1 //===-- llvm/Support/MathExtras.h - Useful math functions -------*- C++ -*-===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file contains some functions that are useful for math stuff.
11 //
12 //===----------------------------------------------------------------------===//
13
14 #ifndef LLVM_SUPPORT_MATHEXTRAS_H
15 #define LLVM_SUPPORT_MATHEXTRAS_H
16
17 #include "llvm/Support/SwapByteOrder.h"
18
19 #ifdef _MSC_VER
20 # include <intrin.h>
21 #endif
22
23 namespace llvm {
24
25 // NOTE: The following support functions use the _32/_64 extensions instead of
26 // type overloading so that signed and unsigned integers can be used without
27 // ambiguity.
28
29 /// Hi_32 - This function returns the high 32 bits of a 64 bit value.
Hi_32(uint64_t Value)30 inline uint32_t Hi_32(uint64_t Value) {
31 return static_cast<uint32_t>(Value >> 32);
32 }
33
34 /// Lo_32 - This function returns the low 32 bits of a 64 bit value.
Lo_32(uint64_t Value)35 inline uint32_t Lo_32(uint64_t Value) {
36 return static_cast<uint32_t>(Value);
37 }
38
39 /// isInt - Checks if an integer fits into the given bit width.
40 template<unsigned N>
isInt(int64_t x)41 inline bool isInt(int64_t x) {
42 return N >= 64 || (-(INT64_C(1)<<(N-1)) <= x && x < (INT64_C(1)<<(N-1)));
43 }
44 // Template specializations to get better code for common cases.
45 template<>
46 inline bool isInt<8>(int64_t x) {
47 return static_cast<int8_t>(x) == x;
48 }
49 template<>
50 inline bool isInt<16>(int64_t x) {
51 return static_cast<int16_t>(x) == x;
52 }
53 template<>
54 inline bool isInt<32>(int64_t x) {
55 return static_cast<int32_t>(x) == x;
56 }
57
58 /// isShiftedInt<N,S> - Checks if a signed integer is an N bit number shifted
59 /// left by S.
60 template<unsigned N, unsigned S>
isShiftedInt(int64_t x)61 inline bool isShiftedInt(int64_t x) {
62 return isInt<N+S>(x) && (x % (1<<S) == 0);
63 }
64
65 /// isUInt - Checks if an unsigned integer fits into the given bit width.
66 template<unsigned N>
isUInt(uint64_t x)67 inline bool isUInt(uint64_t x) {
68 return N >= 64 || x < (UINT64_C(1)<<(N));
69 }
70 // Template specializations to get better code for common cases.
71 template<>
72 inline bool isUInt<8>(uint64_t x) {
73 return static_cast<uint8_t>(x) == x;
74 }
75 template<>
76 inline bool isUInt<16>(uint64_t x) {
77 return static_cast<uint16_t>(x) == x;
78 }
79 template<>
80 inline bool isUInt<32>(uint64_t x) {
81 return static_cast<uint32_t>(x) == x;
82 }
83
84 /// isShiftedUInt<N,S> - Checks if a unsigned integer is an N bit number shifted
85 /// left by S.
86 template<unsigned N, unsigned S>
isShiftedUInt(uint64_t x)87 inline bool isShiftedUInt(uint64_t x) {
88 return isUInt<N+S>(x) && (x % (1<<S) == 0);
89 }
90
91 /// isUIntN - Checks if an unsigned integer fits into the given (dynamic)
92 /// bit width.
isUIntN(unsigned N,uint64_t x)93 inline bool isUIntN(unsigned N, uint64_t x) {
94 return x == (x & (~0ULL >> (64 - N)));
95 }
96
97 /// isIntN - Checks if an signed integer fits into the given (dynamic)
98 /// bit width.
isIntN(unsigned N,int64_t x)99 inline bool isIntN(unsigned N, int64_t x) {
100 return N >= 64 || (-(INT64_C(1)<<(N-1)) <= x && x < (INT64_C(1)<<(N-1)));
101 }
102
103 /// isMask_32 - This function returns true if the argument is a sequence of ones
104 /// starting at the least significant bit with the remainder zero (32 bit
105 /// version). Ex. isMask_32(0x0000FFFFU) == true.
isMask_32(uint32_t Value)106 inline bool isMask_32(uint32_t Value) {
107 return Value && ((Value + 1) & Value) == 0;
108 }
109
110 /// isMask_64 - This function returns true if the argument is a sequence of ones
111 /// starting at the least significant bit with the remainder zero (64 bit
112 /// version).
isMask_64(uint64_t Value)113 inline bool isMask_64(uint64_t Value) {
114 return Value && ((Value + 1) & Value) == 0;
115 }
116
117 /// isShiftedMask_32 - This function returns true if the argument contains a
118 /// sequence of ones with the remainder zero (32 bit version.)
