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 // Slightly adapted for inclusion in V8. |
| 6 // Copyright 2014 the V8 project authors. All rights reserved. |
| 7 |
| 8 #ifndef BASE_SAFE_MATH_IMPL_H_ |
| 9 #define BASE_SAFE_MATH_IMPL_H_ |
| 10 |
| 11 #include <stdint.h> |
| 12 |
| 13 #include <cmath> |
| 14 #include <cstdlib> |
| 15 #include <limits> |
| 16 |
| 17 #include "src/base/macros.h" |
| 18 #include "src/base/safe_conversions.h" |
| 19 |
| 20 namespace v8 { |
| 21 namespace internal { |
| 22 |
| 23 |
| 24 // From Chromium's base/template_util.h: |
| 25 |
| 26 template<class T, T v> |
| 27 struct integral_constant { |
| 28 static const T value = v; |
| 29 typedef T value_type; |
| 30 typedef integral_constant<T, v> type; |
| 31 }; |
| 32 |
| 33 template <class T, T v> const T integral_constant<T, v>::value; |
| 34 |
| 35 typedef integral_constant<bool, true> true_type; |
| 36 typedef integral_constant<bool, false> false_type; |
| 37 |
| 38 template <class T, class U> struct is_same : public false_type {}; |
| 39 template <class T> struct is_same<T, T> : true_type {}; |
| 40 |
| 41 template<bool B, class T = void> |
| 42 struct enable_if {}; |
| 43 |
| 44 template<class T> |
| 45 struct enable_if<true, T> { typedef T type; }; |
| 46 |
| 47 // </template_util.h> |
| 48 |
| 49 |
| 50 // Everything from here up to the floating point operations is portable C++, |
| 51 // but it may not be fast. This code could be split based on |
| 52 // platform/architecture and replaced with potentially faster implementations. |
| 53 |
| 54 // Integer promotion templates used by the portable checked integer arithmetic. |
| 55 template <size_t Size, bool IsSigned> |
| 56 struct IntegerForSizeAndSign; |
| 57 template <> |
| 58 struct IntegerForSizeAndSign<1, true> { |
| 59 typedef int8_t type; |
| 60 }; |
| 61 template <> |
| 62 struct IntegerForSizeAndSign<1, false> { |
| 63 typedef uint8_t type; |
| 64 }; |
| 65 template <> |
| 66 struct IntegerForSizeAndSign<2, true> { |
| 67 typedef int16_t type; |
| 68 }; |
| 69 template <> |
| 70 struct IntegerForSizeAndSign<2, false> { |
| 71 typedef uint16_t type; |
| 72 }; |
| 73 template <> |
| 74 struct IntegerForSizeAndSign<4, true> { |
| 75 typedef int32_t type; |
| 76 }; |
| 77 template <> |
| 78 struct IntegerForSizeAndSign<4, false> { |
| 79 typedef uint32_t type; |
| 80 }; |
| 81 template <> |
| 82 struct IntegerForSizeAndSign<8, true> { |
| 83 typedef int64_t type; |
| 84 }; |
| 85 template <> |
| 86 struct IntegerForSizeAndSign<8, false> { |
| 87 typedef uint64_t type; |
| 88 }; |
| 89 |
| 90 // WARNING: We have no IntegerForSizeAndSign<16, *>. If we ever add one to |
| 91 // support 128-bit math, then the ArithmeticPromotion template below will need |
| 92 // to be updated (or more likely replaced with a decltype expression). |
| 93 |
| 94 template <typename Integer> |
| 95 struct UnsignedIntegerForSize { |
| 96 typedef typename enable_if< |
| 97 std::numeric_limits<Integer>::is_integer, |
| 98 typename IntegerForSizeAndSign<sizeof(Integer), false>::type>::type type; |
| 99 }; |
| 100 |
| 101 template <typename Integer> |
| 102 struct SignedIntegerForSize { |
