that copying doubles around changes the bits of NaNs on some hosts,
641 /// notably x86, so this routine cannot be used if these bits are needed.
642 inline uint64_t DoubleToBits(double Double) {
643   uint64_t Bits;
644   static_assert(sizeof(uint64_t) == sizeof(double), "Unexpected type sizes");
645   memcpy(&Bits, &Double, sizeof(Double));
646   return Bits;
647 }
648 
649 /// This function takes a float and returns the bit equivalent 32-bit integer.
650 /// Note that copying floats around changes the bits of NaNs on some hosts,
651 /// notably x86, so this routine cannot be used if these bits are needed.
652 inline uint32_t FloatToBits(float Float) {
653   uint32_t Bits;
654   static_assert(sizeof(uint32_t) == sizeof(float), "Unexpected type sizes");
655   memcpy(&Bits, &Float, sizeof(Float));
656   return Bits;
657 }
658 
659 /// A and B are either alignments or offsets. Return the minimum alignment that
660 /// may be assumed after adding the two together.
661 constexpr inline uint64_t MinAlign(uint64_t A, uint64_t B) {
662   // The largest power of 2 that divides both A and B.
663   //
664   // Replace "-Value" by "1+~Value" in the following commented code to avoid
665   // MSVC warning C4146
666   //    return (A | B) & -(A | B);
667   return (A | B) & (1 + ~(A | B));
668 }
669 
670 /// Returns the next power of two (in 64-bits) that is strictly greater than A.
671 /// Returns zero on overflow.
672 inline uint64_t NextPowerOf2(uint64_t A) {
673   A |= (A >> 1);
674   A |= (A >> 2);
675   A |= (A >> 4);
676   A |= (A >> 8);
677   A |= (A >> 16);
678   A |= (A >> 32);
679   return A + 1;
680 }
681 
682 /// Returns the power of two which is less than or equal to the given value.
683 /// Essentially, it is a floor operation across the domain of powers of two.
684 inline uint64_t PowerOf2Floor(uint64_t A) {
685   if (!A) return 0;
686   return 1ull << (63 - countLeadingZeros(A, ZB_Undefined));
687 }
688 
689 /// Returns the power of two which is greater than or equal to the given value.
690 /// Essentially, it is a ceil operation across the domain of powers of two.
691 inline uint64_t PowerOf2Ceil(uint64_t A) {
692   if (!A)
693     return 0;
694   return NextPowerOf2(A - 1);
695 }
696 
697 /// Returns the next integer (mod 2**64) that is greater than or equal to
698 /// \p Value and is a multiple of \p Align. \p Align must be non-zero.
699 ///
700 /// If non-zero \p Skew is specified, the return value will be a minimal
701 /// integer that is greater than or equal to \p Value and equal to
702 /// \p Align * N + \p Skew for some integer N. If \p Skew is larger than
703 /// \p Align, its value is adjusted to '\p Skew mod \p Align'.
704 ///
705 /// Examples:
706 /// \code
707 ///   alignTo(5, 8) = 8
708 ///   alignTo(17, 8) = 24
709 ///   alignTo(~0LL, 8) = 0
710 ///   alignTo(321, 255) = 510
711 ///
712 ///   alignTo(5, 8, 7) = 7
713 ///   alignTo(17, 8, 1) = 17
714 ///   alignTo(~0LL, 8, 3) = 3
715 ///   alignTo(321, 255, 42) = 552
716 /// \endcode
717 inline uint64_t alignTo(uint64_t Value, uint64_t Align, uint64_t Skew = 0) {
718   assert(Align != 0u && "Align can't be 0.");
719   Skew %= Align;
720   return (Value + Align - 1 - Skew) / Align * Align + Skew;
721 }
722 
723 /// Returns the next integer (mod 2**64) that is greater than or equal to
724 /// \p Value and is a multiple of \c Align. \c Align must be non-zero.
725 template <uint64_t Align> constexpr inline uint64_t alignTo(uint64_t Value) {
726   static_assert(Align != 0u, "Align must be non-zero");
727   return (Value + Align - 1) / Align * Align;
728 }
729 
730 /// Returns the integer ceil(Numerator / Denominator).
731 inline uint64_t divideCeil(uint64_t Numerator, uint64_t Denominator) {
732   return alignTo(Numerator, Denominator) / Denominator;
733 }
734 
735 /// Returns the integer nearest(Numerator / Denominator).
