]> git.saurik.com Git - apple/objc4.git/blob - runtime/llvm-MathExtras.h
objc4-706.tar.gz
[apple/objc4.git] / runtime / llvm-MathExtras.h
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 // Taken from llvmCore-3425.0.31.
15
16 #ifndef LLVM_SUPPORT_MATHEXTRAS_H
17 #define LLVM_SUPPORT_MATHEXTRAS_H
18
19 namespace objc {
20
21 // NOTE: The following support functions use the _32/_64 extensions instead of
22 // type overloading so that signed and unsigned integers can be used without
23 // ambiguity.
24
25 /// Hi_32 - This function returns the high 32 bits of a 64 bit value.
26 inline uint32_t Hi_32(uint64_t Value) {
27 return static_cast<uint32_t>(Value >> 32);
28 }
29
30 /// Lo_32 - This function returns the low 32 bits of a 64 bit value.
31 inline uint32_t Lo_32(uint64_t Value) {
32 return static_cast<uint32_t>(Value);
33 }
34
35 /// isInt - Checks if an integer fits into the given bit width.
36 template<unsigned N>
37 inline bool isInt(int64_t x) {
38 return N >= 64 || (-(INT64_C(1)<<(N-1)) <= x && x < (INT64_C(1)<<(N-1)));
39 }
40 // Template specializations to get better code for common cases.
41 template<>
42 inline bool isInt<8>(int64_t x) {
43 return static_cast<int8_t>(x) == x;
44 }
45 template<>
46 inline bool isInt<16>(int64_t x) {
47 return static_cast<int16_t>(x) == x;
48 }
49 template<>
50 inline bool isInt<32>(int64_t x) {
51 return static_cast<int32_t>(x) == x;
52 }
53
54 /// isShiftedInt<N,S> - Checks if a signed integer is an N bit number shifted
55 /// left by S.
56 template<unsigned N, unsigned S>
57 inline bool isShiftedInt(int64_t x) {
58 return isInt<N+S>(x) && (x % (1<<S) == 0);
59 }
60
61 /// isUInt - Checks if an unsigned integer fits into the given bit width.
62 template<unsigned N>
63 inline bool isUInt(uint64_t x) {
64 return N >= 64 || x < (UINT64_C(1)<<N);
65 }
66 // Template specializations to get better code for common cases.
67 template<>
68 inline bool isUInt<8>(uint64_t x) {
69 return static_cast<uint8_t>(x) == x;
70 }
71 template<>
72 inline bool isUInt<16>(uint64_t x) {
73 return static_cast<uint16_t>(x) == x;
74 }
75 template<>
76 inline bool isUInt<32>(uint64_t x) {
77 return static_cast<uint32_t>(x) == x;
78 }
79
80 /// isShiftedUInt<N,S> - Checks if a unsigned integer is an N bit number shifted
81 /// left by S.
82 template<unsigned N, unsigned S>
83 inline bool isShiftedUInt(uint64_t x) {
84 return isUInt<N+S>(x) && (x % (1<<S) == 0);
85 }
86
87 /// isUIntN - Checks if an unsigned integer fits into the given (dynamic)
88 /// bit width.
89 inline bool isUIntN(unsigned N, uint64_t x) {
90 return x == (x & (~0ULL >> (64 - N)));
91 }
92
93 /// isIntN - Checks if an signed integer fits into the given (dynamic)
94 /// bit width.
95 inline bool isIntN(unsigned N, int64_t x) {
96 return N >= 64 || (-(INT64_C(1)<<(N-1)) <= x && x < (INT64_C(1)<<(N-1)));
97 }
98
99 /// isMask_32 - This function returns true if the argument is a sequence of ones
100 /// starting at the least significant bit with the remainder zero (32 bit
101 /// version). Ex. isMask_32(0x0000FFFFU) == true.
102 inline bool isMask_32(uint32_t Value) {
103 return Value && ((Value + 1) & Value) == 0;
104 }
105
106 /// isMask_64 - This function returns true if the argument is a sequence of ones
107 /// starting at the least significant bit with the remainder zero (64 bit
108 /// version).
