Home | History | Annotate | Download | only in objects
      1 // Copyright 2017 the V8 project 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 // Parts of the implementation below:
      6 
      7 // Copyright (c) 2014 the Dart project authors.  Please see the AUTHORS file [1]
      8 // for details. All rights reserved. Use of this source code is governed by a
      9 // BSD-style license that can be found in the LICENSE file [2].
     10 //
     11 // [1] https://github.com/dart-lang/sdk/blob/master/AUTHORS
     12 // [2] https://github.com/dart-lang/sdk/blob/master/LICENSE
     13 
     14 // Copyright 2009 The Go Authors. All rights reserved.
     15 // Use of this source code is governed by a BSD-style
     16 // license that can be found in the LICENSE file [3].
     17 //
     18 // [3] https://golang.org/LICENSE
     19 
     20 #include "src/objects/bigint.h"
     21 
     22 #include "src/double.h"
     23 #include "src/objects-inl.h"
     24 
     25 namespace v8 {
     26 namespace internal {
     27 
     28 // The MutableBigInt class is an implementation detail designed to prevent
     29 // accidental mutation of a BigInt after its construction. Step-by-step
     30 // construction of a BigInt must happen in terms of MutableBigInt, the
     31 // final result is then passed through MutableBigInt::MakeImmutable and not
     32 // modified further afterwards.
     33 // Many of the functions in this class use arguments of type {BigIntBase},
     34 // indicating that they will be used in a read-only capacity, and both
     35 // {BigInt} and {MutableBigInt} objects can be passed in.
     36 class MutableBigInt : public FreshlyAllocatedBigInt,
     37                       public NeverReadOnlySpaceObject {
     38  public:
     39   using NeverReadOnlySpaceObject::GetHeap;
     40   using NeverReadOnlySpaceObject::GetIsolate;
     41 
     42   // Bottleneck for converting MutableBigInts to BigInts.
     43   static MaybeHandle<BigInt> MakeImmutable(MaybeHandle<MutableBigInt> maybe);
     44   static Handle<BigInt> MakeImmutable(Handle<MutableBigInt> result);
     45 
     46   // Allocation helpers.
     47   static MaybeHandle<MutableBigInt> New(Isolate* isolate, int length,
     48                                         PretenureFlag pretenure = NOT_TENURED);
     49   static Handle<BigInt> NewFromInt(Isolate* isolate, int value);
     50   static Handle<BigInt> NewFromDouble(Isolate* isolate, double value);
     51   void InitializeDigits(int length, byte value = 0);
     52   static Handle<MutableBigInt> Copy(Isolate* isolate,
     53                                     Handle<BigIntBase> source);
     54   static Handle<BigInt> Zero(Isolate* isolate) {
     55     // TODO(jkummerow): Consider caching a canonical zero-BigInt.
     56     return MakeImmutable(New(isolate, 0)).ToHandleChecked();
     57   }
     58 
     59   static Handle<MutableBigInt> Cast(Handle<FreshlyAllocatedBigInt> bigint) {
     60     SLOW_DCHECK(bigint->IsBigInt());
     61     return Handle<MutableBigInt>::cast(bigint);
     62   }
     63 
     64   // Internal helpers.
     65   static MaybeHandle<MutableBigInt> BitwiseAnd(Isolate* isolate,
     66                                                Handle<BigInt> x,
     67                                                Handle<BigInt> y);
     68   static MaybeHandle<MutableBigInt> BitwiseXor(Isolate* isolate,
     69                                                Handle<BigInt> x,
     70                                                Handle<BigInt> y);
     71   static MaybeHandle<MutableBigInt> BitwiseOr(Isolate* isolate,
     72                                               Handle<BigInt> x,
     73                                               Handle<BigInt> y);
     74 
     75   static Handle<BigInt> TruncateToNBits(Isolate* isolate, int n,
     76                                         Handle<BigInt> x);
     77   static Handle<BigInt> TruncateAndSubFromPowerOfTwo(Isolate* isolate, int n,
     78                                                      Handle<BigInt> x,
     79                                                      bool result_sign);
     80 
     81   static MaybeHandle<BigInt> AbsoluteAdd(Isolate* isolate, Handle<BigInt> x,
     82                                          Handle<BigInt> y, bool result_sign);
     83   static Handle<BigInt> AbsoluteSub(Isolate* isolate, Handle<BigInt> x,
     84                                     Handle<BigInt> y, bool result_sign);
     85   static MaybeHandle<MutableBigInt> AbsoluteAddOne(
     86       Isolate* isolate, Handle<BigIntBase> x, bool sign,
     87       MutableBigInt* result_storage = nullptr);
     88   static Handle<MutableBigInt> AbsoluteSubOne(Isolate* isolate,
     89                                               Handle<BigIntBase> x);
     90   static MaybeHandle<MutableBigInt> AbsoluteSubOne(Isolate* isolate,
     91                                                    Handle<BigIntBase> x,
     92                                                    int result_length);
     93 
     94   enum ExtraDigitsHandling { kCopy, kSkip };
     95   enum SymmetricOp { kSymmetric, kNotSymmetric };
     96   static inline Handle<MutableBigInt> AbsoluteBitwiseOp(
     97       Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
     98       MutableBigInt* result_storage, ExtraDigitsHandling extra_digits,
     99       SymmetricOp symmetric, std::function<digit_t(digit_t, digit_t)> op);
    100   static Handle<MutableBigInt> AbsoluteAnd(
    101       Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
    102       MutableBigInt* result_storage = nullptr);
    103   static Handle<MutableBigInt> AbsoluteAndNot(
    104       Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
    105       MutableBigInt* result_storage = nullptr);
    106   static Handle<MutableBigInt> AbsoluteOr(
    107       Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
    108       MutableBigInt* result_storage = nullptr);
    109   static Handle<MutableBigInt> AbsoluteXor(
    110       Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
    111       MutableBigInt* result_storage = nullptr);
    112 
    113   static int AbsoluteCompare(Handle<BigIntBase> x, Handle<BigIntBase> y);
    114 
    115   static void MultiplyAccumulate(Handle<BigIntBase> multiplicand,
    116                                  digit_t multiplier,
    117                                  Handle<MutableBigInt> accumulator,
    118                                  int accumulator_index);
    119   static void InternalMultiplyAdd(BigIntBase* source, digit_t factor,
    120                                   digit_t summand, int n,
    121                                   MutableBigInt* result);
    122   void InplaceMultiplyAdd(uintptr_t factor, uintptr_t summand);
    123 
    124   // Specialized helpers for Divide/Remainder.
    125   static void AbsoluteDivSmall(Isolate* isolate, Handle<BigIntBase> x,
    126                                digit_t divisor, Handle<MutableBigInt>* quotient,
    127                                digit_t* remainder);
    128   static bool AbsoluteDivLarge(Isolate* isolate, Handle<BigIntBase> dividend,
    129                                Handle<BigIntBase> divisor,
    130                                Handle<MutableBigInt>* quotient,
    131                                Handle<MutableBigInt>* remainder);
    132   static bool ProductGreaterThan(digit_t factor1, digit_t factor2, digit_t high,
    133                                  digit_t low);
    134   digit_t InplaceAdd(Handle<BigIntBase> summand, int start_index);
    135   digit_t InplaceSub(Handle<BigIntBase> subtrahend, int start_index);
    136   void InplaceRightShift(int shift);
    137   enum SpecialLeftShiftMode {
    138     kSameSizeResult,
    139     kAlwaysAddOneDigit,
    140   };
    141   static MaybeHandle<MutableBigInt> SpecialLeftShift(Isolate* isolate,
    142                                                      Handle<BigIntBase> x,
    143                                                      int shift,
    144                                                      SpecialLeftShiftMode mode);
    145 
    146   // Specialized helpers for shift operations.
    147   static MaybeHandle<BigInt> LeftShiftByAbsolute(Isolate* isolate,
    148                                                  Handle<BigIntBase> x,
    149                                                  Handle<BigIntBase> y);
    150   static Handle<BigInt> RightShiftByAbsolute(Isolate* isolate,
    151                                              Handle<BigIntBase> x,
    152                                              Handle<BigIntBase> y);
    153   static Handle<BigInt> RightShiftByMaximum(Isolate* isolate, bool sign);
    154   static Maybe<digit_t> ToShiftAmount(Handle<BigIntBase> x);
    155 
    156   static MaybeHandle<String> ToStringBasePowerOfTwo(Isolate* isolate,
    157                                                     Handle<BigIntBase> x,
    158                                                     int radix);
    159   static MaybeHandle<String> ToStringGeneric(Isolate* isolate,
    160                                              Handle<BigIntBase> x, int radix);
    161 
    162   static double ToDouble(Handle<BigIntBase> x);
    163   enum Rounding { kRoundDown, kTie, kRoundUp };
    164   static Rounding DecideRounding(Handle<BigIntBase> x, int mantissa_bits_unset,
    165                                  int digit_index, uint64_t current_digit);
    166 
    167   // Returns the least significant 64 bits, simulating two's complement
    168   // representation.
    169   static uint64_t GetRawBits(BigIntBase* x, bool* lossless);
    170 
    171   // Digit arithmetic helpers.
    172   static inline digit_t digit_add(digit_t a, digit_t b, digit_t* carry);
    173   static inline digit_t digit_sub(digit_t a, digit_t b, digit_t* borrow);
    174   static inline digit_t digit_mul(digit_t a, digit_t b, digit_t* high);
    175   static inline digit_t digit_div(digit_t high, digit_t low, digit_t divisor,
    176                                   digit_t* remainder);
    177   static digit_t digit_pow(digit_t base, digit_t exponent);
    178   static inline bool digit_ismax(digit_t x) {
    179     return static_cast<digit_t>(~x) == 0;
    180   }
    181 
    182 // Internal field setters. Non-mutable BigInts don't have these.
    183 #include "src/objects/object-macros.h"
    184   inline void set_sign(bool new_sign) {
    185     intptr_t bitfield = READ_INTPTR_FIELD(this, kBitfieldOffset);
    186     bitfield = SignBits::update(static_cast<uint32_t>(bitfield), new_sign);
    187     WRITE_INTPTR_FIELD(this, kBitfieldOffset, bitfield);
    188   }
    189   inline void set_length(int new_length) {
    190     intptr_t bitfield = READ_INTPTR_FIELD(this, kBitfieldOffset);
    191     bitfield = LengthBits::update(static_cast<uint32_t>(bitfield), new_length);
    192     WRITE_INTPTR_FIELD(this, kBitfieldOffset, bitfield);
    193   }
    194   inline void initialize_bitfield(bool sign, int length) {
    195     intptr_t bitfield = LengthBits::encode(length) | SignBits::encode(sign);
    196     WRITE_INTPTR_FIELD(this, kBitfieldOffset, bitfield);
    197   }
    198   inline void set_digit(int n, digit_t value) {
    199     SLOW_DCHECK(0 <= n && n < length());
    200     Address address = FIELD_ADDR(this, kDigitsOffset + n * kDigitSize);
    201     (*reinterpret_cast<digit_t*>(address)) = value;
    202   }
    203 #include "src/objects/object-macros-undef.h"
    204 
    205   void set_64_bits(uint64_t bits);
    206 };
    207 
    208 MaybeHandle<MutableBigInt> MutableBigInt::New(Isolate* isolate, int length,
    209                                               PretenureFlag pretenure) {
    210   if (length > BigInt::kMaxLength) {
    211     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
    212                     MutableBigInt);
    213   }
    214   Handle<MutableBigInt> result =
    215       Cast(isolate->factory()->NewBigInt(length, pretenure));
    216   result->initialize_bitfield(false, length);
    217 #if DEBUG
    218   result->InitializeDigits(length, 0xBF);
    219 #endif
    220   return result;
    221 }
    222 
    223 Handle<BigInt> MutableBigInt::NewFromInt(Isolate* isolate, int value) {
    224   if (value == 0) return Zero(isolate);
    225   Handle<MutableBigInt> result = Cast(isolate->factory()->NewBigInt(1));
    226   bool sign = value < 0;
    227   result->initialize_bitfield(sign, 1);
    228   if (!sign) {
    229     result->set_digit(0, value);
    230   } else {
    231     if (value == kMinInt) {
    232       STATIC_ASSERT(kMinInt == -kMaxInt - 1);
    233       result->set_digit(0, static_cast<BigInt::digit_t>(kMaxInt) + 1);
    234     } else {
    235       result->set_digit(0, -value);
    236     }
    237   }
    238   return MakeImmutable(result);
    239 }
    240 
    241 Handle<BigInt> MutableBigInt::NewFromDouble(Isolate* isolate, double value) {
    242   DCHECK_EQ(value, std::floor(value));
    243   if (value == 0) return Zero(isolate);
    244 
    245   bool sign = value < 0;  // -0 was already handled above.
