| 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 ANGLEBASE_NUMERICS_SAFE_MATH_IMPL_H_ |
| 6 | #define ANGLEBASE_NUMERICS_SAFE_MATH_IMPL_H_ |
| 7 | |
| 8 | #include <stddef.h> |
| 9 | #include <stdint.h> |
| 10 | |
| 11 | #include <climits> |
| 12 | #include <cmath> |
| 13 | #include <cstdlib> |
| 14 | #include <limits> |
| 15 | #include <type_traits> |
| 16 | |
| 17 | #include "anglebase/numerics/safe_conversions.h" |
| 18 | |
| 19 | namespace angle |
| 20 | { |
| 21 | |
| 22 | namespace base |
| 23 | { |
| 24 | namespace internal |
| 25 | { |
| 26 | |
| 27 | // Everything from here up to the floating point operations is portable C++, |
| 28 | // but it may not be fast. This code could be split based on |
| 29 | // platform/architecture and replaced with potentially faster implementations. |
| 30 | |
| 31 | // Integer promotion templates used by the portable checked integer arithmetic. |
| 32 | template <size_t Size, bool IsSigned> |
| 33 | struct IntegerForSizeAndSign; |
| 34 | template <> |
| 35 | struct IntegerForSizeAndSign<1, true> |
| 36 | { |
| 37 | typedef int8_t type; |
| 38 | }; |
| 39 | template <> |
| 40 | struct IntegerForSizeAndSign<1, false> |
| 41 | { |
| 42 | typedef uint8_t type; |
| 43 | }; |
| 44 | template <> |
| 45 | struct IntegerForSizeAndSign<2, true> |
| 46 | { |
| 47 | typedef int16_t type; |
| 48 | }; |
| 49 | template <> |
| 50 | struct IntegerForSizeAndSign<2, false> |
| 51 | { |
| 52 | typedef uint16_t type; |
| 53 | }; |
| 54 | template <> |
| 55 | struct IntegerForSizeAndSign<4, true> |
| 56 | { |
| 57 | typedef int32_t type; |
| 58 | }; |
| 59 | template <> |
| 60 | struct IntegerForSizeAndSign<4, false> |
| 61 | { |
| 62 | typedef uint32_t type; |
| 63 | }; |
| 64 | template <> |
| 65 | struct IntegerForSizeAndSign<8, true> |
| 66 | { |
| 67 | typedef int64_t type; |
| 68 | }; |
| 69 | template <> |
| 70 | struct IntegerForSizeAndSign<8, false> |
| 71 | { |
| 72 | typedef uint64_t type; |
| 73 | }; |
| 74 | |
| 75 | // WARNING: We have no IntegerForSizeAndSign<16, *>. If we ever add one to |
| 76 | // support 128-bit math, then the ArithmeticPromotion template below will need |
| 77 | // to be updated (or more likely replaced with a decltype expression). |
| 78 | |
| 79 | template <typename Integer> |
| 80 | struct UnsignedIntegerForSize |
| 81 | { |
| 82 | typedef |
| 83 | typename std::enable_if<std::numeric_limits<Integer>::is_integer, |
| 84 | typename IntegerForSizeAndSign<sizeof(Integer), false>::type>::type |
| 85 | type; |
| 86 | }; |
| 87 | |
| 88 | template <typename Integer> |
| 89 | struct SignedIntegerForSize |
| 90 | { |
| 91 | typedef |
| 92 | typename std::enable_if<std::numeric_limits<Integer>::is_integer, |
| 93 | typename IntegerForSizeAndSign<sizeof(Integer), true>::type>::type |
| 94 | type; |
| 95 | }; |
| 96 | |
| 97 | template <typename Integer> |
| 98 | struct TwiceWiderInteger |
| 99 | { |
| 100 | typedef typename std::enable_if< |
| 101 | std::numeric_limits<Integer>::is_integer, |
| 102 | typename IntegerForSizeAndSign<sizeof(Integer) * 2, |