| 103 typedef typename enable_if< |
| 104 std::numeric_limits<Integer>::is_integer, |
| 105 typename IntegerForSizeAndSign<sizeof(Integer), true>::type>::type type; |
| 106 }; |
| 107 |
| 108 template <typename Integer> |
| 109 struct TwiceWiderInteger { |
| 110 typedef typename enable_if< |
| 111 std::numeric_limits<Integer>::is_integer, |
| 112 typename IntegerForSizeAndSign< |
| 113 sizeof(Integer) * 2, |
| 114 std::numeric_limits<Integer>::is_signed>::type>::type type; |
| 115 }; |
| 116 |
| 117 template <typename Integer> |
| 118 struct PositionOfSignBit { |
| 119 static const typename enable_if<std::numeric_limits<Integer>::is_integer, |
| 120 size_t>::type value = 8 * sizeof(Integer) - 1; |
| 121 }; |
| 122 |
| 123 // Helper templates for integer manipulations. |
| 124 |
| 125 template <typename T> |
| 126 bool HasSignBit(T x) { |
| 127 // Cast to unsigned since right shift on signed is undefined. |
| 128 return !!(static_cast<typename UnsignedIntegerForSize<T>::type>(x) >> |
| 129 PositionOfSignBit<T>::value); |
| 130 } |
| 131 |
| 132 // This wrapper undoes the standard integer promotions. |
| 133 template <typename T> |
| 134 T BinaryComplement(T x) { |
| 135 return ~x; |
| 136 } |
| 137 |
| 138 // Here are the actual portable checked integer math implementations. |
| 139 // TODO(jschuh): Break this code out from the enable_if pattern and find a clean |
| 140 // way to coalesce things into the CheckedNumericState specializations below. |
| 141 |
| 142 template <typename T> |
| 143 typename enable_if<std::numeric_limits<T>::is_integer, T>::type |
| 144 CheckedAdd(T x, T y, RangeConstraint* validity) { |
| 145 // Since the value of x+y is undefined if we have a signed type, we compute |
| 146 // it using the unsigned type of the same size. |
| 147 typedef typename UnsignedIntegerForSize<T>::type UnsignedDst; |
| 148 UnsignedDst ux = static_cast<UnsignedDst>(x); |
| 149 UnsignedDst uy = static_cast<UnsignedDst>(y); |
| 150 UnsignedDst uresult = ux + uy; |
| 151 // Addition is valid if the sign of (x + y) is equal to either that of x or |
| 152 // that of y. |
| 153 if (std::numeric_limits<T>::is_signed) { |
| 154 if (HasSignBit(BinaryComplement((uresult ^ ux) & (uresult ^ uy)))) |
| 155 *validity = RANGE_VALID; |
| 156 else // Direction of wrap is inverse of result sign. |
| 157 *validity = HasSignBit(uresult) ? RANGE_OVERFLOW : RANGE_UNDERFLOW; |
| 158 |
| 159 } else { // Unsigned is either valid or overflow. |
| 160 *validity = BinaryComplement(x) >= y ? RANGE_VALID : RANGE_OVERFLOW; |
| 161 } |
| 162 return static_cast<T>(uresult); |
| 163 } |
| 164 |
| 165 template <typename T> |
| 166 typename enable_if<std::numeric_limits<T>::is_integer, T>::type |
| 167 CheckedSub(T x, T y, RangeConstraint* validity) { |
| 168 // Since the value of x+y is undefined if we have a signed type, we compute |
| 169 // it using the unsigned type of the same size. |
| 170 typedef typename UnsignedIntegerForSize<T>::type UnsignedDst; |
| 171 UnsignedDst ux = static_cast<UnsignedDst>(x); |
| 172 UnsignedDst uy = static_cast<UnsignedDst>(y); |
| 173 UnsignedDst uresult = ux - uy; |