    246   uint64_t double_bits = bit_cast<uint64_t>(value);
    247   int raw_exponent =
    248       static_cast<int>(double_bits >> Double::kPhysicalSignificandSize) & 0x7FF;
    249   DCHECK_NE(raw_exponent, 0x7FF);
    250   DCHECK_GE(raw_exponent, 0x3FF);
    251   int exponent = raw_exponent - 0x3FF;
    252   int digits = exponent / kDigitBits + 1;
    253   Handle<MutableBigInt> result = Cast(isolate->factory()->NewBigInt(digits));
    254   result->initialize_bitfield(sign, digits);
    255 
    256   // We construct a BigInt from the double {value} by shifting its mantissa
    257   // according to its exponent and mapping the bit pattern onto digits.
    258   //
    259   //               <----------- bitlength = exponent + 1 ----------->
    260   //                <----- 52 ------> <------ trailing zeroes ------>
    261   // mantissa:     1yyyyyyyyyyyyyyyyy 0000000000000000000000000000000
    262   // digits:    0001xxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
    263   //                <-->          <------>
    264   //          msd_topbit         kDigitBits
    265   //
    266   uint64_t mantissa =
    267       (double_bits & Double::kSignificandMask) | Double::kHiddenBit;
    268   const int kMantissaTopBit = Double::kSignificandSize - 1;  // 0-indexed.
    269   // 0-indexed position of most significant bit in the most significant digit.
    270   int msd_topbit = exponent % kDigitBits;
    271   // Number of unused bits in {mantissa}. We'll keep them shifted to the
    272   // left (i.e. most significant part) of the underlying uint64_t.
    273   int remaining_mantissa_bits = 0;
    274   // Next digit under construction.
    275   digit_t digit;
    276 
    277   // First, build the MSD by shifting the mantissa appropriately.
    278   if (msd_topbit < kMantissaTopBit) {
    279     remaining_mantissa_bits = kMantissaTopBit - msd_topbit;
    280     digit = mantissa >> remaining_mantissa_bits;
    281     mantissa = mantissa << (64 - remaining_mantissa_bits);
    282   } else {
    283     DCHECK_GE(msd_topbit, kMantissaTopBit);
    284     digit = mantissa << (msd_topbit - kMantissaTopBit);
    285     mantissa = 0;
    286   }
    287   result->set_digit(digits - 1, digit);
    288   // Then fill in the rest of the digits.
    289   for (int digit_index = digits - 2; digit_index >= 0; digit_index--) {
    290     if (remaining_mantissa_bits > 0) {
    291       remaining_mantissa_bits -= kDigitBits;
    292       if (sizeof(digit) == 4) {
    293         digit = mantissa >> 32;
    294         mantissa = mantissa << 32;
    295       } else {
    296         DCHECK_EQ(sizeof(digit), 8);
    297         digit = mantissa;
    298         mantissa = 0;
    299       }
    300     } else {
    301       digit = 0;
    302     }
    303     result->set_digit(digit_index, digit);
    304   }
    305   return MakeImmutable(result);
    306 }
    307 
    308 Handle<MutableBigInt> MutableBigInt::Copy(Isolate* isolate,
    309                                           Handle<BigIntBase> source) {
    310   int length = source->length();
    311   // Allocating a BigInt of the same length as an existing BigInt cannot throw.
    312   Handle<MutableBigInt> result = New(isolate, length).ToHandleChecked();
    313   memcpy(reinterpret_cast<void*>(result->address() + BigIntBase::kHeaderSize),
    314          reinterpret_cast<void*>(source->address() + BigIntBase::kHeaderSize),
    315          BigInt::SizeFor(length) - BigIntBase::kHeaderSize);
    316   return result;
    317 }
    318 
    319 void MutableBigInt::InitializeDigits(int length, byte value) {
    320   memset(reinterpret_cast<void*>(reinterpret_cast<Address>(this) +
    321                                  kDigitsOffset - kHeapObjectTag),
    322          value, length * kDigitSize);
    323 }
    324 
    325 MaybeHandle<BigInt> MutableBigInt::MakeImmutable(
    326     MaybeHandle<MutableBigInt> maybe) {
    327   Handle<MutableBigInt> result;
    328   if (!maybe.ToHandle(&result)) return MaybeHandle<BigInt>();
    329   return MakeImmutable(result);
    330 }
    331 
    332 Handle<BigInt> MutableBigInt::MakeImmutable(Handle<MutableBigInt> result) {
    333   // Check if we need to right-trim any leading zero-digits.
    334   int old_length = result->length();
    335   int new_length = old_length;
    336   while (new_length > 0 && result->digit(new_length - 1) == 0) new_length--;
    337   int to_trim = old_length - new_length;
    338   if (to_trim != 0) {
    339     int size_delta = to_trim * kDigitSize;
    340     Address new_end = result->address() + BigInt::SizeFor(new_length);
    341     Heap* heap = result->GetHeap();
    342     heap->CreateFillerObjectAt(new_end, size_delta, ClearRecordedSlots::kNo);
    343     result->set_length(new_length);
    344 
    345     // Canonicalize -0n.
    346     if (new_length == 0) {
    347       result->set_sign(false);
    348       // TODO(jkummerow): If we cache a canonical 0n, return that here.
    349     }
    350   }
    351   DCHECK_IMPLIES(result->length() > 0,
    352                  result->digit(result->length() - 1) != 0);  // MSD is non-zero.
    353   return Handle<BigInt>(reinterpret_cast<BigInt**>(result.location()));
    354 }
    355 
    356 Handle<BigInt> BigInt::Zero(Isolate* isolate) {
    357   return MutableBigInt::Zero(isolate);
    358 }
    359 
    360 Handle<BigInt> BigInt::UnaryMinus(Isolate* isolate, Handle<BigInt> x) {
    361   // Special case: There is no -0n.
    362   if (x->is_zero()) {
    363     return x;
    364   }
    365   Handle<MutableBigInt> result = MutableBigInt::Copy(isolate, x);
    366   result->set_sign(!x->sign());
    367   return MutableBigInt::MakeImmutable(result);
    368 }
    369 
    370 MaybeHandle<BigInt> BigInt::BitwiseNot(Isolate* isolate, Handle<BigInt> x) {
    371   MaybeHandle<MutableBigInt> result;
    372   if (x->sign()) {
    373     // ~(-x) == ~(~(x-1)) == x-1
    374     result = MutableBigInt::AbsoluteSubOne(isolate, x, x->length());
    375   } else {
    376     // ~x == -x-1 == -(x+1)
    377     result = MutableBigInt::AbsoluteAddOne(isolate, x, true);
    378   }
    379   return MutableBigInt::MakeImmutable(result);
    380 }
    381 
    382 MaybeHandle<BigInt> BigInt::Exponentiate(Isolate* isolate, Handle<BigInt> base,
    383                                          Handle<BigInt> exponent) {
    384   // 1. If exponent is < 0, throw a RangeError exception.
    385   if (exponent->sign()) {
    386     THROW_NEW_ERROR(isolate,
    387                     NewRangeError(MessageTemplate::kBigIntNegativeExponent),
    388                     BigInt);
    389   }
    390   // 2. If base is 0n and exponent is 0n, return 1n.
    391   if (exponent->is_zero()) {
    392     return MutableBigInt::NewFromInt(isolate, 1);
    393   }
    394   // 3. Return a BigInt representing the mathematical value of base raised
    395   //    to the power exponent.
    396   if (base->is_zero()) return base;
    397   if (base->length() == 1 && base->digit(0) == 1) {
    398     // (-1) ** even_number == 1.
    399     if (base->sign() && (exponent->digit(0) & 1) == 0) {
    400       return UnaryMinus(isolate, base);
    401     }
    402     // (-1) ** odd_number == -1; 1 ** anything == 1.
    403     return base;
    404   }
    405   // For all bases >= 2, very large exponents would lead to unrepresentable
    406   // results.
    407   STATIC_ASSERT(kMaxLengthBits < std::numeric_limits<digit_t>::max());
    408   if (exponent->length() > 1) {
    409     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
    410                     BigInt);
    411   }
    412   digit_t exp_value = exponent->digit(0);
    413   if (exp_value == 1) return base;
    414   if (exp_value >= kMaxLengthBits) {
    415     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
    416                     BigInt);
    417   }
    418   STATIC_ASSERT(kMaxLengthBits <= kMaxInt);
    419   int n = static_cast<int>(exp_value);
    420   if (base->length() == 1 && base->digit(0) == 2) {
    421     // Fast path for 2^n.
    422     int needed_digits = 1 + (n / kDigitBits);
    423     Handle<MutableBigInt> result;
    424     if (!MutableBigInt::New(isolate, needed_digits).ToHandle(&result)) {
    425       return MaybeHandle<BigInt>();
    426     }
    427     result->InitializeDigits(needed_digits);
    428     // All bits are zero. Now set the n-th bit.
    429     digit_t msd = static_cast<digit_t>(1) << (n % kDigitBits);
    430     result->set_digit(needed_digits - 1, msd);
    431     // Result is negative for odd powers of -2n.
    432     if (base->sign()) result->set_sign((n & 1) != 0);
    433     return MutableBigInt::MakeImmutable(result);
    434   }
    435   Handle<BigInt> result;
    436   Handle<BigInt> running_square = base;
    437   // This implicitly sets the result's sign correctly.
    438   if (n & 1) result = base;
    439   n >>= 1;
    440   for (; n != 0; n >>= 1) {
    441     MaybeHandle<BigInt> maybe_result =
    442         Multiply(isolate, running_square, running_square);
    443     if (!maybe_result.ToHandle(&running_square)) return maybe_result;
    444     if (n & 1) {
    445       if (result.is_null()) {
    446         result = running_square;
    447       } else {
    448         maybe_result = Multiply(isolate, result, running_square);
    449         if (!maybe_result.ToHandle(&result)) return maybe_result;
    450       }
    451     }
    452   }
    453   return result;
    454 }
    455 
    456 MaybeHandle<BigInt> BigInt::Multiply(Isolate* isolate, Handle<BigInt> x,
    457                                      Handle<BigInt> y) {
    458   if (x->is_zero()) return x;
    459   if (y->is_zero()) return y;
    460   int result_length = x->length() + y->length();
    461   Handle<MutableBigInt> result;
    462   if (!MutableBigInt::New(isolate, result_length).ToHandle(&result)) {
    463     return MaybeHandle<BigInt>();
    464   }
    465   result->InitializeDigits(result_length);
    466   for (int i = 0; i < x->length(); i++) {
    467     MutableBigInt::MultiplyAccumulate(y, x->digit(i), result, i);
    468   }
    469   result->set_sign(x->sign() != y->sign());
    470   return MutableBigInt::MakeImmutable(result);
    471 }
    472 
    473 MaybeHandle<BigInt> BigInt::Divide(Isolate* isolate, Handle<BigInt> x,
    474                                    Handle<BigInt> y) {
    475   // 1. If y is 0n, throw a RangeError exception.
    476   if (y->is_zero()) {
    477     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntDivZero),
    478                     BigInt);
    479   }
    480   // 2. Let quotient be the mathematical value of x divided by y.
    481   // 3. Return a BigInt representing quotient rounded towards 0 to the next
    482   //    integral value.
    483   if (MutableBigInt::AbsoluteCompare(x, y) < 0) {
    484     return Zero(isolate);
    485   }
    486   Handle<MutableBigInt> quotient;
    487   bool result_sign = x->sign() != y->sign();
    488   if (y->length() == 1) {
    489     digit_t divisor = y->digit(0);
    490     if (divisor == 1) {
    491       return result_sign == x->sign() ? x : UnaryMinus(isolate, x);
    492     }
    493     digit_t remainder;
    494     MutableBigInt::AbsoluteDivSmall(isolate, x, divisor, &quotient, &remainder);
    495   } else {
    496     if (!MutableBigInt::AbsoluteDivLarge(isolate, x, y, &quotient, nullptr)) {
    497       return MaybeHandle<BigInt>();
    498     }
    499   }
    500   quotient->set_sign(x->sign() != y->sign());
    501   return MutableBigInt::MakeImmutable(quotient);
    502 }
    503 
    504 MaybeHandle<BigInt> BigInt::Remainder(Isolate* isolate, Handle<BigInt> x,
    505                                       Handle<BigInt> y) {
    506   // 1. If y is 0n, throw a RangeError exception.
    507   if (y->is_zero()) {
    508     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntDivZero),
    509                     BigInt);
    510   }
    511   // 2. Return the BigInt representing x modulo y.
    512   // See https://github.com/tc39/proposal-bigint/issues/84 though.