| 103 | std::numeric_limits<Integer>::is_signed>::type>::type type; |
| 104 | }; |
| 105 | |
| 106 | template <typename Integer> |
| 107 | struct PositionOfSignBit |
| 108 | { |
| 109 | static const typename std::enable_if<std::numeric_limits<Integer>::is_integer, size_t>::type |
| 110 | value = CHAR_BIT * sizeof(Integer) - 1; |
| 111 | }; |
| 112 | |
| 113 | // This is used for UnsignedAbs, where we need to support floating-point |
| 114 | // template instantiations even though we don't actually support the operations. |
| 115 | // However, there is no corresponding implementation of e.g. CheckedUnsignedAbs, |
| 116 | // so the float versions will not compile. |
| 117 | template <typename Numeric, |
| 118 | bool IsInteger = std::numeric_limits<Numeric>::is_integer, |
| 119 | bool IsFloat = std::numeric_limits<Numeric>::is_iec559> |
| 120 | struct UnsignedOrFloatForSize; |
| 121 | |
| 122 | template <typename Numeric> |
| 123 | struct UnsignedOrFloatForSize<Numeric, true, false> |
| 124 | { |
| 125 | typedef typename UnsignedIntegerForSize<Numeric>::type type; |
| 126 | }; |
| 127 | |
| 128 | template <typename Numeric> |
| 129 | struct UnsignedOrFloatForSize<Numeric, false, true> |
| 130 | { |
| 131 | typedef Numeric type; |
| 132 | }; |
| 133 | |
| 134 | // Helper templates for integer manipulations. |
| 135 | |
| 136 | template <typename T> |
| 137 | constexpr bool HasSignBit(T x) |
| 138 | { |
| 139 | // Cast to unsigned since right shift on signed is undefined. |
| 140 | return !!(static_cast<typename UnsignedIntegerForSize<T>::type>(x) >> |
| 141 | PositionOfSignBit<T>::value); |
| 142 | } |
| 143 | |
| 144 | // This wrapper undoes the standard integer promotions. |
| 145 | template <typename T> |
| 146 | constexpr T BinaryComplement(T x) |
| 147 | { |
| 148 | return static_cast<T>(~x); |
| 149 | } |
| 150 | |
| 151 | // Here are the actual portable checked integer math implementations. |
| 152 | // TODO(jschuh): Break this code out from the enable_if pattern and find a clean |
| 153 | // way to coalesce things into the CheckedNumericState specializations below. |
| 154 | |
| 155 | template <typename T> |
| 156 | typename std::enable_if<std::numeric_limits<T>::is_integer, T>::type |
| 157 | CheckedAdd(T x, T y, RangeConstraint *validity) |
| 158 | { |
| 159 | // Since the value of x+y is undefined if we have a signed type, we compute |
| 160 | // it using the unsigned type of the same size. |
| 161 | typedef typename UnsignedIntegerForSize<T>::type UnsignedDst; |
| 162 | UnsignedDst ux = static_cast<UnsignedDst>(x); |
| 163 | UnsignedDst uy = static_cast<UnsignedDst>(y); |
| 164 | UnsignedDst uresult = static_cast<UnsignedDst>(ux + uy); |
| 165 | // Addition is valid if the sign of (x + y) is equal to either that of x or |
| 166 | // that of y. |
| 167 | if (std::numeric_limits<T>::is_signed) |
| 168 | { |
| 169 | if (HasSignBit(BinaryComplement(static_cast<UnsignedDst>((uresult ^ ux) & (uresult ^ uy))))) |
| 170 | { |
| 171 | *validity = RANGE_VALID; |
| 172 | } |
| 173 | else |