736 inline uint64_t divideNearest(uint64_t Numerator, uint64_t Denominator) {
737   return (Numerator + (Denominator / 2)) / Denominator;
738 }
739 
740 /// Returns the largest uint64_t less than or equal to \p Value and is
741 /// \p Skew mod \p Align. \p Align must be non-zero
742 inline uint64_t alignDown(uint64_t Value, uint64_t Align, uint64_t Skew = 0) {
743   assert(Align != 0u && "Align can't be 0.");
744   Skew %= Align;
745   return (Value - Skew) / Align * Align + Skew;
746 }
747 
748 /// Sign-extend the number in the bottom B bits of X to a 32-bit integer.
749 /// Requires 0 < B <= 32.
750 template <unsigned B> constexpr inline int32_t SignExtend32(uint32_t X) {
751   static_assert(B > 0, "Bit width can't be 0.");
752   static_assert(B <= 32, "Bit width out of range.");
753   return int32_t(X << (32 - B)) >> (32 - B);
754 }
755 
756 /// Sign-extend the number in the bottom B bits of X to a 32-bit integer.
757 /// Requires 0 < B < 32.
758 inline int32_t SignExtend32(uint32_t X, unsigned B) {
759   assert(B > 0 && "Bit width can't be 0.");
760   assert(B <= 32 && "Bit width out of range.");
761   return int32_t(X << (32 - B)) >> (32 - B);
762 }
763 
764 /// Sign-extend the number in the bottom B bits of X to a 64-bit integer.
765 /// Requires 0 < B < 64.
766 template <unsigned B> constexpr inline int64_t SignExtend64(uint64_t x) {
767   static_assert(B > 0, "Bit width can't be 0.");
768   static_assert(B <= 64, "Bit width out of range.");
769   return int64_t(x << (64 - B)) >> (64 - B);
770 }
771 
772 /// Sign-extend the number in the bottom B bits of X to a 64-bit integer.
773 /// Requires 0 < B < 64.
774 inline int64_t SignExtend64(uint64_t X, unsigned B) {
775   assert(B > 0 && "Bit width can't be 0.");
776   assert(B <= 64 && "Bit width out of range.");
777   return int64_t(X << (64 - B)) >> (64 - B);
778 }
779 
780 /// Subtract two unsigned integers, X and Y, of type T and return the absolute
781 /// value of the result.
782 template <typename T>
783 typename std::enable_if<std::is_unsigned<T>::value, T>::type
784 AbsoluteDifference(T X, T Y) {
785   return std::max(X, Y) - std::min(X, Y);
786 }
787 
788 /// Add two unsigned integers, X and Y, of type T.  Clamp the result to the
789 /// maximum representable value of T on overflow.  ResultOverflowed indicates if
790 /// the result is larger than the maximum representable value of type T.
791 template <typename T>
792 typename std::enable_if<std::is_unsigned<T>::value, T>::type
793 SaturatingAdd(T X, T Y, bool *ResultOverflowed = nullptr) {
794   bool Dummy;
795   bool &Overflowed = ResultOverflowed ? *ResultOverflowed : Dummy;
796   // Hacker's Delight, p. 29
797   T Z = X + Y;
798   Overflowed = (Z < X || Z < Y);
799   if (Overflowed)
800     return std::numeric_limits<T>::max();
801   else
802     return Z;
803 }
804 
805 /// Multiply two unsigned integers, X and Y, of type T.  Clamp the result to the
806 /// maximum representable value of T on overflow.  ResultOverflowed indicates if
807 /// the result is larger than the maximum representable value of type T.
808 template <typename T>
809 typename std::enable_if<std::is_unsigned<T>::value, T>::type
810 SaturatingMultiply(T X, T Y, bool *ResultOverflowed = nullptr) {
811   bool Dummy;
812   bool &Overflowed = ResultOverflowed ? *ResultOverflowed : Dummy;
813 
814   // Hacker's Delight, p. 30 has a different algorithm, but we don't use that
815   // because it fails for uint16_t (where multiplication can have undefined
816   // behavior due to promotion to int), and requires a division in addition
817   // to the multiplication.
818 
819   Overflowed = false;
820 
821   // Log2(Z) would be either Log2Z or Log2Z + 1.
822   // Special case: if X or Y is 0, Log2_64 gives -1, and Log2Z
823   // will necessarily be less than Log2Max as desired.
824   int Log2Z = Log2_64(X) + Log2_64(Y);
825   const T Max = std::numeric_limits<T>::max();
826   int Log2Max = Log2_64(Max);
827   if (Log2Z < Log2Max) {
828     return X * Y;
829   }
830   if (Log2Z > Log2Max) {
831     Overflowed = true;
832     return Max;
833   }
834 
835   // We're going to use the top bit, and maybe overflow one
836   // bit past it. Multiply all but the bottom bit then add
837   // that on at the end.
