| OLD | NEW |
| (Empty) |
| 1 // Copyright 2014 The Chromium Authors. All rights reserved. | |
| 2 // Use of this source code is governed by a BSD-style license that can be | |
| 3 // found in the LICENSE file. | |
| 4 | |
| 5 #ifndef SAFE_MATH_IMPL_H_ | |
| 6 #define SAFE_MATH_IMPL_H_ | |
| 7 | |
| 8 #include <stdint.h> | |
| 9 | |
| 10 #include <cmath> | |
| 11 #include <cstdlib> | |
| 12 #include <limits> | |
| 13 | |
| 14 #include "../macros.h" | |
| 15 #include "../template_util.h" | |
| 16 #include "safe_conversions.h" | |
| 17 | |
| 18 namespace base { | |
| 19 namespace internal { | |
| 20 | |
| 21 // Everything from here up to the floating point operations is portable C++, | |
| 22 // but it may not be fast. This code could be split based on | |
| 23 // platform/architecture and replaced with potentially faster implementations. | |
| 24 | |
| 25 // Integer promotion templates used by the portable checked integer arithmetic. | |
| 26 template <size_t Size, bool IsSigned> | |
| 27 struct IntegerForSizeAndSign; | |
| 28 template <> | |
| 29 struct IntegerForSizeAndSign<1, true> { | |
| 30 typedef int8_t type; | |
| 31 }; | |
| 32 template <> | |
| 33 struct IntegerForSizeAndSign<1, false> { | |
| 34 typedef uint8_t type; | |
| 35 }; | |
| 36 template <> | |
| 37 struct IntegerForSizeAndSign<2, true> { | |
| 38 typedef int16_t type; | |
| 39 }; | |
| 40 template <> | |
| 41 struct IntegerForSizeAndSign<2, false> { | |
| 42 typedef uint16_t type; | |
| 43 }; | |
| 44 template <> | |
| 45 struct IntegerForSizeAndSign<4, true> { | |
| 46 typedef int32_t type; | |
| 47 }; | |
| 48 template <> | |
| 49 struct IntegerForSizeAndSign<4, false> { | |
| 50 typedef uint32_t type; | |
| 51 }; | |
| 52 template <> | |
| 53 struct IntegerForSizeAndSign<8, true> { | |
| 54 typedef int64_t type; | |
| 55 }; | |
| 56 template <> | |
| 57 struct IntegerForSizeAndSign<8, false> { | |
| 58 typedef uint64_t type; | |
| 59 }; | |
| 60 | |
| 61 // WARNING: We have no IntegerForSizeAndSign<16, *>. If we ever add one to | |
| 62 // support 128-bit math, then the ArithmeticPromotion template below will need | |
| 63 // to be updated (or more likely replaced with a decltype expression). | |
| 64 | |
| 65 template <typename Integer> | |
| 66 struct UnsignedIntegerForSize { | |
| 67 typedef typename enable_if< | |
| 68 std::numeric_limits<Integer>::is_integer, | |
| 69 typename IntegerForSizeAndSign<sizeof(Integer), false>::type>::type type; | |
| 70 }; | |
| 71 | |
| 72 template <typename Integer> | |
| 73 struct SignedIntegerForSize { | |
| 74 typedef typename enable_if< | |
| 75 std::numeric_limits<Integer>::is_integer, | |
| 76 typename IntegerForSizeAndSign<sizeof(Integer), true>::type>::type type; | |
| 77 }; | |
| 78 | |
| 79 template <typename Integer> | |
| 80 struct TwiceWiderInteger { | |
| 81 typedef typename enable_if< | |
| 82 std::numeric_limits<Integer>::is_integer, | |
| 83 typename IntegerForSizeAndSign< | |
| 84 sizeof(Integer) * 2, | |
| 85 std::numeric_limits<Integer>::is_signed>::type>::type type; | |
| 86 }; | |
| 87 | |
| 88 template <typename Integer> | |
| 89 struct PositionOfSignBit { | |
| 90 static const typename enable_if<std::numeric_limits<Integer>::is_integer, | |