| 174 // Subtraction is valid if either x and y have same sign, or (x-y) and x have |
| 175 // the same sign. |
| 176 if (std::numeric_limits<T>::is_signed) { |
| 177 if (HasSignBit(BinaryComplement((uresult ^ ux) & (ux ^ uy)))) |
| 178 *validity = RANGE_VALID; |
| 179 else // Direction of wrap is inverse of result sign. |
| 180 *validity = HasSignBit(uresult) ? RANGE_OVERFLOW : RANGE_UNDERFLOW; |
| 181 |
| 182 } else { // Unsigned is either valid or underflow. |
| 183 *validity = x >= y ? RANGE_VALID : RANGE_UNDERFLOW; |
| 184 } |
| 185 return static_cast<T>(uresult); |
| 186 } |
| 187 |
| 188 // Integer multiplication is a bit complicated. In the fast case we just |
| 189 // we just promote to a twice wider type, and range check the result. In the |
| 190 // slow case we need to manually check that the result won't be truncated by |
| 191 // checking with division against the appropriate bound. |
| 192 template <typename T> |
| 193 typename enable_if< |
| 194 std::numeric_limits<T>::is_integer && sizeof(T) * 2 <= sizeof(uintmax_t), |
| 195 T>::type |
| 196 CheckedMul(T x, T y, RangeConstraint* validity) { |
| 197 typedef typename TwiceWiderInteger<T>::type IntermediateType; |
| 198 IntermediateType tmp = |
| 199 static_cast<IntermediateType>(x) * static_cast<IntermediateType>(y); |
| 200 *validity = DstRangeRelationToSrcRange<T>(tmp); |
| 201 return static_cast<T>(tmp); |
| 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 // if either side is zero then the result will be zero. |
| 211 if (!(x || y)) { |
| 212 return RANGE_VALID; |
| 213 |
| 214 } else if (x > 0) { |
| 215 if (y > 0) |
| 216 *validity = |
| 217 x <= std::numeric_limits<T>::max() / y ? RANGE_VALID : RANGE_OVERFLOW; |
| 218 else |
| 219 *validity = y >= std::numeric_limits<T>::min() / x ? RANGE_VALID |
| 220 : RANGE_UNDERFLOW; |
| 221 |
| 222 } else { |
| 223 if (y > 0) |
| 224 *validity = x >= std::numeric_limits<T>::min() / y ? RANGE_VALID |
| 225 : RANGE_UNDERFLOW; |
| 226 else |
| 227 *validity = |
| 228 y >= std::numeric_limits<T>::max() / x ? RANGE_VALID : RANGE_OVERFLOW; |
| 229 } |
| 230 |
| 231 return x * y; |
| 232 } |
| 233 |
| 234 template <typename T> |
| 235 typename enable_if<std::numeric_limits<T>::is_integer && |
| 236 !std::numeric_limits<T>::is_signed && |
| 237 (sizeof(T) * 2 > sizeof(uintmax_t)), |
| 238 T>::type |
| 239 CheckedMul(T x, T y, RangeConstraint* validity) { |
| 240 *validity = (y == 0 || x <= std::numeric_limits<T>::max() / y) |
| 241 ? RANGE_VALID |
| 242 : RANGE_OVERFLOW; |
| 243 return x * y; |
| 244 } |
| 245 |
| 246 // Division just requires a check for an invalid negation on signed min/-1. |
| 247 template <typename T> |
| 248 T CheckedDiv( |
| 249 T x, |
| 250 T y, |
| 251 RangeConstraint* validity, |
| 252 typename enable_if<std::numeric_limits<T>::is_integer, int>::type = 0) { |
| 253 if (std::numeric_limits<T>::is_signed && x == std::numeric_limits<T>::min() && |
| 254 y == static_cast<T>(-1)) { |
| 255 *validity = RANGE_OVERFLOW; |
| 256 return std::numeric_limits<T>::min(); |
| 257 } |
| 258 |
| 259 *validity = RANGE_VALID; |
| 260 return x / y; |
| 261 } |
| 262 |
| 263 template <typename T> |