838   T Z = (X >> 1) * Y;
839   if (Z & ~(Max >> 1)) {
840     Overflowed = true;
841     return Max;
842   }
843   Z <<= 1;
844   if (X & 1)
845     return SaturatingAdd(Z, Y, ResultOverflowed);
846 
847   return Z;
848 }
849 
850 /// Multiply two unsigned integers, X and Y, and add the unsigned integer, A to
851 /// the product. Clamp the result to the maximum representable value of T on
852 /// overflow. ResultOverflowed indicates if the result is larger than the
853 /// maximum representable value of type T.
854 template <typename T>
855 typename std::enable_if<std::is_unsigned<T>::value, T>::type
856 SaturatingMultiplyAdd(T X, T Y, T A, bool *ResultOverflowed = nullptr) {
857   bool Dummy;
858   bool &Overflowed = ResultOverflowed ? *ResultOverflowed : Dummy;
859 
860   T Product = SaturatingMultiply(X, Y, &Overflowed);
861   if (Overflowed)
862     return Product;
863 
864   return SaturatingAdd(A, Product, &Overflowed);
865 }
866 
867 /// Use this rather than HUGE_VALF; the latter causes warnings on MSVC.
868 extern const float huge_valf;
869 
870 
871 /// Add two signed integers, computing the two's complement truncated result,
872 /// returning true if overflow occured.
873 template <typename T>
874 typename std::enable_if<std::is_signed<T>::value, T>::type
875 AddOverflow(T X, T Y, T &Result) {
876 #if __has_builtin(__builtin_add_overflow)
877   return __builtin_add_overflow(X, Y, &Result);
878 #else
879   // Perform the unsigned addition.
880   using U = typename std::make_unsigned<T>::type;
881   const U UX = static_cast<U>(X);
882   const U UY = static_cast<U>(Y);
883   const U UResult = UX + UY;
884 
885   // Convert to signed.
886   Result = static_cast<T>(UResult);
887 
888   // Adding two positive numbers should result in a positive number.
889   if (X > 0 && Y > 0)
890     return Result <= 0;
891   // Adding two negatives should result in a negative number.
892   if (X < 0 && Y < 0)
893     return Result >= 0;
894   return false;
895 #endif
896 }
897 
898 /// Subtract two signed integers, computing the two's complement truncated
899 /// result, returning true if an overflow ocurred.
900 template <typename T>
901 typename std::enable_if<std::is_signed<T>::value, T>::type
902 SubOverflow(T X, T Y, T &Result) {
903 #if __has_builtin(__builtin_sub_overflow)
904   return __builtin_sub_overflow(X, Y, &Result);
905 #else
906   // Perform the unsigned addition.
907   using U = typename std::make_unsigned<T>::type;
908   const U UX = static_cast<U>(X);
909   const U UY = static_cast<U>(Y);
910   const U UResult = UX - UY;
911 
912   // Convert to signed.
913   Result = static_cast<T>(UResult);
914 
915   // Subtracting a positive number from a negative results in a negative number.
916   if (X <= 0 && Y > 0)
917     return Result >= 0;
918   // Subtracting a negative number from a positive results in a positive number.
919   if (X >= 0 && Y < 0)
920     return Result <= 0;
921   return false;
922 #endif
923 }
924 
925 
926 /// Multiply two signed integers, computing the two's complement truncated
927 /// result, returning true if an overflow ocurred.
928 template <typename T>
929 typename std::enable_if<std::is_signed<T>::value, T>::type
930 MulOverflow(T X, T Y, T &Result) {
931   // Perform the unsigned multiplication on absolute values.
932   using U = typename std::make_unsigned<T>::type;
933   const U UX = X < 0 ? (0 - static_cast<U>(X)) : static_cast<U>(X);
934   const U UY = Y < 0 ? (0 - static_cast<U>(Y)) : static_cast<U>(Y);
935   const U UResult = UX * UY;
936 
937   // Convert to signed.
938   const bool IsNegative = (X < 0) ^ (Y < 0);
939   Result = IsNegative ? (0 - UResult) : UResult;
940 
941   // If any of the args was 0, result is 0 and no overflow occurs.
942   if (UX == 0 || UY == 0)
943     return false;
944 
945   // UX and UY are in [1, 2^n], where n is the number of digits.
946   // Check how the max allowed absolute value (2^n for negative, 2^(n-1) for
947   // positive) divided by an argument compares to the other.
948   if (IsNegative)
949     return UX > (static_cast<U>(std::numeric_limits<T>::max()) + U(1)) / UY;
950   else
951     return UX > (static_cast<U>(std::numeric_limits<T>::max())) / UY;
952 }
953 
954 } // End llvm namespace
955 
956 #endif
957