    513   if (MutableBigInt::AbsoluteCompare(x, y) < 0) return x;
    514   Handle<MutableBigInt> remainder;
    515   if (y->length() == 1) {
    516     digit_t divisor = y->digit(0);
    517     if (divisor == 1) return Zero(isolate);
    518     digit_t remainder_digit;
    519     MutableBigInt::AbsoluteDivSmall(isolate, x, divisor, nullptr,
    520                                     &remainder_digit);
    521     if (remainder_digit == 0) {
    522       return Zero(isolate);
    523     }
    524     remainder = MutableBigInt::New(isolate, 1).ToHandleChecked();
    525     remainder->set_digit(0, remainder_digit);
    526   } else {
    527     if (!MutableBigInt::AbsoluteDivLarge(isolate, x, y, nullptr, &remainder)) {
    528       return MaybeHandle<BigInt>();
    529     }
    530   }
    531   remainder->set_sign(x->sign());
    532   return MutableBigInt::MakeImmutable(remainder);
    533 }
    534 
    535 MaybeHandle<BigInt> BigInt::Add(Isolate* isolate, Handle<BigInt> x,
    536                                 Handle<BigInt> y) {
    537   bool xsign = x->sign();
    538   if (xsign == y->sign()) {
    539     // x + y == x + y
    540     // -x + -y == -(x + y)
    541     return MutableBigInt::AbsoluteAdd(isolate, x, y, xsign);
    542   }
    543   // x + -y == x - y == -(y - x)
    544   // -x + y == y - x == -(x - y)
    545   if (MutableBigInt::AbsoluteCompare(x, y) >= 0) {
    546     return MutableBigInt::AbsoluteSub(isolate, x, y, xsign);
    547   }
    548   return MutableBigInt::AbsoluteSub(isolate, y, x, !xsign);
    549 }
    550 
    551 MaybeHandle<BigInt> BigInt::Subtract(Isolate* isolate, Handle<BigInt> x,
    552                                      Handle<BigInt> y) {
    553   bool xsign = x->sign();
    554   if (xsign != y->sign()) {
    555     // x - (-y) == x + y
    556     // (-x) - y == -(x + y)
    557     return MutableBigInt::AbsoluteAdd(isolate, x, y, xsign);
    558   }
    559   // x - y == -(y - x)
    560   // (-x) - (-y) == y - x == -(x - y)
    561   if (MutableBigInt::AbsoluteCompare(x, y) >= 0) {
    562     return MutableBigInt::AbsoluteSub(isolate, x, y, xsign);
    563   }
    564   return MutableBigInt::AbsoluteSub(isolate, y, x, !xsign);
    565 }
    566 
    567 MaybeHandle<BigInt> BigInt::LeftShift(Isolate* isolate, Handle<BigInt> x,
    568                                       Handle<BigInt> y) {
    569   if (y->is_zero() || x->is_zero()) return x;
    570   if (y->sign()) return MutableBigInt::RightShiftByAbsolute(isolate, x, y);
    571   return MutableBigInt::LeftShiftByAbsolute(isolate, x, y);
    572 }
    573 
    574 MaybeHandle<BigInt> BigInt::SignedRightShift(Isolate* isolate, Handle<BigInt> x,
    575                                              Handle<BigInt> y) {
    576   if (y->is_zero() || x->is_zero()) return x;
    577   if (y->sign()) return MutableBigInt::LeftShiftByAbsolute(isolate, x, y);
    578   return MutableBigInt::RightShiftByAbsolute(isolate, x, y);
    579 }
    580 
    581 MaybeHandle<BigInt> BigInt::UnsignedRightShift(Isolate* isolate,
    582                                                Handle<BigInt> x,
    583                                                Handle<BigInt> y) {
    584   THROW_NEW_ERROR(isolate, NewTypeError(MessageTemplate::kBigIntShr), BigInt);
    585 }
    586 
    587 namespace {
    588 
    589 // Produces comparison result for {left_negative} == sign(x) != sign(y).
    590 ComparisonResult UnequalSign(bool left_negative) {
    591   return left_negative ? ComparisonResult::kLessThan
    592                        : ComparisonResult::kGreaterThan;
    593 }
    594 
    595 // Produces result for |x| > |y|, with {both_negative} == sign(x) == sign(y);
    596 ComparisonResult AbsoluteGreater(bool both_negative) {
    597   return both_negative ? ComparisonResult::kLessThan
    598                        : ComparisonResult::kGreaterThan;
    599 }
    600 
    601 // Produces result for |x| < |y|, with {both_negative} == sign(x) == sign(y).
    602 ComparisonResult AbsoluteLess(bool both_negative) {
    603   return both_negative ? ComparisonResult::kGreaterThan
    604                        : ComparisonResult::kLessThan;
    605 }
    606 
    607 }  // namespace
    608 
    609 // (Never returns kUndefined.)
    610 ComparisonResult BigInt::CompareToBigInt(Handle<BigInt> x, Handle<BigInt> y) {
    611   bool x_sign = x->sign();
    612   if (x_sign != y->sign()) return UnequalSign(x_sign);
    613 
    614   int result = MutableBigInt::AbsoluteCompare(x, y);
    615   if (result > 0) return AbsoluteGreater(x_sign);
    616   if (result < 0) return AbsoluteLess(x_sign);
    617   return ComparisonResult::kEqual;
    618 }
    619 
    620 bool BigInt::EqualToBigInt(BigInt* x, BigInt* y) {
    621   if (x->sign() != y->sign()) return false;
    622   if (x->length() != y->length()) return false;
    623   for (int i = 0; i < x->length(); i++) {
    624     if (x->digit(i) != y->digit(i)) return false;
    625   }
    626   return true;
    627 }
    628 
    629 MaybeHandle<BigInt> BigInt::BitwiseAnd(Isolate* isolate, Handle<BigInt> x,
    630                                        Handle<BigInt> y) {
    631   return MutableBigInt::MakeImmutable(MutableBigInt::BitwiseAnd(isolate, x, y));
    632 }
    633 
    634 MaybeHandle<MutableBigInt> MutableBigInt::BitwiseAnd(Isolate* isolate,
    635                                                      Handle<BigInt> x,
    636                                                      Handle<BigInt> y) {
    637   if (!x->sign() && !y->sign()) {
    638     return AbsoluteAnd(isolate, x, y);
    639   } else if (x->sign() && y->sign()) {
    640     int result_length = Max(x->length(), y->length()) + 1;
    641     // (-x) & (-y) == ~(x-1) & ~(y-1) == ~((x-1) | (y-1))
    642     // == -(((x-1) | (y-1)) + 1)
    643     Handle<MutableBigInt> result;
    644     if (!AbsoluteSubOne(isolate, x, result_length).ToHandle(&result)) {
    645       return MaybeHandle<MutableBigInt>();
    646     }
    647     Handle<MutableBigInt> y_1 = AbsoluteSubOne(isolate, y);
    648     result = AbsoluteOr(isolate, result, y_1, *result);
    649     return AbsoluteAddOne(isolate, result, true, *result);
    650   } else {
    651     DCHECK(x->sign() != y->sign());
    652     // Assume that x is the positive BigInt.
    653     if (x->sign()) std::swap(x, y);
    654     // x & (-y) == x & ~(y-1) == x &~ (y-1)
    655     return AbsoluteAndNot(isolate, x, AbsoluteSubOne(isolate, y));
    656   }
    657 }
    658 
    659 MaybeHandle<BigInt> BigInt::BitwiseXor(Isolate* isolate, Handle<BigInt> x,
    660                                        Handle<BigInt> y) {
    661   return MutableBigInt::MakeImmutable(MutableBigInt::BitwiseXor(isolate, x, y));
    662 }
    663 
    664 MaybeHandle<MutableBigInt> MutableBigInt::BitwiseXor(Isolate* isolate,
    665                                                      Handle<BigInt> x,
    666                                                      Handle<BigInt> y) {
    667   if (!x->sign() && !y->sign()) {
    668     return AbsoluteXor(isolate, x, y);
    669   } else if (x->sign() && y->sign()) {
    670     int result_length = Max(x->length(), y->length());
    671     // (-x) ^ (-y) == ~(x-1) ^ ~(y-1) == (x-1) ^ (y-1)
    672     Handle<MutableBigInt> result =
    673         AbsoluteSubOne(isolate, x, result_length).ToHandleChecked();
    674     Handle<MutableBigInt> y_1 = AbsoluteSubOne(isolate, y);
    675     return AbsoluteXor(isolate, result, y_1, *result);
    676   } else {
    677     DCHECK(x->sign() != y->sign());
    678     int result_length = Max(x->length(), y->length()) + 1;
    679     // Assume that x is the positive BigInt.
    680     if (x->sign()) std::swap(x, y);
    681     // x ^ (-y) == x ^ ~(y-1) == ~(x ^ (y-1)) == -((x ^ (y-1)) + 1)
    682     Handle<MutableBigInt> result;
    683     if (!AbsoluteSubOne(isolate, y, result_length).ToHandle(&result)) {
    684       return MaybeHandle<MutableBigInt>();
    685     }
    686     result = AbsoluteXor(isolate, result, x, *result);
    687     return AbsoluteAddOne(isolate, result, true, *result);
    688   }
    689 }
    690 
    691 MaybeHandle<BigInt> BigInt::BitwiseOr(Isolate* isolate, Handle<BigInt> x,
    692                                       Handle<BigInt> y) {
    693   return MutableBigInt::MakeImmutable(MutableBigInt::BitwiseOr(isolate, x, y));
    694 }
    695 
    696 MaybeHandle<MutableBigInt> MutableBigInt::BitwiseOr(Isolate* isolate,
    697                                                     Handle<BigInt> x,
    698                                                     Handle<BigInt> y) {
    699   int result_length = Max(x->length(), y->length());
    700   if (!x->sign() && !y->sign()) {
    701     return AbsoluteOr(isolate, x, y);
    702   } else if (x->sign() && y->sign()) {
    703     // (-x) | (-y) == ~(x-1) | ~(y-1) == ~((x-1) & (y-1))
    704     // == -(((x-1) & (y-1)) + 1)
    705     Handle<MutableBigInt> result =
    706         AbsoluteSubOne(isolate, x, result_length).ToHandleChecked();
    707     Handle<MutableBigInt> y_1 = AbsoluteSubOne(isolate, y);
    708     result = AbsoluteAnd(isolate, result, y_1, *result);
    709     return AbsoluteAddOne(isolate, result, true, *result);
    710   } else {
    711     DCHECK(x->sign() != y->sign());
    712     // Assume that x is the positive BigInt.
    713     if (x->sign()) std::swap(x, y);
    714     // x | (-y) == x | ~(y-1) == ~((y-1) &~ x) == -(((y-1) &~ x) + 1)
    715     Handle<MutableBigInt> result =
    716         AbsoluteSubOne(isolate, y, result_length).ToHandleChecked();
    717     result = AbsoluteAndNot(isolate, result, x, *result);
    718     return AbsoluteAddOne(isolate, result, true, *result);
    719   }
    720 }
    721 
    722 MaybeHandle<BigInt> BigInt::Increment(Isolate* isolate, Handle<BigInt> x) {
    723   if (x->sign()) {
    724     Handle<MutableBigInt> result = MutableBigInt::AbsoluteSubOne(isolate, x);
    725     result->set_sign(true);
    726     return MutableBigInt::MakeImmutable(result);
    727   } else {
    728     return MutableBigInt::MakeImmutable(
    729         MutableBigInt::AbsoluteAddOne(isolate, x, false));
    730   }
    731 }
    732 
    733 MaybeHandle<BigInt> BigInt::Decrement(Isolate* isolate, Handle<BigInt> x) {
    734   MaybeHandle<MutableBigInt> result;
    735   if (x->sign()) {
    736     result = MutableBigInt::AbsoluteAddOne(isolate, x, true);
    737   } else if (x->is_zero()) {
    738     // TODO(jkummerow): Consider caching a canonical -1n BigInt.
    739     return MutableBigInt::NewFromInt(isolate, -1);
    740   } else {
    741     result = MutableBigInt::AbsoluteSubOne(isolate, x);
    742   }
    743   return MutableBigInt::MakeImmutable(result);
    744 }
    745 
    746 ComparisonResult BigInt::CompareToString(Isolate* isolate, Handle<BigInt> x,
    747                                          Handle<String> y) {
    748   // a. Let ny be StringToBigInt(y);
    749   MaybeHandle<BigInt> maybe_ny = StringToBigInt(isolate, y);
    750   // b. If ny is NaN, return undefined.
    751   Handle<BigInt> ny;
    752   if (!maybe_ny.ToHandle(&ny)) {
    753     DCHECK(!isolate->has_pending_exception());
    754     return ComparisonResult::kUndefined;
    755   }
    756   // c. Return BigInt::lessThan(x, ny).
    757   return CompareToBigInt(x, ny);
    758 }
    759 
    760 bool BigInt::EqualToString(Isolate* isolate, Handle<BigInt> x,
    761                            Handle<String> y) {
    762   // a. Let n be StringToBigInt(y).