| 264 typename enable_if< |
| 265 std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 266 T>::type |
| 267 CheckedMod(T x, T y, RangeConstraint* validity) { |
| 268 *validity = y > 0 ? RANGE_VALID : RANGE_INVALID; |
| 269 return x % y; |
| 270 } |
| 271 |
| 272 template <typename T> |
| 273 typename enable_if< |
| 274 std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 275 T>::type |
| 276 CheckedMod(T x, T y, RangeConstraint* validity) { |
| 277 *validity = RANGE_VALID; |
| 278 return x % y; |
| 279 } |
| 280 |
| 281 template <typename T> |
| 282 typename enable_if< |
| 283 std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 284 T>::type |
| 285 CheckedNeg(T value, RangeConstraint* validity) { |
| 286 *validity = |
| 287 value != std::numeric_limits<T>::min() ? RANGE_VALID : RANGE_OVERFLOW; |
| 288 // The negation of signed min is min, so catch that one. |
| 289 return -value; |
| 290 } |
| 291 |
| 292 template <typename T> |
| 293 typename enable_if< |
| 294 std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 295 T>::type |
| 296 CheckedNeg(T value, RangeConstraint* validity) { |
| 297 // The only legal unsigned negation is zero. |
| 298 *validity = value ? RANGE_UNDERFLOW : RANGE_VALID; |
| 299 return static_cast<T>( |
| 300 -static_cast<typename SignedIntegerForSize<T>::type>(value)); |
| 301 } |
| 302 |
| 303 template <typename T> |
| 304 typename enable_if< |
| 305 std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 306 T>::type |
| 307 CheckedAbs(T value, RangeConstraint* validity) { |
| 308 *validity = |
| 309 value != std::numeric_limits<T>::min() ? RANGE_VALID : RANGE_OVERFLOW; |
| 310 return std::abs(value); |
| 311 } |
| 312 |
| 313 template <typename T> |
| 314 typename enable_if< |
| 315 std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 316 T>::type |
| 317 CheckedAbs(T value, RangeConstraint* validity) { |
| 318 // Absolute value of a positive is just its identiy. |
| 319 *validity = RANGE_VALID; |
| 320 return value; |
| 321 } |
| 322 |
| 323 // These are the floating point stubs that the compiler needs to see. Only the |
| 324 // negation operation is ever called. |
| 325 #define BASE_FLOAT_ARITHMETIC_STUBS(NAME) \ |
| 326 template <typename T> \ |
| 327 typename enable_if<std::numeric_limits<T>::is_iec559, T>::type \ |
| 328 Checked##NAME(T, T, RangeConstraint*) { \ |
| 329 UNREACHABLE(); \ |
| 330 return 0; \ |
| 331 } |
| 332 |
| 333 BASE_FLOAT_ARITHMETIC_STUBS(Add) |
| 334 BASE_FLOAT_ARITHMETIC_STUBS(Sub) |
| 335 BASE_FLOAT_ARITHMETIC_STUBS(Mul) |
| 336 BASE_FLOAT_ARITHMETIC_STUBS(Div) |
| 337 BASE_FLOAT_ARITHMETIC_STUBS(Mod) |
| 338 |
| 339 #undef BASE_FLOAT_ARITHMETIC_STUBS |
| 340 |
| 341 template <typename T> |
| 342 typename enable_if<std::numeric_limits<T>::is_iec559, T>::type CheckedNeg( |
| 343 T value, |
| 344 RangeConstraint*) { |
| 345 return -value; |
| 346 } |
| 347 |
| 348 template <typename T> |
| 349 typename enable_if<std::numeric_limits<T>::is_iec559, T>::type CheckedAbs( |
| 350 T value, |
| 351 RangeConstraint*) { |
| 352 return std::abs(value); |
| 353 } |
| 354 |
| 355 // Floats carry around their validity state with them, but integers do not. So, |