| 91 size_t>::type value = 8 * sizeof(Integer) - 1; | |
| 92 }; | |
| 93 | |
| 94 // Helper templates for integer manipulations. | |
| 95 | |
| 96 template <typename T> | |
| 97 bool HasSignBit(T x) { | |
| 98 // Cast to unsigned since right shift on signed is undefined. | |
| 99 return !!(static_cast<typename UnsignedIntegerForSize<T>::type>(x) >> | |
| 100 PositionOfSignBit<T>::value); | |
| 101 } | |
| 102 | |
| 103 // This wrapper undoes the standard integer promotions. | |
| 104 template <typename T> | |
| 105 T BinaryComplement(T x) { | |
| 106 return ~x; | |
| 107 } | |
| 108 | |
| 109 // Here are the actual portable checked integer math implementations. | |
| 110 // TODO(jschuh): Break this code out from the enable_if pattern and find a clean | |
| 111 // way to coalesce things into the CheckedNumericState specializations below. | |
| 112 | |
| 113 template <typename T> | |
| 114 typename enable_if<std::numeric_limits<T>::is_integer, T>::type | |
| 115 CheckedAdd(T x, T y, RangeConstraint* validity) { | |
| 116 // Since the value of x+y is undefined if we have a signed type, we compute | |
| 117 // it using the unsigned type of the same size. | |
| 118 typedef typename UnsignedIntegerForSize<T>::type UnsignedDst; | |
| 119 UnsignedDst ux = static_cast<UnsignedDst>(x); | |
| 120 UnsignedDst uy = static_cast<UnsignedDst>(y); | |
| 121 UnsignedDst uresult = ux + uy; | |
| 122 // Addition is valid if the sign of (x + y) is equal to either that of x or | |
| 123 // that of y. | |
| 124 if (std::numeric_limits<T>::is_signed) { | |
| 125 if (HasSignBit(BinaryComplement((uresult ^ ux) & (uresult ^ uy)))) | |
| 126 *validity = RANGE_VALID; | |
| 127 else // Direction of wrap is inverse of result sign. | |
| 128 *validity = HasSignBit(uresult) ? RANGE_OVERFLOW : RANGE_UNDERFLOW; | |
| 129 | |
| 130 } else { // Unsigned is either valid or overflow. | |
| 131 *validity = BinaryComplement(x) >= y ? RANGE_VALID : RANGE_OVERFLOW; | |
| 132 } | |
| 133 return static_cast<T>(uresult); | |
| 134 } | |
| 135 | |
| 136 template <typename T> | |
| 137 typename enable_if<std::numeric_limits<T>::is_integer, T>::type | |
| 138 CheckedSub(T x, T y, RangeConstraint* validity) { | |
| 139 // Since the value of x+y is undefined if we have a signed type, we compute | |
| 140 // it using the unsigned type of the same size. | |
| 141 typedef typename UnsignedIntegerForSize<T>::type UnsignedDst; | |
| 142 UnsignedDst ux = static_cast<UnsignedDst>(x); | |
| 143 UnsignedDst uy = static_cast<UnsignedDst>(y); | |
| 144 UnsignedDst uresult = ux - uy; | |
| 145 // Subtraction is valid if either x and y have same sign, or (x-y) and x have | |
| 146 // the same sign. | |
| 147 if (std::numeric_limits<T>::is_signed) { | |
| 148 if (HasSignBit(BinaryComplement((uresult ^ ux) & (ux ^ uy)))) | |
| 149 *validity = RANGE_VALID; | |
| 150 else // Direction of wrap is inverse of result sign. | |
| 151 *validity = HasSignBit(uresult) ? RANGE_OVERFLOW : RANGE_UNDERFLOW; | |
| 152 | |
| 153 } else { // Unsigned is either valid or underflow. | |
| 154 *validity = x >= y ? RANGE_VALID : RANGE_UNDERFLOW; | |
| 155 } | |
| 156 return static_cast<T>(uresult); | |
| 157 } | |
| 158 | |
| 159 // Integer multiplication is a bit complicated. In the fast case we just | |