| 174 | { // Direction of wrap is inverse of result sign. |
| 175 | *validity = HasSignBit(uresult) ? RANGE_OVERFLOW : RANGE_UNDERFLOW; |
| 176 | } |
| 177 | } |
| 178 | else |
| 179 | { // Unsigned is either valid or overflow. |
| 180 | *validity = BinaryComplement(x) >= y ? RANGE_VALID : RANGE_OVERFLOW; |
| 181 | } |
| 182 | return static_cast<T>(uresult); |
| 183 | } |
| 184 | |
| 185 | template <typename T> |
| 186 | typename std::enable_if<std::numeric_limits<T>::is_integer, T>::type |
| 187 | CheckedSub(T x, T y, RangeConstraint *validity) |
| 188 | { |
| 189 | // Since the value of x+y is undefined if we have a signed type, we compute |
| 190 | // it using the unsigned type of the same size. |
| 191 | typedef typename UnsignedIntegerForSize<T>::type UnsignedDst; |
| 192 | UnsignedDst ux = static_cast<UnsignedDst>(x); |
| 193 | UnsignedDst uy = static_cast<UnsignedDst>(y); |
| 194 | UnsignedDst uresult = static_cast<UnsignedDst>(ux - uy); |
| 195 | // Subtraction is valid if either x and y have same sign, or (x-y) and x have |
| 196 | // the same sign. |
| 197 | if (std::numeric_limits<T>::is_signed) |
| 198 | { |
| 199 | if (HasSignBit(BinaryComplement(static_cast<UnsignedDst>((uresult ^ ux) & (ux ^ uy))))) |
| 200 | { |
| 201 | *validity = RANGE_VALID; |
| 202 | } |
| 203 | else |
| 204 | { // Direction of wrap is inverse of result sign. |
| 205 | *validity = HasSignBit(uresult) ? RANGE_OVERFLOW : RANGE_UNDERFLOW; |
| 206 | } |
| 207 | } |
| 208 | else |
| 209 | { // Unsigned is either valid or underflow. |
| 210 | *validity = x >= y ? RANGE_VALID : RANGE_UNDERFLOW; |
| 211 | } |
| 212 | return static_cast<T>(uresult); |
| 213 | } |
| 214 | |
| 215 | // Integer multiplication is a bit complicated. In the fast case we just |
| 216 | // we just promote to a twice wider type, and range check the result. In the |
| 217 | // slow case we need to manually check that the result won't be truncated by |
| 218 | // checking with division against the appropriate bound. |
| 219 | template <typename T> |
| 220 | typename std::enable_if<std::numeric_limits<T>::is_integer && sizeof(T) * 2 <= sizeof(uintmax_t), |
| 221 | T>::type |
| 222 | CheckedMul(T x, T y, RangeConstraint *validity) |
| 223 | { |
| 224 | typedef typename TwiceWiderInteger<T>::type IntermediateType; |
| 225 | IntermediateType tmp = static_cast<IntermediateType>(x) * static_cast<IntermediateType>(y); |
| 226 | *validity = DstRangeRelationToSrcRange<T>(tmp); |
| 227 | return static_cast<T>(tmp); |
| 228 | } |
| 229 | |
| 230 | template <typename T> |
| 231 | typename std::enable_if<std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed && |
| 232 | (sizeof(T) * 2 > sizeof(uintmax_t)), |
| 233 | T>::type |
| 234 | CheckedMul(T x, T y, RangeConstraint *validity) |
| 235 | { |
| 236 | // If either side is zero then the result will be zero. |
| 237 | if (!x || !y) |
| 238 | { |
| 239 | *validity = RANGE_VALID; |
| 240 | return static_cast<T>(0); |
| 241 | } |
| 242 | else if (x > 0) |
| 243 | { |
| 244 | if (y > 0) |