| 356 // we wrap the underlying value in a specialization in order to hide that detail |
| 357 // and expose an interface via accessors. |
| 358 enum NumericRepresentation { |
| 359 NUMERIC_INTEGER, |
| 360 NUMERIC_FLOATING, |
| 361 NUMERIC_UNKNOWN |
| 362 }; |
| 363 |
| 364 template <typename NumericType> |
| 365 struct GetNumericRepresentation { |
| 366 static const NumericRepresentation value = |
| 367 std::numeric_limits<NumericType>::is_integer |
| 368 ? NUMERIC_INTEGER |
| 369 : (std::numeric_limits<NumericType>::is_iec559 ? NUMERIC_FLOATING |
| 370 : NUMERIC_UNKNOWN); |
| 371 }; |
| 372 |
| 373 template <typename T, NumericRepresentation type = |
| 374 GetNumericRepresentation<T>::value> |
| 375 class CheckedNumericState {}; |
| 376 |
| 377 // Integrals require quite a bit of additional housekeeping to manage state. |
| 378 template <typename T> |
| 379 class CheckedNumericState<T, NUMERIC_INTEGER> { |
| 380 private: |
| 381 T value_; |
| 382 RangeConstraint validity_; |
| 383 |
| 384 public: |
| 385 template <typename Src, NumericRepresentation type> |
| 386 friend class CheckedNumericState; |
| 387 |
| 388 CheckedNumericState() : value_(0), validity_(RANGE_VALID) {} |
| 389 |
| 390 template <typename Src> |
| 391 CheckedNumericState(Src value, RangeConstraint validity) |
| 392 : value_(value), |
| 393 validity_(GetRangeConstraint(validity | |
| 394 DstRangeRelationToSrcRange<T>(value))) { |
| 395 // Argument must be numeric. |
| 396 STATIC_ASSERT(std::numeric_limits<Src>::is_specialized); |
| 397 } |
| 398 |
| 399 // Copy constructor. |
| 400 template <typename Src> |
| 401 CheckedNumericState(const CheckedNumericState<Src>& rhs) |
| 402 : value_(static_cast<T>(rhs.value())), |
| 403 validity_(GetRangeConstraint( |
| 404 rhs.validity() | DstRangeRelationToSrcRange<T>(rhs.value()))) {} |
| 405 |
| 406 template <typename Src> |
| 407 explicit CheckedNumericState( |
| 408 Src value, |
| 409 typename enable_if<std::numeric_limits<Src>::is_specialized, int>::type = |
| 410 0) |
| 411 : value_(static_cast<T>(value)), |
| 412 validity_(DstRangeRelationToSrcRange<T>(value)) {} |
| 413 |
| 414 RangeConstraint validity() const { return validity_; } |
| 415 T value() const { return value_; } |
| 416 }; |
| 417 |
| 418 // Floating points maintain their own validity, but need translation wrappers. |
| 419 template <typename T> |
| 420 class CheckedNumericState<T, NUMERIC_FLOATING> { |
| 421 private: |
| 422 T value_; |
| 423 |
| 424 public: |
| 425 template <typename Src, NumericRepresentation type> |
| 426 friend class CheckedNumericState; |
| 427 |
| 428 CheckedNumericState() : value_(0.0) {} |
| 429 |
| 430 template <typename Src> |
| 431 CheckedNumericState( |
| 432 Src value, |
| 433 RangeConstraint validity, |
| 434 typename enable_if<std::numeric_limits<Src>::is_integer, int>::type = 0) { |
| 435 switch (DstRangeRelationToSrcRange<T>(value)) { |
| 436 case RANGE_VALID: |
| 437 value_ = static_cast<T>(value); |
| 438 break; |
| 439 |
| 440 case RANGE_UNDERFLOW: |
| 441 value_ = -std::numeric_limits<T>::infinity(); |
| 442 break; |
| 443 |
| 444 case RANGE_OVERFLOW: |