119 /// Ex. isShiftedMask_32(0x0000FF00U) == true.
isShiftedMask_32(uint32_t Value)120 inline bool isShiftedMask_32(uint32_t Value) {
121 return isMask_32((Value - 1) | Value);
122 }
123
124 /// isShiftedMask_64 - This function returns true if the argument contains a
125 /// sequence of ones with the remainder zero (64 bit version.)
isShiftedMask_64(uint64_t Value)126 inline bool isShiftedMask_64(uint64_t Value) {
127 return isMask_64((Value - 1) | Value);
128 }
129
130 /// isPowerOf2_32 - This function returns true if the argument is a power of
131 /// two > 0. Ex. isPowerOf2_32(0x00100000U) == true (32 bit edition.)
isPowerOf2_32(uint32_t Value)132 inline bool isPowerOf2_32(uint32_t Value) {
133 return Value && !(Value & (Value - 1));
134 }
135
136 /// isPowerOf2_64 - This function returns true if the argument is a power of two
137 /// > 0 (64 bit edition.)
isPowerOf2_64(uint64_t Value)138 inline bool isPowerOf2_64(uint64_t Value) {
139 return Value && !(Value & (Value - int64_t(1L)));
140 }
141
142 /// ByteSwap_16 - This function returns a byte-swapped representation of the
143 /// 16-bit argument, Value.
ByteSwap_16(uint16_t Value)144 inline uint16_t ByteSwap_16(uint16_t Value) {
145 return sys::SwapByteOrder_16(Value);
146 }
147
148 /// ByteSwap_32 - This function returns a byte-swapped representation of the
149 /// 32-bit argument, Value.
ByteSwap_32(uint32_t Value)150 inline uint32_t ByteSwap_32(uint32_t Value) {
151 return sys::SwapByteOrder_32(Value);
152 }
153
154 /// ByteSwap_64 - This function returns a byte-swapped representation of the
155 /// 64-bit argument, Value.
ByteSwap_64(uint64_t Value)156 inline uint64_t ByteSwap_64(uint64_t Value) {
157 return sys::SwapByteOrder_64(Value);
158 }
159
160 /// CountLeadingZeros_32 - this function performs the platform optimal form of
161 /// counting the number of zeros from the most significant bit to the first one
162 /// bit. Ex. CountLeadingZeros_32(0x00F000FF) == 8.
163 /// Returns 32 if the word is zero.
CountLeadingZeros_32(uint32_t Value)164 inline unsigned CountLeadingZeros_32(uint32_t Value) {
165 unsigned Count; // result
166 #if __GNUC__ >= 4
167 // PowerPC is defined for __builtin_clz(0)
168 #if !defined(__ppc__) && !defined(__ppc64__)
169 if (!Value) return 32;
170 #endif
171 Count = __builtin_clz(Value);
172 #else
173 if (!Value) return 32;
174 Count = 0;
175 // bisection method for count leading zeros
176 for (unsigned Shift = 32 >> 1; Shift; Shift >>= 1) {
177 uint32_t Tmp = Value >> Shift;
178 if (Tmp) {
179 Value = Tmp;
180 } else {
181 Count |= Shift;
182 }
183 }
184 #endif
185 return Count;
186 }
187
188 /// CountLeadingOnes_32 - this function performs the operation of
189 /// counting the number of ones from the most significant bit to the first zero
190 /// bit. Ex. CountLeadingOnes_32(0xFF0FFF00) == 8.
191 /// Returns 32 if the word is all ones.
CountLeadingOnes_32(uint32_t Value)192 inline unsigned CountLeadingOnes_32(uint32_t Value) {
193 return CountLeadingZeros_32(~Value);
194 }
195
196 /// CountLeadingZeros_64 - This function performs the platform optimal form
197 /// of counting the number of zeros from the most significant bit to the first
198 /// one bit (64 bit edition.)
199 /// Returns 64 if the word is zero.