| 160 // we just promote to a twice wider type, and range check the result. In the | |
| 161 // slow case we need to manually check that the result won't be truncated by | |
| 162 // checking with division against the appropriate bound. | |
| 163 template <typename T> | |
| 164 typename enable_if< | |
| 165 std::numeric_limits<T>::is_integer && sizeof(T) * 2 <= sizeof(uintmax_t), | |
| 166 T>::type | |
| 167 CheckedMul(T x, T y, RangeConstraint* validity) { | |
| 168 typedef typename TwiceWiderInteger<T>::type IntermediateType; | |
| 169 IntermediateType tmp = | |
| 170 static_cast<IntermediateType>(x) * static_cast<IntermediateType>(y); | |
| 171 *validity = DstRangeRelationToSrcRange<T>(tmp); | |
| 172 return static_cast<T>(tmp); | |
| 173 } | |
| 174 | |
| 175 template <typename T> | |
| 176 typename enable_if<std::numeric_limits<T>::is_integer&& std::numeric_limits< | |
| 177 T>::is_signed&&(sizeof(T) * 2 > sizeof(uintmax_t)), | |
| 178 T>::type | |
| 179 CheckedMul(T x, T y, RangeConstraint* validity) { | |
| 180 // if either side is zero then the result will be zero. | |
| 181 if (!(x || y)) { | |
| 182 return RANGE_VALID; | |
| 183 | |
| 184 } else if (x > 0) { | |
| 185 if (y > 0) | |
| 186 *validity = | |
| 187 x <= std::numeric_limits<T>::max() / y ? RANGE_VALID : RANGE_OVERFLOW; | |
| 188 else | |
| 189 *validity = y >= std::numeric_limits<T>::min() / x ? RANGE_VALID | |
| 190 : RANGE_UNDERFLOW; | |
| 191 | |
| 192 } else { | |
| 193 if (y > 0) | |
| 194 *validity = x >= std::numeric_limits<T>::min() / y ? RANGE_VALID | |
| 195 : RANGE_UNDERFLOW; | |
| 196 else | |
| 197 *validity = | |
| 198 y >= std::numeric_limits<T>::max() / x ? RANGE_VALID : RANGE_OVERFLOW; | |
| 199 } | |
| 200 | |
| 201 return x * y; | |
| 202 } | |
| 203 | |
| 204 template <typename T> | |
| 205 typename enable_if<std::numeric_limits<T>::is_integer && | |
| 206 !std::numeric_limits<T>::is_signed && | |
| 207 (sizeof(T) * 2 > sizeof(uintmax_t)), | |
| 208 T>::type | |
| 209 CheckedMul(T x, T y, RangeConstraint* validity) { | |
| 210 *validity = (y == 0 || x <= std::numeric_limits<T>::max() / y) | |
| 211 ? RANGE_VALID | |
| 212 : RANGE_OVERFLOW; | |
| 213 return x * y; | |
| 214 } | |
| 215 | |
| 216 // Division just requires a check for an invalid negation on signed min/-1. | |
| 217 template <typename T> | |
| 218 T CheckedDiv( | |
| 219 T x, | |
| 220 T y, | |
| 221 RangeConstraint* validity, | |
| 222 typename enable_if<std::numeric_limits<T>::is_integer, int>::type = 0) { | |
| 223 if (std::numeric_limits<T>::is_signed && x == std::numeric_limits<T>::min() && | |
| 224 y == static_cast<T>(-1)) { | |
| 225 *validity = RANGE_OVERFLOW; | |
| 226 return std::numeric_limits<T>::min(); | |
| 227 } | |
| 228 | |
| 229 *validity = RANGE_VALID; | |
| 230 return x / y; | |
| 231 } | |
| 232 | |
| 233 template <typename T> | |
| 234 typename enable_if< | |
| 235 std::numeric_limits<T>::is_integer&& std::numeric_limits<T>::is_signed, | |
| 236 T>::type | |
| 237 CheckedMod(T x, T y, RangeConstraint* validity) { | |
| 238 *validity = y > 0 ? RANGE_VALID : RANGE_INVALID; | |
| 239 return x % y; | |
| 240 } | |
| 241 | |
| 242 template <typename T> | |
| 243 typename enable_if< | |