| 445 value_ = std::numeric_limits<T>::infinity(); |
| 446 break; |
| 447 |
| 448 case RANGE_INVALID: |
| 449 value_ = std::numeric_limits<T>::quiet_NaN(); |
| 450 break; |
| 451 } |
| 452 } |
| 453 |
| 454 template <typename Src> |
| 455 explicit CheckedNumericState( |
| 456 Src value, |
| 457 typename enable_if<std::numeric_limits<Src>::is_specialized, int>::type = |
| 458 0) |
| 459 : value_(static_cast<T>(value)) {} |
| 460 |
| 461 // Copy constructor. |
| 462 template <typename Src> |
| 463 CheckedNumericState(const CheckedNumericState<Src>& rhs) |
| 464 : value_(static_cast<T>(rhs.value())) {} |
| 465 |
| 466 RangeConstraint validity() const { |
| 467 return GetRangeConstraint(value_ <= std::numeric_limits<T>::max(), |
| 468 value_ >= -std::numeric_limits<T>::max()); |
| 469 } |
| 470 T value() const { return value_; } |
| 471 }; |
| 472 |
| 473 // For integers less than 128-bit and floats 32-bit or larger, we can distil |
| 474 // C/C++ arithmetic promotions down to two simple rules: |
| 475 // 1. The type with the larger maximum exponent always takes precedence. |
| 476 // 2. The resulting type must be promoted to at least an int. |
| 477 // The following template specializations implement that promotion logic. |
| 478 enum ArithmeticPromotionCategory { |
| 479 LEFT_PROMOTION, |
| 480 RIGHT_PROMOTION, |
| 481 DEFAULT_PROMOTION |
| 482 }; |
| 483 |
| 484 template <typename Lhs, |
| 485 typename Rhs = Lhs, |
| 486 ArithmeticPromotionCategory Promotion = |
| 487 (MaxExponent<Lhs>::value > MaxExponent<Rhs>::value) |
| 488 ? (MaxExponent<Lhs>::value > MaxExponent<int>::value |
| 489 ? LEFT_PROMOTION |
| 490 : DEFAULT_PROMOTION) |
| 491 : (MaxExponent<Rhs>::value > MaxExponent<int>::value |
| 492 ? RIGHT_PROMOTION |
| 493 : DEFAULT_PROMOTION) > |
| 494 struct ArithmeticPromotion; |
| 495 |
| 496 template <typename Lhs, typename Rhs> |
| 497 struct ArithmeticPromotion<Lhs, Rhs, LEFT_PROMOTION> { |
| 498 typedef Lhs type; |
| 499 }; |
| 500 |
| 501 template <typename Lhs, typename Rhs> |
| 502 struct ArithmeticPromotion<Lhs, Rhs, RIGHT_PROMOTION> { |
| 503 typedef Rhs type; |
| 504 }; |
| 505 |
| 506 template <typename Lhs, typename Rhs> |
| 507 struct ArithmeticPromotion<Lhs, Rhs, DEFAULT_PROMOTION> { |
| 508 typedef int type; |
| 509 }; |
| 510 |
| 511 // We can statically check if operations on the provided types can wrap, so we |
| 512 // can skip the checked operations if they're not needed. So, for an integer we |
| 513 // care if the destination type preserves the sign and is twice the width of |
| 514 // the source. |
| 515 template <typename T, typename Lhs, typename Rhs> |
| 516 struct IsIntegerArithmeticSafe { |
| 517 static const bool value = !std::numeric_limits<T>::is_iec559 && |
| 518 StaticDstRangeRelationToSrcRange<T, Lhs>::value == |
| 519 NUMERIC_RANGE_CONTAINED && |
| 520 sizeof(T) >= (2 * sizeof(Lhs)) && |
| 521 StaticDstRangeRelationToSrcRange<T, Rhs>::value != |
| 522 NUMERIC_RANGE_CONTAINED && |
| 523 sizeof(T) >= (2 * sizeof(Rhs)); |
| 524 }; |
| 525 |
| 526 } // namespace internal |
| 527 } // namespace v8 |
| 528 |
| 529 #endif // BASE_SAFE_MATH_IMPL_H_ |
OLD | NEW |