CountLeadingZeros_64(uint64_t Value)200 inline unsigned CountLeadingZeros_64(uint64_t Value) {
201 unsigned Count; // result
202 #if __GNUC__ >= 4
203 // PowerPC is defined for __builtin_clzll(0)
204 #if !defined(__ppc__) && !defined(__ppc64__)
205 if (!Value) return 64;
206 #endif
207 Count = __builtin_clzll(Value);
208 #else
209 if (sizeof(long) == sizeof(int64_t)) {
210 if (!Value) return 64;
211 Count = 0;
212 // bisection method for count leading zeros
213 for (unsigned Shift = 64 >> 1; Shift; Shift >>= 1) {
214 uint64_t Tmp = Value >> Shift;
215 if (Tmp) {
216 Value = Tmp;
217 } else {
218 Count |= Shift;
219 }
220 }
221 } else {
222 // get hi portion
223 uint32_t Hi = Hi_32(Value);
224
225 // if some bits in hi portion
226 if (Hi) {
227 // leading zeros in hi portion plus all bits in lo portion
228 Count = CountLeadingZeros_32(Hi);
229 } else {
230 // get lo portion
231 uint32_t Lo = Lo_32(Value);
232 // same as 32 bit value
233 Count = CountLeadingZeros_32(Lo)+32;
234 }
235 }
236 #endif
237 return Count;
238 }
239
240 /// CountLeadingOnes_64 - This function performs the operation
241 /// of counting the number of ones from the most significant bit to the first
242 /// zero bit (64 bit edition.)
243 /// Returns 64 if the word is all ones.
CountLeadingOnes_64(uint64_t Value)244 inline unsigned CountLeadingOnes_64(uint64_t Value) {
245 return CountLeadingZeros_64(~Value);
246 }
247
248 /// CountTrailingZeros_32 - this function performs the platform optimal form of
249 /// counting the number of zeros from the least significant bit to the first one
250 /// bit. Ex. CountTrailingZeros_32(0xFF00FF00) == 8.
251 /// Returns 32 if the word is zero.
CountTrailingZeros_32(uint32_t Value)252 inline unsigned CountTrailingZeros_32(uint32_t Value) {
253 #if __GNUC__ >= 4
254 return Value ? __builtin_ctz(Value) : 32;
255 #else
256 static const unsigned Mod37BitPosition[] = {
257 32, 0, 1, 26, 2, 23, 27, 0, 3, 16, 24, 30, 28, 11, 0, 13,
258 4, 7, 17, 0, 25, 22, 31, 15, 29, 10, 12, 6, 0, 21, 14, 9,
259 5, 20, 8, 19, 18
260 };
261 // Replace "-Value" by "1+~Value" in the following commented code to avoid
262 // MSVC warning C4146
263 // return Mod37BitPosition[(-Value & Value) % 37];
264 return Mod37BitPosition[((1 + ~Value) & Value) % 37];
265 #endif
266 }
267
268 /// CountTrailingOnes_32 - this function performs the operation of
269 /// counting the number of ones from the least significant bit to the first zero
270 /// bit. Ex. CountTrailingOnes_32(0x00FF00FF) == 8.
271 /// Returns 32 if the word is all ones.
CountTrailingOnes_32(uint32_t Value)272 inline unsigned CountTrailingOnes_32(uint32_t Value) {
273 return CountTrailingZeros_32(~Value);
274 }
275
276 /// CountTrailingZeros_64 - This function performs the platform optimal form
277 /// of counting the number of zeros from the least significant bit to the first
278 /// one bit (64 bit edition.)
279 /// Returns 64 if the word is zero.
CountTrailingZeros_64(uint64_t Value)280 inline unsigned CountTrailingZeros_64(uint64_t Value) {
281 #if __GNUC__ >= 4
282 return Value ? __builtin_ctzll(Value) : 64;
283 #else
284 static const unsigned Mod67Position[] = {
285 64, 0, 1, 39, 2, 15, 40, 23, 3, 12, 16, 59, 41, 19, 24, 54,
286 4, 64, 13, 10, 17, 62, 60, 28, 42, 30, 20, 51, 25, 44, 55,
287 47, 5, 32, 65, 38, 14, 22, 11, 58, 18, 53, 63, 9, 61, 27,
288 29, 50, 43, 46, 31, 37, 21, 57, 52, 8, 26, 49, 45, 36, 56,
289 7, 48, 35, 6, 34, 33, 0
290 };
291 // Replace "-Value" by "1+~Value" in the following commented code to avoid
292 // MSVC warning C4146
293 // return Mod67Position[(-Value & Value) % 67];
294 return Mod67Position[((1 + ~Value) & Value) % 67];
295 #endif
296 }
297
298 /// CountTrailingOnes_64 - This function performs the operation
299 /// of counting the number of ones from the least significant bit to the first
300 /// zero bit (64 bit edition.)
301 /// Returns 64 if the word is all ones.