| 244 std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, | |
| 245 T>::type | |
| 246 CheckedMod(T x, T y, RangeConstraint* validity) { | |
| 247 *validity = RANGE_VALID; | |
| 248 return x % y; | |
| 249 } | |
| 250 | |
| 251 template <typename T> | |
| 252 typename enable_if< | |
| 253 std::numeric_limits<T>::is_integer&& std::numeric_limits<T>::is_signed, | |
| 254 T>::type | |
| 255 CheckedNeg(T value, RangeConstraint* validity) { | |
| 256 *validity = | |
| 257 value != std::numeric_limits<T>::min() ? RANGE_VALID : RANGE_OVERFLOW; | |
| 258 // The negation of signed min is min, so catch that one. | |
| 259 return -value; | |
| 260 } | |
| 261 | |
| 262 template <typename T> | |
| 263 typename enable_if< | |
| 264 std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, | |
| 265 T>::type | |
| 266 CheckedNeg(T value, RangeConstraint* validity) { | |
| 267 // The only legal unsigned negation is zero. | |
| 268 *validity = value ? RANGE_UNDERFLOW : RANGE_VALID; | |
| 269 return static_cast<T>( | |
| 270 -static_cast<typename SignedIntegerForSize<T>::type>(value)); | |
| 271 } | |
| 272 | |
| 273 template <typename T> | |
| 274 typename enable_if< | |
| 275 std::numeric_limits<T>::is_integer&& std::numeric_limits<T>::is_signed, | |
| 276 T>::type | |
| 277 CheckedAbs(T value, RangeConstraint* validity) { | |
| 278 *validity = | |
| 279 value != std::numeric_limits<T>::min() ? RANGE_VALID : RANGE_OVERFLOW; | |
| 280 return std::abs(value); | |
| 281 } | |
| 282 | |
| 283 template <typename T> | |
| 284 typename enable_if< | |
| 285 std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, | |
| 286 T>::type | |
| 287 CheckedAbs(T value, RangeConstraint* validity) { | |
| 288 // Absolute value of a positive is just its identiy. | |
| 289 *validity = RANGE_VALID; | |
| 290 return value; | |
| 291 } | |
| 292 | |
| 293 // These are the floating point stubs that the compiler needs to see. Only the | |
| 294 // negation operation is ever called. | |
| 295 #define BASE_FLOAT_ARITHMETIC_STUBS(NAME) \ | |
| 296 template <typename T> \ | |
| 297 typename enable_if<std::numeric_limits<T>::is_iec559, T>::type \ | |
| 298 Checked##NAME(T, T, RangeConstraint*) { \ | |
| 299 NOTREACHED(); \ | |
| 300 return 0; \ | |
| 301 } | |
| 302 | |
| 303 BASE_FLOAT_ARITHMETIC_STUBS(Add) | |
| 304 BASE_FLOAT_ARITHMETIC_STUBS(Sub) | |
| 305 BASE_FLOAT_ARITHMETIC_STUBS(Mul) | |
| 306 BASE_FLOAT_ARITHMETIC_STUBS(Div) | |
| 307 BASE_FLOAT_ARITHMETIC_STUBS(Mod) | |
| 308 | |
| 309 #undef BASE_FLOAT_ARITHMETIC_STUBS | |
| 310 | |
| 311 template <typename T> | |
| 312 typename enable_if<std::numeric_limits<T>::is_iec559, T>::type CheckedNeg( | |
| 313 T value, | |
| 314 RangeConstraint*) { | |
| 315 return -value; | |
| 316 } | |
| 317 | |
| 318 template <typename T> | |
| 319 typename enable_if<std::numeric_limits<T>::is_iec559, T>::type CheckedAbs( | |
| 320 T value, | |
| 321 RangeConstraint*) { | |
| 322 return std::abs(value); | |
| 323 } | |
| 324 | |
| 325 // Floats carry around their validity state with them, but integers do not. So, | |
| 326 // we wrap the underlying value in a specialization in order to hide that detail | |
| 327 // and expose an interface via accessors. | |
| 328 enum NumericRepresentation { | |