109 inline bool isMask_64(uint64_t Value) {
110 return Value && ((Value + 1) & Value) == 0;
111 }
112
113 /// isShiftedMask_32 - This function returns true if the argument contains a
114 /// sequence of ones with the remainder zero (32 bit version.)
115 /// Ex. isShiftedMask_32(0x0000FF00U) == true.
116 inline bool isShiftedMask_32(uint32_t Value) {
117 return isMask_32((Value - 1) | Value);
118 }
119
120 /// isShiftedMask_64 - This function returns true if the argument contains a
121 /// sequence of ones with the remainder zero (64 bit version.)
122 inline bool isShiftedMask_64(uint64_t Value) {
123 return isMask_64((Value - 1) | Value);
124 }
125
126 /// isPowerOf2_32 - This function returns true if the argument is a power of
127 /// two > 0. Ex. isPowerOf2_32(0x00100000U) == true (32 bit edition.)
128 inline bool isPowerOf2_32(uint32_t Value) {
129 return Value && !(Value & (Value - 1));
130 }
131
132 /// isPowerOf2_64 - This function returns true if the argument is a power of two
133 /// > 0 (64 bit edition.)
134 inline bool isPowerOf2_64(uint64_t Value) {
135 return Value && !(Value & (Value - int64_t(1L)));
136 }
137
138 /// CountLeadingZeros_32 - this function performs the platform optimal form of
139 /// counting the number of zeros from the most significant bit to the first one
140 /// bit. Ex. CountLeadingZeros_32(0x00F000FF) == 8.
141 /// Returns 32 if the word is zero.
142 inline unsigned CountLeadingZeros_32(uint32_t Value) {
143 unsigned Count; // result
144 #if __GNUC__ >= 4
145 // PowerPC is defined for __builtin_clz(0)
146 #if !defined(__ppc__) && !defined(__ppc64__)
147 if (!Value) return 32;
148 #endif
149 Count = __builtin_clz(Value);
150 #else
151 if (!Value) return 32;
152 Count = 0;
153 // bisection method for count leading zeros
154 for (unsigned Shift = 32 >> 1; Shift; Shift >>= 1) {
155 uint32_t Tmp = Value >> Shift;
156 if (Tmp) {
157 Value = Tmp;
158 } else {
159 Count |= Shift;
160 }
161 }
162 #endif
163 return Count;
164 }
165
166 /// CountLeadingOnes_32 - this function performs the operation of
167 /// counting the number of ones from the most significant bit to the first zero
168 /// bit. Ex. CountLeadingOnes_32(0xFF0FFF00) == 8.
169 /// Returns 32 if the word is all ones.
170 inline unsigned CountLeadingOnes_32(uint32_t Value) {
171 return CountLeadingZeros_32(~Value);
172 }
173
174 /// CountLeadingZeros_64 - This function performs the platform optimal form
175 /// of counting the number of zeros from the most significant bit to the first
176 /// one bit (64 bit edition.)
177 /// Returns 64 if the word is zero.
178 inline unsigned CountLeadingZeros_64(uint64_t Value) {
179 unsigned Count; // result
180 #if __GNUC__ >= 4
181 // PowerPC is defined for __builtin_clzll(0)
182 #if !defined(__ppc__) && !defined(__ppc64__)
183 if (!Value) return 64;
184 #endif
185 Count = __builtin_clzll(Value);
186 #else
187 if (sizeof(long) == sizeof(int64_t)) {
188 if (!Value) return 64;
189 Count = 0;
190 // bisection method for count leading zeros
191 for (unsigned Shift = 64 >> 1; Shift; Shift >>= 1) {
192 uint64_t Tmp = Value >> Shift;
193 if (Tmp) {
194 Value = Tmp;
195 } else {
196 Count |= Shift;
197 }
198 }
199 } else {
200 // get hi portion
201 uint32_t Hi = Hi_32(Value);
202
203 // if some bits in hi portion
204 if (Hi) {
205 // leading zeros in hi portion plus all bits in lo portion
206 Count = CountLeadingZeros_32(Hi);
207 } else {
208 // get lo portion
209 uint32_t Lo = Lo_32(Value);
210 // same as 32 bit value
211 Count = CountLeadingZeros_32(Lo)+32;
212 }
213 }
214 #endif
215 return Count;
216 }
217
218 /// CountLeadingOnes_64 - This function performs the operation
219 /// of counting the number of ones from the most significant bit to the first
220 /// zero bit (64 bit edition.)