CountTrailingOnes_64(uint64_t Value)302 inline unsigned CountTrailingOnes_64(uint64_t Value) {
303 return CountTrailingZeros_64(~Value);
304 }
305
306 /// CountPopulation_32 - this function counts the number of set bits in a value.
307 /// Ex. CountPopulation(0xF000F000) = 8
308 /// Returns 0 if the word is zero.
CountPopulation_32(uint32_t Value)309 inline unsigned CountPopulation_32(uint32_t Value) {
310 #if __GNUC__ >= 4
311 return __builtin_popcount(Value);
312 #else
313 uint32_t v = Value - ((Value >> 1) & 0x55555555);
314 v = (v & 0x33333333) + ((v >> 2) & 0x33333333);
315 return ((v + (v >> 4) & 0xF0F0F0F) * 0x1010101) >> 24;
316 #endif
317 }
318
319 /// CountPopulation_64 - this function counts the number of set bits in a value,
320 /// (64 bit edition.)
CountPopulation_64(uint64_t Value)321 inline unsigned CountPopulation_64(uint64_t Value) {
322 #if __GNUC__ >= 4
323 return __builtin_popcountll(Value);
324 #else
325 uint64_t v = Value - ((Value >> 1) & 0x5555555555555555ULL);
326 v = (v & 0x3333333333333333ULL) + ((v >> 2) & 0x3333333333333333ULL);
327 v = (v + (v >> 4)) & 0x0F0F0F0F0F0F0F0FULL;
328 return unsigned((uint64_t)(v * 0x0101010101010101ULL) >> 56);
329 #endif
330 }
331
332 /// Log2_32 - This function returns the floor log base 2 of the specified value,
333 /// -1 if the value is zero. (32 bit edition.)
334 /// Ex. Log2_32(32) == 5, Log2_32(1) == 0, Log2_32(0) == -1, Log2_32(6) == 2
Log2_32(uint32_t Value)335 inline unsigned Log2_32(uint32_t Value) {
336 return 31 - CountLeadingZeros_32(Value);
337 }
338
339 /// Log2_64 - This function returns the floor log base 2 of the specified value,
340 /// -1 if the value is zero. (64 bit edition.)
Log2_64(uint64_t Value)341 inline unsigned Log2_64(uint64_t Value) {
342 return 63 - CountLeadingZeros_64(Value);
343 }
344
345 /// Log2_32_Ceil - This function returns the ceil log base 2 of the specified
346 /// value, 32 if the value is zero. (32 bit edition).
347 /// Ex. Log2_32_Ceil(32) == 5, Log2_32_Ceil(1) == 0, Log2_32_Ceil(6) == 3
Log2_32_Ceil(uint32_t Value)348 inline unsigned Log2_32_Ceil(uint32_t Value) {
349 return 32-CountLeadingZeros_32(Value-1);
350 }
351
352 /// Log2_64_Ceil - This function returns the ceil log base 2 of the specified
353 /// value, 64 if the value is zero. (64 bit edition.)
Log2_64_Ceil(uint64_t Value)354 inline unsigned Log2_64_Ceil(uint64_t Value) {
355 return 64-CountLeadingZeros_64(Value-1);
356 }
357
358 /// GreatestCommonDivisor64 - Return the greatest common divisor of the two
359 /// values using Euclid's algorithm.
GreatestCommonDivisor64(uint64_t A,uint64_t B)360 inline uint64_t GreatestCommonDivisor64(uint64_t A, uint64_t B) {
361 while (B) {
362 uint64_t T = B;
363 B = A % B;
364 A = T;
365 }
366 return A;
367 }
368
369 /// BitsToDouble - This function takes a 64-bit integer and returns the bit
370 /// equivalent double.
BitsToDouble(uint64_t Bits)371 inline double BitsToDouble(uint64_t Bits) {
372 union {
373 uint64_t L;
374 double D;
375 } T;
376 T.L = Bits;
377 return T.D;
378 }
379
380 /// BitsToFloat - This function takes a 32-bit integer and returns the bit
381 /// equivalent float.
BitsToFloat(uint32_t Bits)382 inline float BitsToFloat(uint32_t Bits) {
383 union {
384 uint32_t I;
385 float F;
386 } T;
387 T.I = Bits;
388 return T.F;
389 }
390
391 /// DoubleToBits - This function takes a double and returns the bit
392 /// equivalent 64-bit integer. Note that copying doubles around
393 /// changes the bits of NaNs on some hosts, notably x86, so this
394 /// routine cannot be used if these bits are needed.