| 245 | *validity = x <= std::numeric_limits<T>::max() / y ? RANGE_VALID : RANGE_OVERFLOW; |
| 246 | else |
| 247 | *validity = y >= std::numeric_limits<T>::min() / x ? RANGE_VALID : RANGE_UNDERFLOW; |
| 248 | } |
| 249 | else |
| 250 | { |
| 251 | if (y > 0) |
| 252 | *validity = x >= std::numeric_limits<T>::min() / y ? RANGE_VALID : RANGE_UNDERFLOW; |
| 253 | else |
| 254 | *validity = y >= std::numeric_limits<T>::max() / x ? RANGE_VALID : RANGE_OVERFLOW; |
| 255 | } |
| 256 | |
| 257 | return static_cast<T>(x * y); |
| 258 | } |
| 259 | |
| 260 | template <typename T> |
| 261 | typename std::enable_if<std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed && |
| 262 | (sizeof(T) * 2 > sizeof(uintmax_t)), |
| 263 | T>::type |
| 264 | CheckedMul(T x, T y, RangeConstraint *validity) |
| 265 | { |
| 266 | *validity = (y == 0 || x <= std::numeric_limits<T>::max() / y) ? RANGE_VALID : RANGE_OVERFLOW; |
| 267 | return static_cast<T>(x * y); |
| 268 | } |
| 269 | |
| 270 | // Division just requires a check for an invalid negation on signed min/-1. |
| 271 | template <typename T> |
| 272 | T CheckedDiv(T x, |
| 273 | T y, |
| 274 | RangeConstraint *validity, |
| 275 | typename std::enable_if<std::numeric_limits<T>::is_integer, int>::type = 0) |
| 276 | { |
| 277 | if (std::numeric_limits<T>::is_signed && x == std::numeric_limits<T>::min() && |
| 278 | y == static_cast<T>(-1)) |
| 279 | { |
| 280 | *validity = RANGE_OVERFLOW; |
| 281 | return std::numeric_limits<T>::min(); |
| 282 | } |
| 283 | |
| 284 | *validity = RANGE_VALID; |
| 285 | return static_cast<T>(x / y); |
| 286 | } |
| 287 | |
| 288 | template <typename T> |
| 289 | typename std::enable_if<std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 290 | T>::type |
| 291 | CheckedMod(T x, T y, RangeConstraint *validity) |
| 292 | { |
| 293 | *validity = y > 0 ? RANGE_VALID : RANGE_INVALID; |
| 294 | return static_cast<T>(x % y); |
| 295 | } |
| 296 | |
| 297 | template <typename T> |
| 298 | typename std::enable_if<std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 299 | T>::type |
| 300 | CheckedMod(T x, T y, RangeConstraint *validity) |
| 301 | { |
| 302 | *validity = RANGE_VALID; |
| 303 | return static_cast<T>(x % y); |
| 304 | } |
| 305 | |
| 306 | template <typename T> |
| 307 | typename std::enable_if<std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 308 | T>::type |
| 309 | CheckedNeg(T value, RangeConstraint *validity) |
| 310 | { |
| 311 | *validity = value != std::numeric_limits<T>::min() ? RANGE_VALID : RANGE_OVERFLOW; |
| 312 | // The negation of signed min is min, so catch that one. |
| 313 | return static_cast<T>(-value); |
| 314 | } |
| 315 | |
| 316 | template <typename T> |
| 317 | typename std::enable_if<std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 318 | T>::type |
| 319 | CheckedNeg(T value, RangeConstraint *validity) |
| 320 | { |
| 321 | // The only legal unsigned negation is zero. |
| 322 | *validity = value ? RANGE_UNDERFLOW : RANGE_VALID; |
| 323 | return static_cast<T>(-static_cast<typename SignedIntegerForSize<T>::type>(value)); |