| 329 NUMERIC_INTEGER, | |
| 330 NUMERIC_FLOATING, | |
| 331 NUMERIC_UNKNOWN | |
| 332 }; | |
| 333 | |
| 334 template <typename NumericType> | |
| 335 struct GetNumericRepresentation { | |
| 336 static const NumericRepresentation value = | |
| 337 std::numeric_limits<NumericType>::is_integer | |
| 338 ? NUMERIC_INTEGER | |
| 339 : (std::numeric_limits<NumericType>::is_iec559 ? NUMERIC_FLOATING | |
| 340 : NUMERIC_UNKNOWN); | |
| 341 }; | |
| 342 | |
| 343 template <typename T, NumericRepresentation type = | |
| 344 GetNumericRepresentation<T>::value> | |
| 345 class CheckedNumericState {}; | |
| 346 | |
| 347 // Integrals require quite a bit of additional housekeeping to manage state. | |
| 348 template <typename T> | |
| 349 class CheckedNumericState<T, NUMERIC_INTEGER> { | |
| 350 private: | |
| 351 T value_; | |
| 352 RangeConstraint validity_; | |
| 353 | |
| 354 public: | |
| 355 template <typename Src, NumericRepresentation type> | |
| 356 friend class CheckedNumericState; | |
| 357 | |
| 358 CheckedNumericState() : value_(0), validity_(RANGE_VALID) {} | |
| 359 | |
| 360 template <typename Src> | |
| 361 CheckedNumericState(Src value, RangeConstraint validity) | |
| 362 : value_(value), | |
| 363 validity_(GetRangeConstraint(validity | | |
| 364 DstRangeRelationToSrcRange<T>(value))) { | |
| 365 COMPILE_ASSERT(std::numeric_limits<Src>::is_specialized, | |
| 366 argument_must_be_numeric); | |
| 367 } | |
| 368 | |
| 369 // Copy constructor. | |
| 370 template <typename Src> | |
| 371 CheckedNumericState(const CheckedNumericState<Src>& rhs) | |
| 372 : value_(static_cast<T>(rhs.value())), | |
| 373 validity_(GetRangeConstraint( | |
| 374 rhs.validity() | DstRangeRelationToSrcRange<T>(rhs.value()))) {} | |
| 375 | |
| 376 template <typename Src> | |
| 377 explicit CheckedNumericState( | |
| 378 Src value, | |
| 379 typename enable_if<std::numeric_limits<Src>::is_specialized, int>::type = | |
| 380 0) | |
| 381 : value_(static_cast<T>(value)), | |
| 382 validity_(DstRangeRelationToSrcRange<T>(value)) {} | |
| 383 | |
| 384 RangeConstraint validity() const { return validity_; } | |
| 385 T value() const { return value_; } | |
| 386 }; | |
| 387 | |
| 388 // Floating points maintain their own validity, but need translation wrappers. | |
| 389 template <typename T> | |
| 390 class CheckedNumericState<T, NUMERIC_FLOATING> { | |
| 391 private: | |
| 392 T value_; | |
| 393 | |
| 394 public: | |
| 395 template <typename Src, NumericRepresentation type> | |
| 396 friend class CheckedNumericState; | |
| 397 | |
| 398 CheckedNumericState() : value_(0.0) {} | |
| 399 | |
| 400 template <typename Src> | |
| 401 CheckedNumericState( | |
| 402 Src value, | |
| 403 RangeConstraint validity, | |
| 404 typename enable_if<std::numeric_limits<Src>::is_integer, int>::type = 0) { | |
| 405 switch (DstRangeRelationToSrcRange<T>(value)) { | |
| 406 case RANGE_VALID: | |
| 407 value_ = static_cast<T>(value); | |
| 408 break; | |
| 409 | |
| 410 case RANGE_UNDERFLOW: | |
| 411 value_ = -std::numeric_limits<T>::infinity(); | |
| 412 break; | |
| 413 | |
| 414 case RANGE_OVERFLOW: | |
| 415 value_ = std::numeric_limits<T>::infinity(); | |
| 416 break; | |
| 417 | |
| 418 case RANGE_INVALID: | |