DoubleToBits(double Double)395 inline uint64_t DoubleToBits(double Double) {
396 union {
397 uint64_t L;
398 double D;
399 } T;
400 T.D = Double;
401 return T.L;
402 }
403
404 /// FloatToBits - This function takes a float and returns the bit
405 /// equivalent 32-bit integer. Note that copying floats around
406 /// changes the bits of NaNs on some hosts, notably x86, so this
407 /// routine cannot be used if these bits are needed.
FloatToBits(float Float)408 inline uint32_t FloatToBits(float Float) {
409 union {
410 uint32_t I;
411 float F;
412 } T;
413 T.F = Float;
414 return T.I;
415 }
416
417 /// Platform-independent wrappers for the C99 isnan() function.
418 int IsNAN(float f);
419 int IsNAN(double d);
420
421 /// Platform-independent wrappers for the C99 isinf() function.
422 int IsInf(float f);
423 int IsInf(double d);
424
425 /// MinAlign - A and B are either alignments or offsets. Return the minimum
426 /// alignment that may be assumed after adding the two together.
MinAlign(uint64_t A,uint64_t B)427 inline uint64_t MinAlign(uint64_t A, uint64_t B) {
428 // The largest power of 2 that divides both A and B.
429 //
430 // Replace "-Value" by "1+~Value" in the following commented code to avoid
431 // MSVC warning C4146
432 // return (A | B) & -(A | B);
433 return (A | B) & (1 + ~(A | B));
434 }
435
436 /// NextPowerOf2 - Returns the next power of two (in 64-bits)
437 /// that is strictly greater than A. Returns zero on overflow.
NextPowerOf2(uint64_t A)438 inline uint64_t NextPowerOf2(uint64_t A) {
439 A |= (A >> 1);
440 A |= (A >> 2);
441 A |= (A >> 4);
442 A |= (A >> 8);
443 A |= (A >> 16);
444 A |= (A >> 32);
445 return A + 1;
446 }
447
448 /// Returns the next integer (mod 2**64) that is greater than or equal to
449 /// \p Value and is a multiple of \p Align. \p Align must be non-zero.
450 ///
451 /// Examples:
452 /// \code
453 /// RoundUpToAlignment(5, 8) = 8
454 /// RoundUpToAlignment(17, 8) = 24
455 /// RoundUpToAlignment(~0LL, 8) = 0
456 /// \endcode
RoundUpToAlignment(uint64_t Value,uint64_t Align)457 inline uint64_t RoundUpToAlignment(uint64_t Value, uint64_t Align) {
458 return ((Value + Align - 1) / Align) * Align;
459 }
460
461 /// Returns the offset to the next integer (mod 2**64) that is greater than
462 /// or equal to \p Value and is a multiple of \p Align. \p Align must be
463 /// non-zero.
OffsetToAlignment(uint64_t Value,uint64_t Align)464 inline uint64_t OffsetToAlignment(uint64_t Value, uint64_t Align) {
465 return RoundUpToAlignment(Value, Align) - Value;
466 }
467
468 /// abs64 - absolute value of a 64-bit int. Not all environments support
469 /// "abs" on whatever their name for the 64-bit int type is. The absolute
470 /// value of the largest negative number is undefined, as with "abs".
abs64(int64_t x)471 inline int64_t abs64(int64_t x) {
472 return (x < 0) ? -x : x;
473 }
474
475 /// SignExtend32 - Sign extend B-bit number x to 32-bit int.
476 /// Usage int32_t r = SignExtend32<5>(x);
SignExtend32(uint32_t x)477 template <unsigned B> inline int32_t SignExtend32(uint32_t x) {
478 return int32_t(x << (32 - B)) >> (32 - B);
479 }
480
481 /// \brief Sign extend number in the bottom B bits of X to a 32-bit int.
482 /// Requires 0 < B <= 32.
SignExtend32(uint32_t X,unsigned B)483 inline int32_t SignExtend32(uint32_t X, unsigned B) {
484 return int32_t(X << (32 - B)) >> (32 - B);
485 }
486
487 /// SignExtend64 - Sign extend B-bit number x to 64-bit int.
488 /// Usage int64_t r = SignExtend64<5>(x);
SignExtend64(uint64_t x)489 template <unsigned B> inline int64_t SignExtend64(uint64_t x) {
490 return int64_t(x << (64 - B)) >> (64 - B);
491 }
492
493 /// \brief Sign extend number in the bottom B bits of X to a 64-bit int.
494 /// Requires 0 < B <= 64.
SignExtend64(uint64_t X,unsigned B)495 inline int64_t SignExtend64(uint64_t X, unsigned B) {
496 return int64_t(X << (64 - B)) >> (64 - B);
497 }
498
499 } // End llvm namespace
500
501 #endif
502