| 324 | } |
| 325 | |
| 326 | template <typename T> |
| 327 | typename std::enable_if<std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 328 | T>::type |
| 329 | CheckedAbs(T value, RangeConstraint *validity) |
| 330 | { |
| 331 | *validity = value != std::numeric_limits<T>::min() ? RANGE_VALID : RANGE_OVERFLOW; |
| 332 | return static_cast<T>(std::abs(value)); |
| 333 | } |
| 334 | |
| 335 | template <typename T> |
| 336 | typename std::enable_if<std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 337 | T>::type |
| 338 | CheckedAbs(T value, RangeConstraint *validity) |
| 339 | { |
| 340 | // T is unsigned, so |value| must already be positive. |
| 341 | *validity = RANGE_VALID; |
| 342 | return value; |
| 343 | } |
| 344 | |
| 345 | template <typename T> |
| 346 | typename std::enable_if<std::numeric_limits<T>::is_integer && std::numeric_limits<T>::is_signed, |
| 347 | typename UnsignedIntegerForSize<T>::type>::type |
| 348 | CheckedUnsignedAbs(T value) |
| 349 | { |
| 350 | typedef typename UnsignedIntegerForSize<T>::type UnsignedT; |
| 351 | return value == std::numeric_limits<T>::min() |
| 352 | ? static_cast<UnsignedT>(std::numeric_limits<T>::max()) + 1 |
| 353 | : static_cast<UnsignedT>(std::abs(value)); |
| 354 | } |
| 355 | |
| 356 | template <typename T> |
| 357 | typename std::enable_if<std::numeric_limits<T>::is_integer && !std::numeric_limits<T>::is_signed, |
| 358 | T>::type |
| 359 | CheckedUnsignedAbs(T value) |
| 360 | { |
| 361 | // T is unsigned, so |value| must already be positive. |
| 362 | return static_cast<T>(value); |
| 363 | } |
| 364 | |
| 365 | // These are the floating point stubs that the compiler needs to see. Only the |
| 366 | // negation operation is ever called. |
| 367 | #define ANGLEBASE_FLOAT_ARITHMETIC_STUBS(NAME) \ |
| 368 | template <typename T> \ |
| 369 | typename std::enable_if<std::numeric_limits<T>::is_iec559, T>::type Checked##NAME( \ |
| 370 | T, T, RangeConstraint *) \ |
| 371 | { \ |
| 372 | NOTREACHED(); \ |
| 373 | return static_cast<T>(0); \ |
| 374 | } |
| 375 | |
| 376 | ANGLEBASE_FLOAT_ARITHMETIC_STUBS(Add) |
| 377 | ANGLEBASE_FLOAT_ARITHMETIC_STUBS(Sub) |
| 378 | ANGLEBASE_FLOAT_ARITHMETIC_STUBS(Mul) |
| 379 | ANGLEBASE_FLOAT_ARITHMETIC_STUBS(Div) |
| 380 | ANGLEBASE_FLOAT_ARITHMETIC_STUBS(Mod) |
| 381 | |
| 382 | #undef ANGLEBASE_FLOAT_ARITHMETIC_STUBS |
| 383 | |
| 384 | template <typename T> |
| 385 | typename std::enable_if<std::numeric_limits<T>::is_iec559, T>::type CheckedNeg(T value, |
| 386 | RangeConstraint *) |
| 387 | { |
| 388 | return static_cast<T>(-value); |
| 389 | } |
| 390 | |
| 391 | template <typename T> |
| 392 | typename std::enable_if<std::numeric_limits<T>::is_iec559, T>::type CheckedAbs(T value, |
| 393 | RangeConstraint *) |
| 394 | { |
| 395 | return static_cast<T>(std::abs(value)); |
| 396 | } |
| 397 | |
| 398 | // Floats carry around their validity state with them, but integers do not. So, |
| 399 | // we wrap the underlying value in a specialization in order to hide that detail |
| 400 | // and expose an interface via accessors. |