    763   MaybeHandle<BigInt> maybe_n = StringToBigInt(isolate, y);
    764   // b. If n is NaN, return false.
    765   Handle<BigInt> n;
    766   if (!maybe_n.ToHandle(&n)) {
    767     DCHECK(!isolate->has_pending_exception());
    768     return false;
    769   }
    770   // c. Return the result of x == n.
    771   return EqualToBigInt(*x, *n);
    772 }
    773 
    774 bool BigInt::EqualToNumber(Handle<BigInt> x, Handle<Object> y) {
    775   DCHECK(y->IsNumber());
    776   // a. If x or y are any of NaN, +, or -, return false.
    777   // b. If the mathematical value of x is equal to the mathematical value of y,
    778   //    return true, otherwise return false.
    779   if (y->IsSmi()) {
    780     int value = Smi::ToInt(*y);
    781     if (value == 0) return x->is_zero();
    782     // Any multi-digit BigInt is bigger than a Smi.
    783     STATIC_ASSERT(sizeof(digit_t) >= sizeof(value));
    784     return (x->length() == 1) && (x->sign() == (value < 0)) &&
    785            (x->digit(0) ==
    786             static_cast<digit_t>(std::abs(static_cast<int64_t>(value))));
    787   }
    788   DCHECK(y->IsHeapNumber());
    789   double value = Handle<HeapNumber>::cast(y)->value();
    790   return CompareToDouble(x, value) == ComparisonResult::kEqual;
    791 }
    792 
    793 ComparisonResult BigInt::CompareToNumber(Handle<BigInt> x, Handle<Object> y) {
    794   DCHECK(y->IsNumber());
    795   if (y->IsSmi()) {
    796     bool x_sign = x->sign();
    797     int y_value = Smi::ToInt(*y);
    798     bool y_sign = (y_value < 0);
    799     if (x_sign != y_sign) return UnequalSign(x_sign);
    800 
    801     if (x->is_zero()) {
    802       DCHECK(!y_sign);
    803       return y_value == 0 ? ComparisonResult::kEqual
    804                           : ComparisonResult::kLessThan;
    805     }
    806     // Any multi-digit BigInt is bigger than a Smi.
    807     STATIC_ASSERT(sizeof(digit_t) >= sizeof(y_value));
    808     if (x->length() > 1) return AbsoluteGreater(x_sign);
    809 
    810     digit_t abs_value = std::abs(static_cast<int64_t>(y_value));
    811     digit_t x_digit = x->digit(0);
    812     if (x_digit > abs_value) return AbsoluteGreater(x_sign);
    813     if (x_digit < abs_value) return AbsoluteLess(x_sign);
    814     return ComparisonResult::kEqual;
    815   }
    816   DCHECK(y->IsHeapNumber());
    817   double value = Handle<HeapNumber>::cast(y)->value();
    818   return CompareToDouble(x, value);
    819 }
    820 
    821 ComparisonResult BigInt::CompareToDouble(Handle<BigInt> x, double y) {
    822   if (std::isnan(y)) return ComparisonResult::kUndefined;
    823   if (y == V8_INFINITY) return ComparisonResult::kLessThan;
    824   if (y == -V8_INFINITY) return ComparisonResult::kGreaterThan;
    825   bool x_sign = x->sign();
    826   // Note that this is different from the double's sign bit for -0. That's
    827   // intentional because -0 must be treated like 0.
    828   bool y_sign = (y < 0);
    829   if (x_sign != y_sign) return UnequalSign(x_sign);
    830   if (y == 0) {
    831     DCHECK(!x_sign);
    832     return x->is_zero() ? ComparisonResult::kEqual
    833                         : ComparisonResult::kGreaterThan;
    834   }
    835   if (x->is_zero()) {
    836     DCHECK(!y_sign);
    837     return ComparisonResult::kLessThan;
    838   }
    839   uint64_t double_bits = bit_cast<uint64_t>(y);
    840   int raw_exponent =
    841       static_cast<int>(double_bits >> Double::kPhysicalSignificandSize) & 0x7FF;
    842   uint64_t mantissa = double_bits & Double::kSignificandMask;
    843   // Non-finite doubles are handled above.
    844   DCHECK_NE(raw_exponent, 0x7FF);
    845   int exponent = raw_exponent - 0x3FF;
    846   if (exponent < 0) {
    847     // The absolute value of the double is less than 1. Only 0n has an
    848     // absolute value smaller than that, but we've already covered that case.
    849     DCHECK(!x->is_zero());
    850     return AbsoluteGreater(x_sign);
    851   }
    852   int x_length = x->length();
    853   digit_t x_msd = x->digit(x_length - 1);
    854   int msd_leading_zeros = base::bits::CountLeadingZeros(x_msd);
    855   int x_bitlength = x_length * kDigitBits - msd_leading_zeros;
    856   int y_bitlength = exponent + 1;
    857   if (x_bitlength < y_bitlength) return AbsoluteLess(x_sign);
    858   if (x_bitlength > y_bitlength) return AbsoluteGreater(x_sign);
    859 
    860   // At this point, we know that signs and bit lengths (i.e. position of
    861   // the most significant bit in exponent-free representation) are identical.
    862   // {x} is not zero, {y} is finite and not denormal.
    863   // Now we virtually convert the double to an integer by shifting its
    864   // mantissa according to its exponent, so it will align with the BigInt {x},
    865   // and then we compare them bit for bit until we find a difference or the
    866   // least significant bit.
    867   //                    <----- 52 ------> <-- virtual trailing zeroes -->
    868   // y / mantissa:     1yyyyyyyyyyyyyyyyy 0000000000000000000000000000000
    869   // x / digits:    0001xxxx xxxxxxxx xxxxxxxx ...
    870   //                    <-->          <------>
    871   //              msd_topbit         kDigitBits
    872   //
    873   mantissa |= Double::kHiddenBit;
    874   const int kMantissaTopBit = 52;  // 0-indexed.
    875   // 0-indexed position of {x}'s most significant bit within the {msd}.
    876   int msd_topbit = kDigitBits - 1 - msd_leading_zeros;
    877   DCHECK_EQ(msd_topbit, (x_bitlength - 1) % kDigitBits);
    878   // Shifted chunk of {mantissa} for comparing with {digit}.
    879   digit_t compare_mantissa;
    880   // Number of unprocessed bits in {mantissa}. We'll keep them shifted to
    881   // the left (i.e. most significant part) of the underlying uint64_t.
    882   int remaining_mantissa_bits = 0;
    883 
    884   // First, compare the most significant digit against the beginning of
    885   // the mantissa.
    886   if (msd_topbit < kMantissaTopBit) {
    887     remaining_mantissa_bits = (kMantissaTopBit - msd_topbit);
    888     compare_mantissa = mantissa >> remaining_mantissa_bits;
    889     mantissa = mantissa << (64 - remaining_mantissa_bits);
    890   } else {
    891     DCHECK_GE(msd_topbit, kMantissaTopBit);
    892     compare_mantissa = mantissa << (msd_topbit - kMantissaTopBit);
    893     mantissa = 0;
    894   }
    895   if (x_msd > compare_mantissa) return AbsoluteGreater(x_sign);
    896   if (x_msd < compare_mantissa) return AbsoluteLess(x_sign);
    897 
    898   // Then, compare additional digits against any remaining mantissa bits.
    899   for (int digit_index = x_length - 2; digit_index >= 0; digit_index--) {
    900     if (remaining_mantissa_bits > 0) {
    901       remaining_mantissa_bits -= kDigitBits;
    902       if (sizeof(mantissa) != sizeof(x_msd)) {
    903         compare_mantissa = mantissa >> (64 - kDigitBits);
    904         // "& 63" to appease compilers. kDigitBits is 32 here anyway.
    905         mantissa = mantissa << (kDigitBits & 63);
    906       } else {
    907         compare_mantissa = mantissa;
    908         mantissa = 0;
    909       }
    910     } else {
    911       compare_mantissa = 0;
    912     }
    913     digit_t digit = x->digit(digit_index);
    914     if (digit > compare_mantissa) return AbsoluteGreater(x_sign);
    915     if (digit < compare_mantissa) return AbsoluteLess(x_sign);
    916   }
    917 
    918   // Integer parts are equal; check whether {y} has a fractional part.
    919   if (mantissa != 0) {
    920     DCHECK_GT(remaining_mantissa_bits, 0);
    921     return AbsoluteLess(x_sign);
    922   }
    923   return ComparisonResult::kEqual;
    924 }
    925 
    926 MaybeHandle<String> BigInt::ToString(Isolate* isolate, Handle<BigInt> bigint,
    927                                      int radix) {
    928   if (bigint->is_zero()) {
    929     return isolate->factory()->NewStringFromStaticChars("0");
    930   }
    931   if (base::bits::IsPowerOfTwo(radix)) {
    932     return MutableBigInt::ToStringBasePowerOfTwo(isolate, bigint, radix);
    933   }
    934   return MutableBigInt::ToStringGeneric(isolate, bigint, radix);
    935 }
    936 
    937 MaybeHandle<BigInt> BigInt::FromNumber(Isolate* isolate,
    938                                        Handle<Object> number) {
    939   DCHECK(number->IsNumber());
    940   if (number->IsSmi()) {
    941     return MutableBigInt::NewFromInt(isolate, Smi::ToInt(*number));
    942   }
    943   double value = HeapNumber::cast(*number)->value();
    944   if (!std::isfinite(value) || (DoubleToInteger(value) != value)) {
    945     THROW_NEW_ERROR(isolate,
    946                     NewRangeError(MessageTemplate::kBigIntFromNumber, number),
    947                     BigInt);
    948   }
    949   return MutableBigInt::NewFromDouble(isolate, value);
    950 }
    951 
    952 MaybeHandle<BigInt> BigInt::FromObject(Isolate* isolate, Handle<Object> obj) {
    953   if (obj->IsJSReceiver()) {
    954     ASSIGN_RETURN_ON_EXCEPTION(
    955         isolate, obj,
    956         JSReceiver::ToPrimitive(Handle<JSReceiver>::cast(obj),
    957                                 ToPrimitiveHint::kNumber),
    958         BigInt);
    959   }
    960 
    961   if (obj->IsBoolean()) {
    962     return MutableBigInt::NewFromInt(isolate, obj->BooleanValue(isolate));
    963   }
    964   if (obj->IsBigInt()) {
    965     return Handle<BigInt>::cast(obj);
    966   }
    967   if (obj->IsString()) {
    968     Handle<BigInt> n;
    969     if (!StringToBigInt(isolate, Handle<String>::cast(obj)).ToHandle(&n)) {
    970       THROW_NEW_ERROR(isolate,
    971                       NewSyntaxError(MessageTemplate::kBigIntFromObject, obj),
    972                       BigInt);
    973     }
    974     return n;
    975   }
    976 
    977   THROW_NEW_ERROR(
    978       isolate, NewTypeError(MessageTemplate::kBigIntFromObject, obj), BigInt);
    979 }
    980 
    981 Handle<Object> BigInt::ToNumber(Isolate* isolate, Handle<BigInt> x) {
    982   if (x->is_zero()) return Handle<Smi>(Smi::kZero, isolate);
    983   if (x->length() == 1 && x->digit(0) < Smi::kMaxValue) {
    984     int value = static_cast<int>(x->digit(0));
    985     if (x->sign()) value = -value;
    986     return Handle<Smi>(Smi::FromInt(value), isolate);
    987   }
    988   double result = MutableBigInt::ToDouble(x);
    989   return isolate->factory()->NewHeapNumber(result);
    990 }
    991 
    992 double MutableBigInt::ToDouble(Handle<BigIntBase> x) {
    993   if (x->is_zero()) return 0.0;
    994   int x_length = x->length();
    995   digit_t x_msd = x->digit(x_length - 1);
    996   int msd_leading_zeros = base::bits::CountLeadingZeros(x_msd);
    997   int x_bitlength = x_length * kDigitBits - msd_leading_zeros;
    998   if (x_bitlength > 1024) return x->sign() ? -V8_INFINITY : V8_INFINITY;
    999   uint64_t exponent = x_bitlength - 1;
   1000   // We need the most significant bit shifted to the position of a double's
   1001   // "hidden bit". We also need to hide that MSB, so we shift it out.
   1002   uint64_t current_digit = x_msd;
   1003   int digit_index = x_length - 1;
   1004   int shift = msd_leading_zeros + 1 + (64 - kDigitBits);
   1005   DCHECK_LE(1, shift);
   1006   DCHECK_LE(shift, 64);
   1007   uint64_t mantissa = (shift == 64) ? 0 : current_digit << shift;
   1008   mantissa >>= 12;
   1009   int mantissa_bits_unset = shift - 12;
   1010   // If not all mantissa bits are defined yet, get more digits as needed.