| 419 value_ = std::numeric_limits<T>::quiet_NaN(); | |
| 420 break; | |
| 421 | |
| 422 default: | |
| 423 NOTREACHED(); | |
| 424 } | |
| 425 } | |
| 426 | |
| 427 template <typename Src> | |
| 428 explicit CheckedNumericState( | |
| 429 Src value, | |
| 430 typename enable_if<std::numeric_limits<Src>::is_specialized, int>::type = | |
| 431 0) | |
| 432 : value_(static_cast<T>(value)) {} | |
| 433 | |
| 434 // Copy constructor. | |
| 435 template <typename Src> | |
| 436 CheckedNumericState(const CheckedNumericState<Src>& rhs) | |
| 437 : value_(static_cast<T>(rhs.value())) {} | |
| 438 | |
| 439 RangeConstraint validity() const { | |
| 440 return GetRangeConstraint(value_ <= std::numeric_limits<T>::max(), | |
| 441 value_ >= -std::numeric_limits<T>::max()); | |
| 442 } | |
| 443 T value() const { return value_; } | |
| 444 }; | |
| 445 | |
| 446 // For integers less than 128-bit and floats 32-bit or larger, we can distil | |
| 447 // C/C++ arithmetic promotions down to two simple rules: | |
| 448 // 1. The type with the larger maximum exponent always takes precedence. | |
| 449 // 2. The resulting type must be promoted to at least an int. | |
| 450 // The following template specializations implement that promotion logic. | |
| 451 enum ArithmeticPromotionCategory { | |
| 452 LEFT_PROMOTION, | |
| 453 RIGHT_PROMOTION, | |
| 454 DEFAULT_PROMOTION | |
| 455 }; | |
| 456 | |
| 457 template <typename Lhs, | |
| 458 typename Rhs = Lhs, | |
| 459 ArithmeticPromotionCategory Promotion = | |
| 460 (MaxExponent<Lhs>::value > MaxExponent<Rhs>::value) | |
| 461 ? (MaxExponent<Lhs>::value > MaxExponent<int>::value | |
| 462 ? LEFT_PROMOTION | |
| 463 : DEFAULT_PROMOTION) | |
| 464 : (MaxExponent<Rhs>::value > MaxExponent<int>::value | |
| 465 ? RIGHT_PROMOTION | |
| 466 : DEFAULT_PROMOTION) > | |
| 467 struct ArithmeticPromotion; | |
| 468 | |
| 469 template <typename Lhs, typename Rhs> | |
| 470 struct ArithmeticPromotion<Lhs, Rhs, LEFT_PROMOTION> { | |
| 471 typedef Lhs type; | |
| 472 }; | |
| 473 | |
| 474 template <typename Lhs, typename Rhs> | |
| 475 struct ArithmeticPromotion<Lhs, Rhs, RIGHT_PROMOTION> { | |
| 476 typedef Rhs type; | |
| 477 }; | |
| 478 | |
| 479 template <typename Lhs, typename Rhs> | |
| 480 struct ArithmeticPromotion<Lhs, Rhs, DEFAULT_PROMOTION> { | |
| 481 typedef int type; | |
| 482 }; | |
| 483 | |
| 484 // We can statically check if operations on the provided types can wrap, so we | |
| 485 // can skip the checked operations if they're not needed. So, for an integer we | |
| 486 // care if the destination type preserves the sign and is twice the width of | |
| 487 // the source. | |
| 488 template <typename T, typename Lhs, typename Rhs> | |
| 489 struct IsIntegerArithmeticSafe { | |
| 490 static const bool value = !std::numeric_limits<T>::is_iec559 && | |
| 491 StaticDstRangeRelationToSrcRange<T, Lhs>::value == | |
| 492 NUMERIC_RANGE_CONTAINED && | |
| 493 sizeof(T) >= (2 * sizeof(Lhs)) && | |
| 494 StaticDstRangeRelationToSrcRange<T, Rhs>::value != | |
| 495 NUMERIC_RANGE_CONTAINED && | |
| 496 sizeof(T) >= (2 * sizeof(Rhs)); | |
| 497 }; | |
| 498 | |
| 499 } // namespace internal | |
| 500 } // namespace base | |
| 501 | |
| 502 #endif // SAFE_MATH_IMPL_H_ | |
| OLD | NEW |