| 401 | enum NumericRepresentation |
| 402 | { |
| 403 | NUMERIC_INTEGER, |
| 404 | NUMERIC_FLOATING, |
| 405 | NUMERIC_UNKNOWN |
| 406 | }; |
| 407 | |
| 408 | template <typename NumericType> |
| 409 | struct GetNumericRepresentation |
| 410 | { |
| 411 | static const NumericRepresentation value = |
| 412 | std::numeric_limits<NumericType>::is_integer |
| 413 | ? NUMERIC_INTEGER |
| 414 | : (std::numeric_limits<NumericType>::is_iec559 ? NUMERIC_FLOATING : NUMERIC_UNKNOWN); |
| 415 | }; |
| 416 | |
| 417 | template <typename T, NumericRepresentation type = GetNumericRepresentation<T>::value> |
| 418 | class CheckedNumericState |
| 419 | {}; |
| 420 | |
| 421 | // Integrals require quite a bit of additional housekeeping to manage state. |
| 422 | template <typename T> |
| 423 | class CheckedNumericState<T, NUMERIC_INTEGER> |
| 424 | { |
| 425 | private: |
| 426 | T value_; |
| 427 | RangeConstraint validity_ : CHAR_BIT; // Actually requires only two bits. |
| 428 | |
| 429 | public: |
| 430 | template <typename Src, NumericRepresentation type> |
| 431 | friend class CheckedNumericState; |
| 432 | |
| 433 | CheckedNumericState() : value_(0), validity_(RANGE_VALID) {} |
| 434 | |
| 435 | template <typename Src> |
| 436 | CheckedNumericState(Src value, RangeConstraint validity) |
| 437 | : value_(static_cast<T>(value)), |
| 438 | validity_(GetRangeConstraint(validity | DstRangeRelationToSrcRange<T>(value))) |
| 439 | { |
| 440 | static_assert(std::numeric_limits<Src>::is_specialized, "Argument must be numeric." ); |
| 441 | } |
| 442 | |
| 443 | // Copy constructor. |
| 444 | template <typename Src> |
| 445 | CheckedNumericState(const CheckedNumericState<Src> &rhs) |
| 446 | : value_(static_cast<T>(rhs.value())), |
| 447 | validity_(GetRangeConstraint(rhs.validity() | DstRangeRelationToSrcRange<T>(rhs.value()))) |
| 448 | {} |
| 449 | |
| 450 | template <typename Src> |
| 451 | explicit CheckedNumericState( |
| 452 | Src value, |
| 453 | typename std::enable_if<std::numeric_limits<Src>::is_specialized, int>::type = 0) |
| 454 | : value_(static_cast<T>(value)), validity_(DstRangeRelationToSrcRange<T>(value)) |
| 455 | {} |
| 456 | |
| 457 | RangeConstraint validity() const { return validity_; } |
| 458 | T value() const { return value_; } |
| 459 | }; |
| 460 | |
| 461 | // Floating points maintain their own validity, but need translation wrappers. |
| 462 | template <typename T> |
| 463 | class CheckedNumericState<T, NUMERIC_FLOATING> |
| 464 | { |
| 465 | private: |
| 466 | T value_; |
| 467 | |
| 468 | public: |
| 469 | template <typename Src, NumericRepresentation type> |
| 470 | friend class CheckedNumericState; |
| 471 | |
| 472 | CheckedNumericState() : value_(0.0) {} |
| 473 | |
| 474 | template <typename Src> |
| 475 | CheckedNumericState( |
| 476 | Src value, |
| 477 | RangeConstraint validity, |
| 478 | typename std::enable_if<std::numeric_limits<Src>::is_integer, int>::type = 0) |
| 479 | { |
| 480 | switch (DstRangeRelationToSrcRange<T>(value)) |
| 481 | { |
| 482 | case RANGE_VALID: |