   1011   if (mantissa_bits_unset >= kDigitBits && digit_index > 0) {
   1012     digit_index--;
   1013     current_digit = static_cast<uint64_t>(x->digit(digit_index));
   1014     mantissa |= (current_digit << (mantissa_bits_unset - kDigitBits));
   1015     mantissa_bits_unset -= kDigitBits;
   1016   }
   1017   if (mantissa_bits_unset > 0 && digit_index > 0) {
   1018     DCHECK_LT(mantissa_bits_unset, kDigitBits);
   1019     digit_index--;
   1020     current_digit = static_cast<uint64_t>(x->digit(digit_index));
   1021     mantissa |= (current_digit >> (kDigitBits - mantissa_bits_unset));
   1022     mantissa_bits_unset -= kDigitBits;
   1023   }
   1024   // If there are unconsumed digits left, we may have to round.
   1025   Rounding rounding =
   1026       DecideRounding(x, mantissa_bits_unset, digit_index, current_digit);
   1027   if (rounding == kRoundUp || (rounding == kTie && (mantissa & 1) == 1)) {
   1028     mantissa++;
   1029     // Incrementing the mantissa can overflow the mantissa bits. In that case
   1030     // the new mantissa will be all zero (plus hidden bit).
   1031     if ((mantissa >> Double::kPhysicalSignificandSize) != 0) {
   1032       mantissa = 0;
   1033       exponent++;
   1034       // Incrementing the exponent can overflow too.
   1035       if (exponent > 1023) {
   1036         return x->sign() ? -V8_INFINITY : V8_INFINITY;
   1037       }
   1038     }
   1039   }
   1040   // Assemble the result.
   1041   uint64_t sign_bit = x->sign() ? (static_cast<uint64_t>(1) << 63) : 0;
   1042   exponent = (exponent + 0x3FF) << Double::kPhysicalSignificandSize;
   1043   uint64_t double_bits = sign_bit | exponent | mantissa;
   1044   return bit_cast<double>(double_bits);
   1045 }
   1046 
   1047 // This is its own function to keep control flow sane. The meaning of the
   1048 // parameters is defined by {ToDouble}'s local variable usage.
   1049 MutableBigInt::Rounding MutableBigInt::DecideRounding(Handle<BigIntBase> x,
   1050                                                       int mantissa_bits_unset,
   1051                                                       int digit_index,
   1052                                                       uint64_t current_digit) {
   1053   if (mantissa_bits_unset > 0) return kRoundDown;
   1054   int top_unconsumed_bit;
   1055   if (mantissa_bits_unset < 0) {
   1056     // There are unconsumed bits in {current_digit}.
   1057     top_unconsumed_bit = -mantissa_bits_unset - 1;
   1058   } else {
   1059     DCHECK_EQ(mantissa_bits_unset, 0);
   1060     // {current_digit} fit the mantissa exactly; look at the next digit.
   1061     if (digit_index == 0) return kRoundDown;
   1062     digit_index--;
   1063     current_digit = static_cast<uint64_t>(x->digit(digit_index));
   1064     top_unconsumed_bit = kDigitBits - 1;
   1065   }
   1066   // If the most significant remaining bit is 0, round down.
   1067   uint64_t bitmask = static_cast<uint64_t>(1) << top_unconsumed_bit;
   1068   if ((current_digit & bitmask) == 0) {
   1069     return kRoundDown;
   1070   }
   1071   // If any other remaining bit is set, round up.
   1072   bitmask -= 1;
   1073   if ((current_digit & bitmask) != 0) return kRoundUp;
   1074   while (digit_index > 0) {
   1075     digit_index--;
   1076     if (x->digit(digit_index) != 0) return kRoundUp;
   1077   }
   1078   return kTie;
   1079 }
   1080 
   1081 void BigInt::BigIntShortPrint(std::ostream& os) {
   1082   if (sign()) os << "-";
   1083   int len = length();
   1084   if (len == 0) {
   1085     os << "0";
   1086     return;
   1087   }
   1088   if (len > 1) os << "...";
   1089   os << digit(0);
   1090 }
   1091 
   1092 // Internal helpers.
   1093 
   1094 MaybeHandle<BigInt> MutableBigInt::AbsoluteAdd(Isolate* isolate,
   1095                                                Handle<BigInt> x,
   1096                                                Handle<BigInt> y,
   1097                                                bool result_sign) {
   1098   if (x->length() < y->length()) return AbsoluteAdd(isolate, y, x, result_sign);
   1099   if (x->is_zero()) {
   1100     DCHECK(y->is_zero());
   1101     return x;
   1102   }
   1103   if (y->is_zero()) {
   1104     return result_sign == x->sign() ? x : BigInt::UnaryMinus(isolate, x);
   1105   }
   1106   Handle<MutableBigInt> result;
   1107   if (!New(isolate, x->length() + 1).ToHandle(&result)) {
   1108     return MaybeHandle<BigInt>();
   1109   }
   1110   digit_t carry = 0;
   1111   int i = 0;
   1112   for (; i < y->length(); i++) {
   1113     digit_t new_carry = 0;
   1114     digit_t sum = digit_add(x->digit(i), y->digit(i), &new_carry);
   1115     sum = digit_add(sum, carry, &new_carry);
   1116     result->set_digit(i, sum);
   1117     carry = new_carry;
   1118   }
   1119   for (; i < x->length(); i++) {
   1120     digit_t new_carry = 0;
   1121     digit_t sum = digit_add(x->digit(i), carry, &new_carry);
   1122     result->set_digit(i, sum);
   1123     carry = new_carry;
   1124   }
   1125   result->set_digit(i, carry);
   1126   result->set_sign(result_sign);
   1127   return MakeImmutable(result);
   1128 }
   1129 
   1130 Handle<BigInt> MutableBigInt::AbsoluteSub(Isolate* isolate, Handle<BigInt> x,
   1131                                           Handle<BigInt> y, bool result_sign) {
   1132   DCHECK(x->length() >= y->length());
   1133   SLOW_DCHECK(AbsoluteCompare(x, y) >= 0);
   1134   if (x->is_zero()) {
   1135     DCHECK(y->is_zero());
   1136     return x;
   1137   }
   1138   if (y->is_zero()) {
   1139     return result_sign == x->sign() ? x : BigInt::UnaryMinus(isolate, x);
   1140   }
   1141   Handle<MutableBigInt> result = New(isolate, x->length()).ToHandleChecked();
   1142   digit_t borrow = 0;
   1143   int i = 0;
   1144   for (; i < y->length(); i++) {
   1145     digit_t new_borrow = 0;
   1146     digit_t difference = digit_sub(x->digit(i), y->digit(i), &new_borrow);
   1147     difference = digit_sub(difference, borrow, &new_borrow);
   1148     result->set_digit(i, difference);
   1149     borrow = new_borrow;
   1150   }
   1151   for (; i < x->length(); i++) {
   1152     digit_t new_borrow = 0;
   1153     digit_t difference = digit_sub(x->digit(i), borrow, &new_borrow);
   1154     result->set_digit(i, difference);
   1155     borrow = new_borrow;
   1156   }
   1157   DCHECK_EQ(0, borrow);
   1158   result->set_sign(result_sign);
   1159   return MakeImmutable(result);
   1160 }
   1161 
   1162 // Adds 1 to the absolute value of {x} and sets the result's sign to {sign}.
   1163 // {result_storage} is optional; if present, it will be used to store the
   1164 // result, otherwise a new BigInt will be allocated for the result.
   1165 // {result_storage} and {x} may refer to the same BigInt for in-place
   1166 // modification.
   1167 MaybeHandle<MutableBigInt> MutableBigInt::AbsoluteAddOne(
   1168     Isolate* isolate, Handle<BigIntBase> x, bool sign,
   1169     MutableBigInt* result_storage) {
   1170   int input_length = x->length();
   1171   // The addition will overflow into a new digit if all existing digits are
   1172   // at maximum.
   1173   bool will_overflow = true;
   1174   for (int i = 0; i < input_length; i++) {
   1175     if (!digit_ismax(x->digit(i))) {
   1176       will_overflow = false;
   1177       break;
   1178     }
   1179   }
   1180   int result_length = input_length + will_overflow;
   1181   Handle<MutableBigInt> result(result_storage, isolate);
   1182   if (result_storage == nullptr) {
   1183     if (!New(isolate, result_length).ToHandle(&result)) {
   1184       return MaybeHandle<MutableBigInt>();
   1185     }
   1186   } else {
   1187     DCHECK(result->length() == result_length);
   1188   }
   1189   digit_t carry = 1;
   1190   for (int i = 0; i < input_length; i++) {
   1191     digit_t new_carry = 0;
   1192     result->set_digit(i, digit_add(x->digit(i), carry, &new_carry));
   1193     carry = new_carry;
   1194   }
   1195   if (result_length > input_length) {
   1196     result->set_digit(input_length, carry);
   1197   } else {
   1198     DCHECK_EQ(carry, 0);
   1199   }
   1200   result->set_sign(sign);
   1201   return result;
   1202 }
   1203 
   1204 // Subtracts 1 from the absolute value of {x}. {x} must not be zero.
   1205 Handle<MutableBigInt> MutableBigInt::AbsoluteSubOne(Isolate* isolate,
   1206                                                     Handle<BigIntBase> x) {
   1207   DCHECK(!x->is_zero());
   1208   // Requesting a result length identical to an existing BigInt's length
   1209   // cannot overflow the limit.
   1210   return AbsoluteSubOne(isolate, x, x->length()).ToHandleChecked();
   1211 }
   1212 
   1213 // Like the above, but you can specify that the allocated result should have
   1214 // length {result_length}, which must be at least as large as {x->length()}.
   1215 MaybeHandle<MutableBigInt> MutableBigInt::AbsoluteSubOne(Isolate* isolate,
   1216                                                          Handle<BigIntBase> x,
   1217                                                          int result_length) {
   1218   DCHECK(!x->is_zero());
   1219   DCHECK(result_length >= x->length());
   1220   Handle<MutableBigInt> result;
   1221   if (!New(isolate, result_length).ToHandle(&result)) {
   1222     return MaybeHandle<MutableBigInt>();
   1223   }
   1224   int length = x->length();
   1225   digit_t borrow = 1;
   1226   for (int i = 0; i < length; i++) {
   1227     digit_t new_borrow = 0;
   1228     result->set_digit(i, digit_sub(x->digit(i), borrow, &new_borrow));
   1229     borrow = new_borrow;
   1230   }
   1231   DCHECK_EQ(borrow, 0);
   1232   for (int i = length; i < result_length; i++) {
   1233     result->set_digit(i, borrow);
   1234   }
   1235   return result;
   1236 }
   1237 
   1238 // Helper for Absolute{And,AndNot,Or,Xor}.
   1239 // Performs the given binary {op} on digit pairs of {x} and {y}; when the
   1240 // end of the shorter of the two is reached, {extra_digits} configures how
   1241 // remaining digits in the longer input (if {symmetric} == kSymmetric, in
   1242 // {x} otherwise) are handled: copied to the result or ignored.
   1243 // If {result_storage} is non-nullptr, it will be used for the result and
   1244 // any extra digits in it will be zeroed out, otherwise a new BigInt (with
   1245 // the same length as the longer input) will be allocated.
   1246 // {result_storage} may alias {x} or {y} for in-place modification.
   1247 // Example:
   1248 //              y:             [ y2 ][ y1 ][ y0 ]
   1249 //              x:       [ x3 ][ x2 ][ x1 ][ x0 ]
   1250 //                          |     |     |     |
   1251 //                      (kCopy)  (op)  (op)  (op)
   1252 //                          |     |     |     |
   1253 //                          v     v     v     v
   1254 // result_storage: [  0 ][ x3 ][ r2 ][ r1 ][ r0 ]
   1255 inline Handle<MutableBigInt> MutableBigInt::AbsoluteBitwiseOp(
   1256     Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
   1257     MutableBigInt* result_storage, ExtraDigitsHandling extra_digits,
   1258     SymmetricOp symmetric, std::function<digit_t(digit_t, digit_t)> op) {
   1259   int x_length = x->length();
   1260   int y_length = y->length();
   1261   int num_pairs = y_length;
   1262   if (x_length < y_length) {
   1263     num_pairs = x_length;
   1264     if (symmetric == kSymmetric) {
   1265       std::swap(x, y);
   1266       std::swap(x_length, y_length);
   1267     }
   1268   }
   1269   DCHECK(num_pairs == Min(x_length, y_length));
   1270   Handle<MutableBigInt> result(result_storage, isolate);
   1271   int result_length = extra_digits == kCopy ? x_length : num_pairs;
   1272   if (result_storage == nullptr) {
   1273     result = New(isolate, result_length).ToHandleChecked();
   1274   } else {
   1275     DCHECK(result_storage->length() >= result_length);
   1276     result_length = result_storage->length();
   1277   }
   1278   int i = 0;
   1279   for (; i < num_pairs; i++) {
   1280     result->set_digit(i, op(x->digit(i), y->digit(i)));
   1281   }
   1282   if (extra_digits == kCopy) {
   1283     for (; i < x_length; i++) {
   1284       result->set_digit(i, x->digit(i));
   1285     }
   1286   }
   1287   for (; i < result_length; i++) {
   1288     result->set_digit(i, 0);
   1289   }
   1290   return result;
   1291 }
   1292 
   1293 // If {result_storage} is non-nullptr, it will be used for the result,
   1294 // otherwise a new BigInt of appropriate length will be allocated.