221 /// Returns 64 if the word is all ones.
222 inline unsigned CountLeadingOnes_64(uint64_t Value) {
223 return CountLeadingZeros_64(~Value);
224 }
225
226 /// CountTrailingZeros_32 - this function performs the platform optimal form of
227 /// counting the number of zeros from the least significant bit to the first one
228 /// bit. Ex. CountTrailingZeros_32(0xFF00FF00) == 8.
229 /// Returns 32 if the word is zero.
230 inline unsigned CountTrailingZeros_32(uint32_t Value) {
231 #if __GNUC__ >= 4
232 return Value ? __builtin_ctz(Value) : 32;
233 #else
234 static const unsigned Mod37BitPosition[] = {
235 32, 0, 1, 26, 2, 23, 27, 0, 3, 16, 24, 30, 28, 11, 0, 13,
236 4, 7, 17, 0, 25, 22, 31, 15, 29, 10, 12, 6, 0, 21, 14, 9,
237 5, 20, 8, 19, 18
238 };
239 return Mod37BitPosition[(-Value & Value) % 37];
240 #endif
241 }
242
243 /// CountTrailingOnes_32 - this function performs the operation of
244 /// counting the number of ones from the least significant bit to the first zero
245 /// bit. Ex. CountTrailingOnes_32(0x00FF00FF) == 8.
246 /// Returns 32 if the word is all ones.
247 inline unsigned CountTrailingOnes_32(uint32_t Value) {
248 return CountTrailingZeros_32(~Value);
249 }
250
251 /// CountTrailingZeros_64 - This function performs the platform optimal form
252 /// of counting the number of zeros from the least significant bit to the first
253 /// one bit (64 bit edition.)
254 /// Returns 64 if the word is zero.
255 inline unsigned CountTrailingZeros_64(uint64_t Value) {
256 #if __GNUC__ >= 4
257 return Value ? __builtin_ctzll(Value) : 64;
258 #else
259 static const unsigned Mod67Position[] = {
260 64, 0, 1, 39, 2, 15, 40, 23, 3, 12, 16, 59, 41, 19, 24, 54,
261 4, 64, 13, 10, 17, 62, 60, 28, 42, 30, 20, 51, 25, 44, 55,
262 47, 5, 32, 65, 38, 14, 22, 11, 58, 18, 53, 63, 9, 61, 27,
263 29, 50, 43, 46, 31, 37, 21, 57, 52, 8, 26, 49, 45, 36, 56,
264 7, 48, 35, 6, 34, 33, 0
265 };
266 return Mod67Position[(-Value & Value) % 67];
267 #endif
268 }
269
270 /// CountTrailingOnes_64 - This function performs the operation
271 /// of counting the number of ones from the least significant bit to the first
272 /// zero bit (64 bit edition.)
273 /// Returns 64 if the word is all ones.
274 inline unsigned CountTrailingOnes_64(uint64_t Value) {
275 return CountTrailingZeros_64(~Value);
276 }
277
278 /// CountPopulation_32 - this function counts the number of set bits in a value.
279 /// Ex. CountPopulation(0xF000F000) = 8
280 /// Returns 0 if the word is zero.
281 inline unsigned CountPopulation_32(uint32_t Value) {
282 #if __GNUC__ >= 4
283 return __builtin_popcount(Value);
284 #else
285 uint32_t v = Value - ((Value >> 1) & 0x55555555);
286 v = (v & 0x33333333) + ((v >> 2) & 0x33333333);
287 return ((v + (v >> 4) & 0xF0F0F0F) * 0x1010101) >> 24;
288 #endif
289 }
290
291 /// CountPopulation_64 - this function counts the number of set bits in a value,
292 /// (64 bit edition.)