| 483 | value_ = static_cast<T>(value); |
| 484 | break; |
| 485 | |
| 486 | case RANGE_UNDERFLOW: |
| 487 | value_ = -std::numeric_limits<T>::infinity(); |
| 488 | break; |
| 489 | |
| 490 | case RANGE_OVERFLOW: |
| 491 | value_ = std::numeric_limits<T>::infinity(); |
| 492 | break; |
| 493 | |
| 494 | case RANGE_INVALID: |
| 495 | value_ = std::numeric_limits<T>::quiet_NaN(); |
| 496 | break; |
| 497 | |
| 498 | default: |
| 499 | NOTREACHED(); |
| 500 | } |
| 501 | } |
| 502 | |
| 503 | template <typename Src> |
| 504 | explicit CheckedNumericState( |
| 505 | Src value, |
| 506 | typename std::enable_if<std::numeric_limits<Src>::is_specialized, int>::type = 0) |
| 507 | : value_(static_cast<T>(value)) |
| 508 | {} |
| 509 | |
| 510 | // Copy constructor. |
| 511 | template <typename Src> |
| 512 | CheckedNumericState(const CheckedNumericState<Src> &rhs) : value_(static_cast<T>(rhs.value())) |
| 513 | {} |
| 514 | |
| 515 | RangeConstraint validity() const |
| 516 | { |
| 517 | return GetRangeConstraint(value_ <= std::numeric_limits<T>::max(), |
| 518 | value_ >= -std::numeric_limits<T>::max()); |
| 519 | } |
| 520 | T value() const { return value_; } |
| 521 | }; |
| 522 | |
| 523 | // For integers less than 128-bit and floats 32-bit or larger, we have the type |
| 524 | // with the larger maximum exponent take precedence. |
| 525 | enum ArithmeticPromotionCategory |
| 526 | { |
| 527 | LEFT_PROMOTION, |
| 528 | RIGHT_PROMOTION |
| 529 | }; |
| 530 | |
| 531 | template <typename Lhs, |
| 532 | typename Rhs = Lhs, |
| 533 | ArithmeticPromotionCategory Promotion = |
| 534 | (MaxExponent<Lhs>::value > MaxExponent<Rhs>::value) ? LEFT_PROMOTION |
| 535 | : RIGHT_PROMOTION> |
| 536 | struct ArithmeticPromotion; |
| 537 | |
| 538 | template <typename Lhs, typename Rhs> |
| 539 | struct ArithmeticPromotion<Lhs, Rhs, LEFT_PROMOTION> |
| 540 | { |
| 541 | typedef Lhs type; |
| 542 | }; |
| 543 | |
| 544 | template <typename Lhs, typename Rhs> |
| 545 | struct ArithmeticPromotion<Lhs, Rhs, RIGHT_PROMOTION> |
| 546 | { |
| 547 | typedef Rhs type; |
| 548 | }; |
| 549 | |
| 550 | // We can statically check if operations on the provided types can wrap, so we |
| 551 | // can skip the checked operations if they're not needed. So, for an integer we |
| 552 | // care if the destination type preserves the sign and is twice the width of |
| 553 | // the source. |
| 554 | template <typename T, typename Lhs, typename Rhs> |
| 555 | struct IsIntegerArithmeticSafe |
| 556 | { |
| 557 | static const bool value = |
| 558 | !std::numeric_limits<T>::is_iec559 && |
| 559 | StaticDstRangeRelationToSrcRange<T, Lhs>::value == NUMERIC_RANGE_CONTAINED && |
| 560 | sizeof(T) >= (2 * sizeof(Lhs)) && |
| 561 | StaticDstRangeRelationToSrcRange<T, Rhs>::value != NUMERIC_RANGE_CONTAINED && |
| 562 | sizeof(T) >= (2 * sizeof(Rhs)); |
| 563 | }; |
| 564 | |
| 565 | } // namespace internal |
| 566 | } // namespace base |
| 567 | |
| 568 | } // namespace angle |
| 569 | |
| 570 | #endif // ANGLEBASE_NUMERICS_SAFE_MATH_IMPL_H_ |
| 571 | |