   1295 // {result_storage} may alias {x} or {y} for in-place modification.
   1296 Handle<MutableBigInt> MutableBigInt::AbsoluteAnd(
   1297     Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
   1298     MutableBigInt* result_storage) {
   1299   return AbsoluteBitwiseOp(isolate, x, y, result_storage, kSkip, kSymmetric,
   1300                            [](digit_t a, digit_t b) { return a & b; });
   1301 }
   1302 
   1303 // If {result_storage} is non-nullptr, it will be used for the result,
   1304 // otherwise a new BigInt of appropriate length will be allocated.
   1305 // {result_storage} may alias {x} or {y} for in-place modification.
   1306 Handle<MutableBigInt> MutableBigInt::AbsoluteAndNot(
   1307     Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
   1308     MutableBigInt* result_storage) {
   1309   return AbsoluteBitwiseOp(isolate, x, y, result_storage, kCopy, kNotSymmetric,
   1310                            [](digit_t a, digit_t b) { return a & ~b; });
   1311 }
   1312 
   1313 // If {result_storage} is non-nullptr, it will be used for the result,
   1314 // otherwise a new BigInt of appropriate length will be allocated.
   1315 // {result_storage} may alias {x} or {y} for in-place modification.
   1316 Handle<MutableBigInt> MutableBigInt::AbsoluteOr(Isolate* isolate,
   1317                                                 Handle<BigIntBase> x,
   1318                                                 Handle<BigIntBase> y,
   1319                                                 MutableBigInt* result_storage) {
   1320   return AbsoluteBitwiseOp(isolate, x, y, result_storage, kCopy, kSymmetric,
   1321                            [](digit_t a, digit_t b) { return a | b; });
   1322 }
   1323 
   1324 // If {result_storage} is non-nullptr, it will be used for the result,
   1325 // otherwise a new BigInt of appropriate length will be allocated.
   1326 // {result_storage} may alias {x} or {y} for in-place modification.
   1327 Handle<MutableBigInt> MutableBigInt::AbsoluteXor(
   1328     Isolate* isolate, Handle<BigIntBase> x, Handle<BigIntBase> y,
   1329     MutableBigInt* result_storage) {
   1330   return AbsoluteBitwiseOp(isolate, x, y, result_storage, kCopy, kSymmetric,
   1331                            [](digit_t a, digit_t b) { return a ^ b; });
   1332 }
   1333 
   1334 // Returns a positive value if abs(x) > abs(y), a negative value if
   1335 // abs(x) < abs(y), or zero if abs(x) == abs(y).
   1336 int MutableBigInt::AbsoluteCompare(Handle<BigIntBase> x, Handle<BigIntBase> y) {
   1337   int diff = x->length() - y->length();
   1338   if (diff != 0) return diff;
   1339   int i = x->length() - 1;
   1340   while (i >= 0 && x->digit(i) == y->digit(i)) i--;
   1341   if (i < 0) return 0;
   1342   return x->digit(i) > y->digit(i) ? 1 : -1;
   1343 }
   1344 
   1345 // Multiplies {multiplicand} with {multiplier} and adds the result to
   1346 // {accumulator}, starting at {accumulator_index} for the least-significant
   1347 // digit.
   1348 // Callers must ensure that {accumulator} is big enough to hold the result.
   1349 void MutableBigInt::MultiplyAccumulate(Handle<BigIntBase> multiplicand,
   1350                                        digit_t multiplier,
   1351                                        Handle<MutableBigInt> accumulator,
   1352                                        int accumulator_index) {
   1353   // This is a minimum requirement; the DCHECK in the second loop below
   1354   // will enforce more as needed.
   1355   DCHECK(accumulator->length() > multiplicand->length() + accumulator_index);
   1356   if (multiplier == 0L) return;
   1357   digit_t carry = 0;
   1358   digit_t high = 0;
   1359   for (int i = 0; i < multiplicand->length(); i++, accumulator_index++) {
   1360     digit_t acc = accumulator->digit(accumulator_index);
   1361     digit_t new_carry = 0;
   1362     // Add last round's carryovers.
   1363     acc = digit_add(acc, high, &new_carry);
   1364     acc = digit_add(acc, carry, &new_carry);
   1365     // Compute this round's multiplication.
   1366     digit_t m_digit = multiplicand->digit(i);
   1367     digit_t low = digit_mul(multiplier, m_digit, &high);
   1368     acc = digit_add(acc, low, &new_carry);
   1369     // Store result and prepare for next round.
   1370     accumulator->set_digit(accumulator_index, acc);
   1371     carry = new_carry;
   1372   }
   1373   for (; carry != 0 || high != 0; accumulator_index++) {
   1374     DCHECK(accumulator_index < accumulator->length());
   1375     digit_t acc = accumulator->digit(accumulator_index);
   1376     digit_t new_carry = 0;
   1377     acc = digit_add(acc, high, &new_carry);
   1378     high = 0;
   1379     acc = digit_add(acc, carry, &new_carry);
   1380     accumulator->set_digit(accumulator_index, acc);
   1381     carry = new_carry;
   1382   }
   1383 }
   1384 
   1385 // Multiplies {source} with {factor} and adds {summand} to the result.
   1386 // {result} and {source} may be the same BigInt for inplace modification.
   1387 void MutableBigInt::InternalMultiplyAdd(BigIntBase* source, digit_t factor,
   1388                                         digit_t summand, int n,
   1389                                         MutableBigInt* result) {
   1390   DCHECK(source->length() >= n);
   1391   DCHECK(result->length() >= n);
   1392   digit_t carry = summand;
   1393   digit_t high = 0;
   1394   for (int i = 0; i < n; i++) {
   1395     digit_t current = source->digit(i);
   1396     digit_t new_carry = 0;
   1397     // Compute this round's multiplication.
   1398     digit_t new_high = 0;
   1399     current = digit_mul(current, factor, &new_high);
   1400     // Add last round's carryovers.
   1401     current = digit_add(current, high, &new_carry);
   1402     current = digit_add(current, carry, &new_carry);
   1403     // Store result and prepare for next round.
   1404     result->set_digit(i, current);
   1405     carry = new_carry;
   1406     high = new_high;
   1407   }
   1408   if (result->length() > n) {
   1409     result->set_digit(n++, carry + high);
   1410     // Current callers don't pass in such large results, but let's be robust.
   1411     while (n < result->length()) {
   1412       result->set_digit(n++, 0);
   1413     }
   1414   } else {
   1415     CHECK_EQ(carry + high, 0);
   1416   }
   1417 }
   1418 
   1419 // Multiplies {x} with {factor} and then adds {summand} to it.
   1420 void BigInt::InplaceMultiplyAdd(Handle<FreshlyAllocatedBigInt> x,
   1421                                 uintptr_t factor, uintptr_t summand) {
   1422   STATIC_ASSERT(sizeof(factor) == sizeof(digit_t));
   1423   STATIC_ASSERT(sizeof(summand) == sizeof(digit_t));
   1424   Handle<MutableBigInt> bigint = MutableBigInt::Cast(x);
   1425   MutableBigInt::InternalMultiplyAdd(*bigint, factor, summand, bigint->length(),
   1426                                      *bigint);
   1427 }
   1428 
   1429 // Divides {x} by {divisor}, returning the result in {quotient} and {remainder}.
   1430 // Mathematically, the contract is:
   1431 // quotient = (x - remainder) / divisor, with 0 <= remainder < divisor.
   1432 // If {quotient} is an empty handle, an appropriately sized BigInt will be
   1433 // allocated for it; otherwise the caller must ensure that it is big enough.
   1434 // {quotient} can be the same as {x} for an in-place division. {quotient} can
   1435 // also be nullptr if the caller is only interested in the remainder.
   1436 void MutableBigInt::AbsoluteDivSmall(Isolate* isolate, Handle<BigIntBase> x,
   1437                                      digit_t divisor,
   1438                                      Handle<MutableBigInt>* quotient,
   1439                                      digit_t* remainder) {
   1440   DCHECK_NE(divisor, 0);
   1441   DCHECK(!x->is_zero());  // Callers check anyway, no need to handle this.
   1442   *remainder = 0;
   1443   int length = x->length();
   1444   if (quotient != nullptr) {
   1445     if ((*quotient).is_null()) {
   1446       *quotient = New(isolate, length).ToHandleChecked();
   1447     }
   1448     for (int i = length - 1; i >= 0; i--) {
   1449       digit_t q = digit_div(*remainder, x->digit(i), divisor, remainder);
   1450       (*quotient)->set_digit(i, q);
   1451     }
   1452   } else {
   1453     for (int i = length - 1; i >= 0; i--) {
   1454       digit_div(*remainder, x->digit(i), divisor, remainder);
   1455     }
   1456   }
   1457 }
   1458 
   1459 // Divides {dividend} by {divisor}, returning the result in {quotient} and
   1460 // {remainder}. Mathematically, the contract is:
   1461 // quotient = (dividend - remainder) / divisor, with 0 <= remainder < divisor.
   1462 // Both {quotient} and {remainder} are optional, for callers that are only
   1463 // interested in one of them.
   1464 // See Knuth, Volume 2, section 4.3.1, Algorithm D.
   1465 bool MutableBigInt::AbsoluteDivLarge(Isolate* isolate,
   1466                                      Handle<BigIntBase> dividend,
   1467                                      Handle<BigIntBase> divisor,
   1468                                      Handle<MutableBigInt>* quotient,
   1469                                      Handle<MutableBigInt>* remainder) {
   1470   DCHECK_GE(divisor->length(), 2);
   1471   DCHECK(dividend->length() >= divisor->length());
   1472   // The unusual variable names inside this function are consistent with
   1473   // Knuth's book, as well as with Go's implementation of this algorithm.
   1474   // Maintaining this consistency is probably more useful than trying to
   1475   // come up with more descriptive names for them.
   1476   int n = divisor->length();
   1477   int m = dividend->length() - n;
   1478 
   1479   // The quotient to be computed.
   1480   Handle<MutableBigInt> q;
   1481   if (quotient != nullptr) q = New(isolate, m + 1).ToHandleChecked();
   1482   // In each iteration, {qhatv} holds {divisor} * {current quotient digit}.
   1483   // "v" is the book's name for {divisor}, "qhat" the current quotient digit.
   1484   Handle<MutableBigInt> qhatv;
   1485   if (!New(isolate, n + 1).ToHandle(&qhatv)) return false;
   1486 
   1487   // D1.
   1488   // Left-shift inputs so that the divisor's MSB is set. This is necessary
   1489   // to prevent the digit-wise divisions (see digit_div call below) from
   1490   // overflowing (they take a two digits wide input, and return a one digit
   1491   // result).
   1492   int shift = base::bits::CountLeadingZeros(divisor->digit(n - 1));
   1493   if (shift > 0) {
   1494     divisor = SpecialLeftShift(isolate, divisor, shift, kSameSizeResult)
   1495                   .ToHandleChecked();
   1496   }
   1497   // Holds the (continuously updated) remaining part of the dividend, which
   1498   // eventually becomes the remainder.
   1499   Handle<MutableBigInt> u;
   1500   if (!SpecialLeftShift(isolate, dividend, shift, kAlwaysAddOneDigit)
   1501            .ToHandle(&u)) {
   1502     return false;
   1503   }
   1504 
   1505   // D2.
   1506   // Iterate over the dividend's digit (like the "grad school" algorithm).
   1507   // {vn1} is the divisor's most significant digit.
   1508   digit_t vn1 = divisor->digit(n - 1);
   1509   for (int j = m; j >= 0; j--) {
   1510     // D3.
   1511     // Estimate the current iteration's quotient digit (see Knuth for details).
   1512     // {qhat} is the current quotient digit.
   1513     digit_t qhat = std::numeric_limits<digit_t>::max();
   1514     // {ujn} is the dividend's most significant remaining digit.
   1515     digit_t ujn = u->digit(j + n);
   1516     if (ujn != vn1) {
   1517       // {rhat} is the current iteration's remainder.