293 inline unsigned CountPopulation_64(uint64_t Value) {
294 #if __GNUC__ >= 4
295 return __builtin_popcountll(Value);
296 #else
297 uint64_t v = Value - ((Value >> 1) & 0x5555555555555555ULL);
298 v = (v & 0x3333333333333333ULL) + ((v >> 2) & 0x3333333333333333ULL);
299 v = (v + (v >> 4)) & 0x0F0F0F0F0F0F0F0FULL;
300 return unsigned((uint64_t)(v * 0x0101010101010101ULL) >> 56);
301 #endif
302 }
303
304 /// Log2_32 - This function returns the floor log base 2 of the specified value,
305 /// -1 if the value is zero. (32 bit edition.)
306 /// Ex. Log2_32(32) == 5, Log2_32(1) == 0, Log2_32(0) == -1, Log2_32(6) == 2
307 inline unsigned Log2_32(uint32_t Value) {
308 return 31 - CountLeadingZeros_32(Value);
309 }
310
311 /// Log2_64 - This function returns the floor log base 2 of the specified value,
312 /// -1 if the value is zero. (64 bit edition.)
313 inline unsigned Log2_64(uint64_t Value) {
314 return 63 - CountLeadingZeros_64(Value);
315 }
316
317 /// Log2_32_Ceil - This function returns the ceil log base 2 of the specified
318 /// value, 32 if the value is zero. (32 bit edition).
319 /// Ex. Log2_32_Ceil(32) == 5, Log2_32_Ceil(1) == 0, Log2_32_Ceil(6) == 3
320 inline unsigned Log2_32_Ceil(uint32_t Value) {
321 return 32-CountLeadingZeros_32(Value-1);
322 }
323
324 /// Log2_64_Ceil - This function returns the ceil log base 2 of the specified
325 /// value, 64 if the value is zero. (64 bit edition.)
326 inline unsigned Log2_64_Ceil(uint64_t Value) {
327 return 64-CountLeadingZeros_64(Value-1);
328 }
329
330 /// GreatestCommonDivisor64 - Return the greatest common divisor of the two
331 /// values using Euclid's algorithm.
332 inline uint64_t GreatestCommonDivisor64(uint64_t A, uint64_t B) {
333 while (B) {
334 uint64_t T = B;
335 B = A % B;
336 A = T;
337 }
338 return A;
339 }
340
341 /// BitsToDouble - This function takes a 64-bit integer and returns the bit
342 /// equivalent double.
343 inline double BitsToDouble(uint64_t Bits) {
344 union {
345 uint64_t L;
346 double D;
347 } T;
348 T.L = Bits;
349 return T.D;
350 }
351
352 /// BitsToFloat - This function takes a 32-bit integer and returns the bit
353 /// equivalent float.
354 inline float BitsToFloat(uint32_t Bits) {
355 union {
356 uint32_t I;
357 float F;
358 } T;
359 T.I = Bits;
360 return T.F;
361 }
362
363 /// DoubleToBits - This function takes a double and returns the bit
364 /// equivalent 64-bit integer. Note that copying doubles around
365 /// changes the bits of NaNs on some hosts, notably x86, so this
366 /// routine cannot be used if these bits are needed.
367 inline uint64_t DoubleToBits(double Double) {
368 union {
369 uint64_t L;
370 double D;
371 } T;
372 T.D = Double;
373 return T.L;
374 }
375
376 /// FloatToBits - This function takes a float and returns the bit
377 /// equivalent 32-bit integer. Note that copying floats around
378 /// changes the bits of NaNs on some hosts, notably x86, so this
379 /// routine cannot be used if these bits are needed.