   1518       digit_t rhat = 0;
   1519       // Estimate the current quotient digit by dividing the most significant
   1520       // digits of dividend and divisor. The result will not be too small,
   1521       // but could be a bit too large.
   1522       qhat = digit_div(ujn, u->digit(j + n - 1), vn1, &rhat);
   1523 
   1524       // Decrement the quotient estimate as needed by looking at the next
   1525       // digit, i.e. by testing whether
   1526       // qhat * v_{n-2} > (rhat << kDigitBits) + u_{j+n-2}.
   1527       digit_t vn2 = divisor->digit(n - 2);
   1528       digit_t ujn2 = u->digit(j + n - 2);
   1529       while (ProductGreaterThan(qhat, vn2, rhat, ujn2)) {
   1530         qhat--;
   1531         digit_t prev_rhat = rhat;
   1532         rhat += vn1;
   1533         // v[n-1] >= 0, so this tests for overflow.
   1534         if (rhat < prev_rhat) break;
   1535       }
   1536     }
   1537 
   1538     // D4.
   1539     // Multiply the divisor with the current quotient digit, and subtract
   1540     // it from the dividend. If there was "borrow", then the quotient digit
   1541     // was one too high, so we must correct it and undo one subtraction of
   1542     // the (shifted) divisor.
   1543     InternalMultiplyAdd(*divisor, qhat, 0, n, *qhatv);
   1544     digit_t c = u->InplaceSub(qhatv, j);
   1545     if (c != 0) {
   1546       c = u->InplaceAdd(divisor, j);
   1547       u->set_digit(j + n, u->digit(j + n) + c);
   1548       qhat--;
   1549     }
   1550 
   1551     if (quotient != nullptr) q->set_digit(j, qhat);
   1552   }
   1553   if (quotient != nullptr) {
   1554     *quotient = q;  // Caller will right-trim.
   1555   }
   1556   if (remainder != nullptr) {
   1557     u->InplaceRightShift(shift);
   1558     *remainder = u;
   1559   }
   1560   return true;
   1561 }
   1562 
   1563 // Returns whether (factor1 * factor2) > (high << kDigitBits) + low.
   1564 bool MutableBigInt::ProductGreaterThan(digit_t factor1, digit_t factor2,
   1565                                        digit_t high, digit_t low) {
   1566   digit_t result_high;
   1567   digit_t result_low = digit_mul(factor1, factor2, &result_high);
   1568   return result_high > high || (result_high == high && result_low > low);
   1569 }
   1570 
   1571 // Adds {summand} onto {this}, starting with {summand}'s 0th digit
   1572 // at {this}'s {start_index}'th digit. Returns the "carry" (0 or 1).
   1573 BigInt::digit_t MutableBigInt::InplaceAdd(Handle<BigIntBase> summand,
   1574                                           int start_index) {
   1575   digit_t carry = 0;
   1576   int n = summand->length();
   1577   DCHECK(length() >= start_index + n);
   1578   for (int i = 0; i < n; i++) {
   1579     digit_t new_carry = 0;
   1580     digit_t sum =
   1581         digit_add(digit(start_index + i), summand->digit(i), &new_carry);
   1582     sum = digit_add(sum, carry, &new_carry);
   1583     set_digit(start_index + i, sum);
   1584     carry = new_carry;
   1585   }
   1586   return carry;
   1587 }
   1588 
   1589 // Subtracts {subtrahend} from {this}, starting with {subtrahend}'s 0th digit
   1590 // at {this}'s {start_index}-th digit. Returns the "borrow" (0 or 1).
   1591 BigInt::digit_t MutableBigInt::InplaceSub(Handle<BigIntBase> subtrahend,
   1592                                           int start_index) {
   1593   digit_t borrow = 0;
   1594   int n = subtrahend->length();
   1595   DCHECK(length() >= start_index + n);
   1596   for (int i = 0; i < n; i++) {
   1597     digit_t new_borrow = 0;
   1598     digit_t difference =
   1599         digit_sub(digit(start_index + i), subtrahend->digit(i), &new_borrow);
   1600     difference = digit_sub(difference, borrow, &new_borrow);
   1601     set_digit(start_index + i, difference);
   1602     borrow = new_borrow;
   1603   }
   1604   return borrow;
   1605 }
   1606 
   1607 void MutableBigInt::InplaceRightShift(int shift) {
   1608   DCHECK_GE(shift, 0);
   1609   DCHECK_LT(shift, kDigitBits);
   1610   DCHECK_GT(length(), 0);
   1611   DCHECK_EQ(digit(0) & ((static_cast<digit_t>(1) << shift) - 1), 0);
   1612   if (shift == 0) return;
   1613   digit_t carry = digit(0) >> shift;
   1614   int last = length() - 1;
   1615   for (int i = 0; i < last; i++) {
   1616     digit_t d = digit(i + 1);
   1617     set_digit(i, (d << (kDigitBits - shift)) | carry);
   1618     carry = d >> shift;
   1619   }
   1620   set_digit(last, carry);
   1621 }
   1622 
   1623 // Always copies the input, even when {shift} == 0.
   1624 // {shift} must be less than kDigitBits, {x} must be non-zero.
   1625 MaybeHandle<MutableBigInt> MutableBigInt::SpecialLeftShift(
   1626     Isolate* isolate, Handle<BigIntBase> x, int shift,
   1627     SpecialLeftShiftMode mode) {
   1628   DCHECK_GE(shift, 0);
   1629   DCHECK_LT(shift, kDigitBits);
   1630   DCHECK_GT(x->length(), 0);
   1631   int n = x->length();
   1632   int result_length = mode == kAlwaysAddOneDigit ? n + 1 : n;
   1633   Handle<MutableBigInt> result;
   1634   if (!New(isolate, result_length).ToHandle(&result)) {
   1635     return MaybeHandle<MutableBigInt>();
   1636   }
   1637   if (shift == 0) {
   1638     for (int i = 0; i < n; i++) result->set_digit(i, x->digit(i));
   1639     if (mode == kAlwaysAddOneDigit) result->set_digit(n, 0);
   1640     return result;
   1641   }
   1642   DCHECK_GT(shift, 0);
   1643   digit_t carry = 0;
   1644   for (int i = 0; i < n; i++) {
   1645     digit_t d = x->digit(i);
   1646     result->set_digit(i, (d << shift) | carry);
   1647     carry = d >> (kDigitBits - shift);
   1648   }
   1649   if (mode == kAlwaysAddOneDigit) {
   1650     result->set_digit(n, carry);
   1651   } else {
   1652     DCHECK_EQ(mode, kSameSizeResult);
   1653     DCHECK_EQ(carry, 0);
   1654   }
   1655   return result;
   1656 }
   1657 
   1658 MaybeHandle<BigInt> MutableBigInt::LeftShiftByAbsolute(Isolate* isolate,
   1659                                                        Handle<BigIntBase> x,
   1660                                                        Handle<BigIntBase> y) {
   1661   Maybe<digit_t> maybe_shift = ToShiftAmount(y);
   1662   if (maybe_shift.IsNothing()) {
   1663     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
   1664                     BigInt);
   1665   }
   1666   digit_t shift = maybe_shift.FromJust();
   1667   int digit_shift = static_cast<int>(shift / kDigitBits);
   1668   int bits_shift = static_cast<int>(shift % kDigitBits);
   1669   int length = x->length();
   1670   bool grow = bits_shift != 0 &&
   1671               (x->digit(length - 1) >> (kDigitBits - bits_shift)) != 0;
   1672   int result_length = length + digit_shift + grow;
   1673   if (result_length > kMaxLength) {
   1674     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
   1675                     BigInt);
   1676   }
   1677   Handle<MutableBigInt> result;
   1678   if (!New(isolate, result_length).ToHandle(&result)) {
   1679     return MaybeHandle<BigInt>();
   1680   }
   1681   if (bits_shift == 0) {
   1682     int i = 0;
   1683     for (; i < digit_shift; i++) result->set_digit(i, 0ul);
   1684     for (; i < result_length; i++) {
   1685       result->set_digit(i, x->digit(i - digit_shift));
   1686     }
   1687   } else {
   1688     digit_t carry = 0;
   1689     for (int i = 0; i < digit_shift; i++) result->set_digit(i, 0ul);
   1690     for (int i = 0; i < length; i++) {
   1691       digit_t d = x->digit(i);
   1692       result->set_digit(i + digit_shift, (d << bits_shift) | carry);
   1693       carry = d >> (kDigitBits - bits_shift);
   1694     }
   1695     if (grow) {
   1696       result->set_digit(length + digit_shift, carry);
   1697     } else {
   1698       DCHECK_EQ(carry, 0);
   1699     }
   1700   }
   1701   result->set_sign(x->sign());
   1702   return MakeImmutable(result);
   1703 }
   1704 
   1705 Handle<BigInt> MutableBigInt::RightShiftByAbsolute(Isolate* isolate,
   1706                                                    Handle<BigIntBase> x,
   1707                                                    Handle<BigIntBase> y) {
   1708   int length = x->length();
   1709   bool sign = x->sign();
   1710   Maybe<digit_t> maybe_shift = ToShiftAmount(y);
   1711   if (maybe_shift.IsNothing()) {
   1712     return RightShiftByMaximum(isolate, sign);
   1713   }
   1714   digit_t shift = maybe_shift.FromJust();
   1715   int digit_shift = static_cast<int>(shift / kDigitBits);
   1716   int bits_shift = static_cast<int>(shift % kDigitBits);
   1717   int result_length = length - digit_shift;
   1718   if (result_length <= 0) {
   1719     return RightShiftByMaximum(isolate, sign);
   1720   }
   1721   // For negative numbers, round down if any bit was shifted out (so that e.g.
   1722   // -5n >> 1n == -3n and not -2n). Check now whether this will happen and
   1723   // whether it can cause overflow into a new digit. If we allocate the result
   1724   // large enough up front, it avoids having to do a second allocation later.
   1725   bool must_round_down = false;
   1726   if (sign) {
   1727     const digit_t mask = (static_cast<digit_t>(1) << bits_shift) - 1;
   1728     if ((x->digit(digit_shift) & mask) != 0) {
   1729       must_round_down = true;
   1730     } else {
   1731       for (int i = 0; i < digit_shift; i++) {
   1732         if (x->digit(i) != 0) {
   1733           must_round_down = true;
   1734           break;
   1735         }
   1736       }
   1737     }
   1738   }
   1739   // If bits_shift is non-zero, it frees up bits, preventing overflow.
   1740   if (must_round_down && bits_shift == 0) {
   1741     // Overflow cannot happen if the most significant digit has unset bits.
   1742     digit_t msd = x->digit(length - 1);
   1743     bool rounding_can_overflow = digit_ismax(msd);
   1744     if (rounding_can_overflow) result_length++;
   1745   }
   1746 
   1747   DCHECK_LE(result_length, length);
   1748   Handle<MutableBigInt> result = New(isolate, result_length).ToHandleChecked();
   1749   if (bits_shift == 0) {
   1750     for (int i = digit_shift; i < length; i++) {
   1751       result->set_digit(i - digit_shift, x->digit(i));
   1752     }
   1753   } else {
   1754     digit_t carry = x->digit(digit_shift) >> bits_shift;
   1755     int last = length - digit_shift - 1;
   1756     for (int i = 0; i < last; i++) {
   1757       digit_t d = x->digit(i + digit_shift + 1);
   1758       result->set_digit(i, (d << (kDigitBits - bits_shift)) | carry);
   1759       carry = d >> bits_shift;
   1760     }
   1761     result->set_digit(last, carry);
   1762   }
   1763 
   1764   if (sign) {
   1765     result->set_sign(true);
   1766     if (must_round_down) {
   1767       // Since the result is negative, rounding down means adding one to
   1768       // its absolute value. This cannot overflow.
   1769       result = AbsoluteAddOne(isolate, result, true, *result).ToHandleChecked();
   1770     }
   1771   }
   1772   return MakeImmutable(result);
   1773 }
   1774 
   1775 Handle<BigInt> MutableBigInt::RightShiftByMaximum(Isolate* isolate, bool sign) {
   1776   if (sign) {
   1777     // TODO(jkummerow): Consider caching a canonical -1n BigInt.
   1778     return NewFromInt(isolate, -1);
   1779   } else {
   1780     return Zero(isolate);
   1781   }
   1782 }
   1783 
   1784 // Returns the value of {x} if it is less than the maximum bit length of
   1785 // a BigInt, or Nothing otherwise.