380 inline uint32_t FloatToBits(float Float) {
381 union {
382 uint32_t I;
383 float F;
384 } T;
385 T.F = Float;
386 return T.I;
387 }
388
389 /// Platform-independent wrappers for the C99 isnan() function.
390 int IsNAN(float f);
391 int IsNAN(double d);
392
393 /// Platform-independent wrappers for the C99 isinf() function.
394 int IsInf(float f);
395 int IsInf(double d);
396
397 /// MinAlign - A and B are either alignments or offsets. Return the minimum
398 /// alignment that may be assumed after adding the two together.
399 inline uint64_t MinAlign(uint64_t A, uint64_t B) {
400 // The largest power of 2 that divides both A and B.
401 return (A | B) & -(A | B);
402 }
403
404 /// NextPowerOf2 - Returns the next power of two (in 64-bits)
405 /// that is strictly greater than A. Returns zero on overflow.
406 inline uint64_t NextPowerOf2(uint64_t A) {
407 A |= (A >> 1);
408 A |= (A >> 2);
409 A |= (A >> 4);
410 A |= (A >> 8);
411 A |= (A >> 16);
412 A |= (A >> 32);
413 return A + 1;
414 }
415
416 /// NextPowerOf2 - Returns the next power of two (in 32-bits)
417 /// that is strictly greater than A. Returns zero on overflow.
418 inline uint32_t NextPowerOf2(uint32_t A) {
419 A |= (A >> 1);
420 A |= (A >> 2);
421 A |= (A >> 4);
422 A |= (A >> 8);
423 A |= (A >> 16);
424 return A + 1;
425 }
426
427 /// Returns the next integer (mod 2**64) that is greater than or equal to
428 /// \p Value and is a multiple of \p Align. \p Align must be non-zero.
429 ///
430 /// Examples:
431 /// \code
432 /// RoundUpToAlignment(5, 8) = 8
433 /// RoundUpToAlignment(17, 8) = 24
434 /// RoundUpToAlignment(~0LL, 8) = 0
435 /// \endcode
436 inline uint64_t RoundUpToAlignment(uint64_t Value, uint64_t Align) {
437 return ((Value + Align - 1) / Align) * Align;
438 }
439
440 /// Returns the offset to the next integer (mod 2**64) that is greater than
441 /// or equal to \p Value and is a multiple of \p Align. \p Align must be
442 /// non-zero.
443 inline uint64_t OffsetToAlignment(uint64_t Value, uint64_t Align) {
444 return RoundUpToAlignment(Value, Align) - Value;
445 }
446
447 /// abs64 - absolute value of a 64-bit int. Not all environments support
448 /// "abs" on whatever their name for the 64-bit int type is. The absolute
449 /// value of the largest negative number is undefined, as with "abs".
450 inline int64_t abs64(int64_t x) {
451 return (x < 0) ? -x : x;
452 }
453
454 /// SignExtend32 - Sign extend B-bit number x to 32-bit int.
455 /// Usage int32_t r = SignExtend32<5>(x);
456 template <unsigned B> inline int32_t SignExtend32(uint32_t x) {
457 return int32_t(x << (32 - B)) >> (32 - B);
458 }
459
460 /// \brief Sign extend number in the bottom B bits of X to a 32-bit int.
461 /// Requires 0 < B <= 32.
462 inline int32_t SignExtend32(uint32_t X, unsigned B) {
463 return int32_t(X << (32 - B)) >> (32 - B);
464 }
465
466 /// SignExtend64 - Sign extend B-bit number x to 64-bit int.
467 /// Usage int64_t r = SignExtend64<5>(x);
468 template <unsigned B> inline int64_t SignExtend64(uint64_t x) {
469 return int64_t(x << (64 - B)) >> (64 - B);
470 }
471
472 /// \brief Sign extend number in the bottom B bits of X to a 64-bit int.
473 /// Requires 0 < B <= 64.
474 inline int64_t SignExtend64(uint64_t X, unsigned B) {
475 return int64_t(X << (64 - B)) >> (64 - B);
476 }
477
478 } // End llvm namespace
479
480 #endif