   1786 Maybe<BigInt::digit_t> MutableBigInt::ToShiftAmount(Handle<BigIntBase> x) {
   1787   if (x->length() > 1) return Nothing<digit_t>();
   1788   digit_t value = x->digit(0);
   1789   STATIC_ASSERT(kMaxLengthBits < std::numeric_limits<digit_t>::max());
   1790   if (value > kMaxLengthBits) return Nothing<digit_t>();
   1791   return Just(value);
   1792 }
   1793 
   1794 // Lookup table for the maximum number of bits required per character of a
   1795 // base-N string representation of a number. To increase accuracy, the array
   1796 // value is the actual value multiplied by 32. To generate this table:
   1797 // for (var i = 0; i <= 36; i++) { print(Math.ceil(Math.log2(i) * 32) + ","); }
   1798 constexpr uint8_t kMaxBitsPerChar[] = {
   1799     0,   0,   32,  51,  64,  75,  83,  90,  96,  // 0..8
   1800     102, 107, 111, 115, 119, 122, 126, 128,      // 9..16
   1801     131, 134, 136, 139, 141, 143, 145, 147,      // 17..24
   1802     149, 151, 153, 154, 156, 158, 159, 160,      // 25..32
   1803     162, 163, 165, 166,                          // 33..36
   1804 };
   1805 
   1806 static const int kBitsPerCharTableShift = 5;
   1807 static const size_t kBitsPerCharTableMultiplier = 1u << kBitsPerCharTableShift;
   1808 
   1809 MaybeHandle<FreshlyAllocatedBigInt> BigInt::AllocateFor(
   1810     Isolate* isolate, int radix, int charcount, ShouldThrow should_throw,
   1811     PretenureFlag pretenure) {
   1812   DCHECK(2 <= radix && radix <= 36);
   1813   DCHECK_GE(charcount, 0);
   1814   size_t bits_per_char = kMaxBitsPerChar[radix];
   1815   size_t chars = static_cast<size_t>(charcount);
   1816   const int roundup = kBitsPerCharTableMultiplier - 1;
   1817   if (chars <= (std::numeric_limits<size_t>::max() - roundup) / bits_per_char) {
   1818     size_t bits_min = bits_per_char * chars;
   1819     // Divide by 32 (see table), rounding up.
   1820     bits_min = (bits_min + roundup) >> kBitsPerCharTableShift;
   1821     if (bits_min <= static_cast<size_t>(kMaxInt)) {
   1822       // Divide by kDigitsBits, rounding up.
   1823       int length = (static_cast<int>(bits_min) + kDigitBits - 1) / kDigitBits;
   1824       if (length <= kMaxLength) {
   1825         Handle<MutableBigInt> result =
   1826             MutableBigInt::New(isolate, length, pretenure).ToHandleChecked();
   1827         result->InitializeDigits(length);
   1828         return result;
   1829       }
   1830     }
   1831   }
   1832   // All the overflow/maximum checks above fall through to here.
   1833   if (should_throw == kThrowOnError) {
   1834     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
   1835                     FreshlyAllocatedBigInt);
   1836   } else {
   1837     return MaybeHandle<FreshlyAllocatedBigInt>();
   1838   }
   1839 }
   1840 
   1841 Handle<BigInt> BigInt::Finalize(Handle<FreshlyAllocatedBigInt> x, bool sign) {
   1842   Handle<MutableBigInt> bigint = MutableBigInt::Cast(x);
   1843   bigint->set_sign(sign);
   1844   return MutableBigInt::MakeImmutable(bigint);
   1845 }
   1846 
   1847 // The serialization format MUST NOT CHANGE without updating the format
   1848 // version in value-serializer.cc!
   1849 uint32_t BigInt::GetBitfieldForSerialization() const {
   1850   // In order to make the serialization format the same on 32/64 bit builds,
   1851   // we convert the length-in-digits to length-in-bytes for serialization.
   1852   // Being able to do this depends on having enough LengthBits:
   1853   STATIC_ASSERT(kMaxLength * kDigitSize <= LengthBits::kMax);
   1854   int bytelength = length() * kDigitSize;
   1855   return SignBits::encode(sign()) | LengthBits::encode(bytelength);
   1856 }
   1857 
   1858 int BigInt::DigitsByteLengthForBitfield(uint32_t bitfield) {
   1859   return LengthBits::decode(bitfield);
   1860 }
   1861 
   1862 // The serialization format MUST NOT CHANGE without updating the format
   1863 // version in value-serializer.cc!
   1864 void BigInt::SerializeDigits(uint8_t* storage) {
   1865   void* digits = reinterpret_cast<void*>(reinterpret_cast<Address>(this) +
   1866                                          kDigitsOffset - kHeapObjectTag);
   1867 #if defined(V8_TARGET_LITTLE_ENDIAN)
   1868   int bytelength = length() * kDigitSize;
   1869   memcpy(storage, digits, bytelength);
   1870 #elif defined(V8_TARGET_BIG_ENDIAN)
   1871   digit_t* digit_storage = reinterpret_cast<digit_t*>(storage);
   1872   const digit_t* digit = reinterpret_cast<const digit_t*>(digits);
   1873   for (int i = 0; i < length(); i++) {
   1874     *digit_storage = ByteReverse(*digit);
   1875     digit_storage++;
   1876     digit++;
   1877   }
   1878 #endif  // V8_TARGET_BIG_ENDIAN
   1879 }
   1880 
   1881 // The serialization format MUST NOT CHANGE without updating the format
   1882 // version in value-serializer.cc!
   1883 MaybeHandle<BigInt> BigInt::FromSerializedDigits(
   1884     Isolate* isolate, uint32_t bitfield, Vector<const uint8_t> digits_storage,
   1885     PretenureFlag pretenure) {
   1886   int bytelength = LengthBits::decode(bitfield);
   1887   DCHECK(digits_storage.length() == bytelength);
   1888   bool sign = SignBits::decode(bitfield);
   1889   int length = (bytelength + kDigitSize - 1) / kDigitSize;  // Round up.
   1890   Handle<MutableBigInt> result =
   1891       MutableBigInt::Cast(isolate->factory()->NewBigInt(length, pretenure));
   1892   result->initialize_bitfield(sign, length);
   1893   void* digits = reinterpret_cast<void*>(reinterpret_cast<Address>(*result) +
   1894                                          kDigitsOffset - kHeapObjectTag);
   1895 #if defined(V8_TARGET_LITTLE_ENDIAN)
   1896   memcpy(digits, digits_storage.start(), bytelength);
   1897   void* padding_start =
   1898       reinterpret_cast<void*>(reinterpret_cast<Address>(digits) + bytelength);
   1899   memset(padding_start, 0, length * kDigitSize - bytelength);
   1900 #elif defined(V8_TARGET_BIG_ENDIAN)
   1901   digit_t* digit = reinterpret_cast<digit_t*>(digits);
   1902   const digit_t* digit_storage =
   1903       reinterpret_cast<const digit_t*>(digits_storage.start());
   1904   for (int i = 0; i < bytelength / kDigitSize; i++) {
   1905     *digit = ByteReverse(*digit_storage);
   1906     digit_storage++;
   1907     digit++;
   1908   }
   1909   if (bytelength % kDigitSize) {
   1910     *digit = 0;
   1911     byte* digit_byte = reinterpret_cast<byte*>(digit);
   1912     digit_byte += sizeof(*digit) - 1;
   1913     const byte* digit_storage_byte =
   1914         reinterpret_cast<const byte*>(digit_storage);
   1915     for (int i = 0; i < bytelength % kDigitSize; i++) {
   1916       *digit_byte = *digit_storage_byte;
   1917       digit_byte--;
   1918       digit_storage_byte++;
   1919     }
   1920   }
   1921 #endif  // V8_TARGET_BIG_ENDIAN
   1922   return MutableBigInt::MakeImmutable(result);
   1923 }
   1924 
   1925 static const char kConversionChars[] = "0123456789abcdefghijklmnopqrstuvwxyz";
   1926 
   1927 MaybeHandle<String> MutableBigInt::ToStringBasePowerOfTwo(Isolate* isolate,
   1928                                                           Handle<BigIntBase> x,
   1929                                                           int radix) {
   1930   STATIC_ASSERT(base::bits::IsPowerOfTwo(kDigitBits));
   1931   DCHECK(base::bits::IsPowerOfTwo(radix));
   1932   DCHECK(radix >= 2 && radix <= 32);
   1933   DCHECK(!x->is_zero());
   1934 
   1935   const int length = x->length();
   1936   const bool sign = x->sign();
   1937   const int bits_per_char = base::bits::CountTrailingZeros(radix);
   1938   const int char_mask = radix - 1;
   1939   // Compute the length of the resulting string: divide the bit length of the
   1940   // BigInt by the number of bits representable per character (rounding up).
   1941   const digit_t msd = x->digit(length - 1);
   1942   const int msd_leading_zeros = base::bits::CountLeadingZeros(msd);
   1943   const size_t bit_length = length * kDigitBits - msd_leading_zeros;
   1944   const size_t chars_required =
   1945       (bit_length + bits_per_char - 1) / bits_per_char + sign;
   1946 
   1947   if (chars_required > String::kMaxLength) {
   1948     THROW_NEW_ERROR(isolate, NewInvalidStringLengthError(), String);
   1949   }
   1950 
   1951   Handle<SeqOneByteString> result =
   1952       isolate->factory()
   1953           ->NewRawOneByteString(static_cast<int>(chars_required))
   1954           .ToHandleChecked();
   1955   DisallowHeapAllocation no_gc;
   1956   uint8_t* buffer = result->GetChars();
   1957   // Print the number into the string, starting from the last position.
   1958   int pos = static_cast<int>(chars_required - 1);
   1959   digit_t digit = 0;
   1960   // Keeps track of how many unprocessed bits there are in {digit}.
   1961   int available_bits = 0;
   1962   for (int i = 0; i < length - 1; i++) {
   1963     digit_t new_digit = x->digit(i);
   1964     // Take any leftover bits from the last iteration into account.
   1965     int current = (digit | (new_digit << available_bits)) & char_mask;
   1966     buffer[pos--] = kConversionChars[current];
   1967     int consumed_bits = bits_per_char - available_bits;
   1968     digit = new_digit >> consumed_bits;
   1969     available_bits = kDigitBits - consumed_bits;
   1970     while (available_bits >= bits_per_char) {
   1971       buffer[pos--] = kConversionChars[digit & char_mask];
   1972       digit >>= bits_per_char;
   1973       available_bits -= bits_per_char;
   1974     }
   1975   }
   1976   // Take any leftover bits from the last iteration into account.
   1977   int current = (digit | (msd << available_bits)) & char_mask;
   1978   buffer[pos--] = kConversionChars[current];
   1979   digit = msd >> (bits_per_char - available_bits);
   1980   while (digit != 0) {
   1981     buffer[pos--] = kConversionChars[digit & char_mask];
   1982     digit >>= bits_per_char;
   1983   }
   1984   if (sign) buffer[pos--] = '-';
   1985   DCHECK_EQ(pos, -1);
   1986   return result;
   1987 }
   1988 
   1989 MaybeHandle<String> MutableBigInt::ToStringGeneric(Isolate* isolate,
   1990                                                    Handle<BigIntBase> x,
   1991                                                    int radix) {
   1992   DCHECK(radix >= 2 && radix <= 36);
   1993   DCHECK(!x->is_zero());
   1994   Heap* heap = isolate->heap();
   1995 
   1996   const int length = x->length();
   1997   const bool sign = x->sign();
   1998 
   1999   // Compute (an overapproximation of) the length of the resulting string:
   2000   // Divide bit length of the BigInt by bits representable per character.
   2001   const size_t bit_length =
   2002       length * kDigitBits - base::bits::CountLeadingZeros(x->digit(length - 1));
   2003   // Maximum number of bits we can represent with one character. We'll use this
   2004   // to find an appropriate chunk size below.
   2005   const uint8_t max_bits_per_char = kMaxBitsPerChar[radix];
   2006   // For estimating result length, we have to be pessimistic and work with
   2007   // the minimum number of bits one character can represent.
   2008   const uint8_t min_bits_per_char = max_bits_per_char - 1;
   2009   // Perform the following computation with uint64_t to avoid overflows.
   2010   uint64_t chars_required = bit_length;
   2011   chars_required *= kBitsPerCharTableMultiplier;
   2012   chars_required += min_bits_per_char - 1;  // Round up.
   2013   chars_required /= min_bits_per_char;
   2014   chars_required += sign;
   2015 
   2016   if (chars_required > String::kMaxLength) {
   2017     THROW_NEW_ERROR(isolate, NewInvalidStringLengthError(), String);
   2018   }
   2019   Handle<SeqOneByteString> result =
   2020       isolate->factory()
   2021           ->NewRawOneByteString(static_cast<int>(chars_required))
   2022           .ToHandleChecked();
   2023 
   2024 #if DEBUG
   2025   // Zap the string first.
   2026   {
   2027     DisallowHeapAllocation no_gc;
   2028     uint8_t* chars = result->GetChars();
   2029     for (int i = 0; i < static_cast<int>(chars_required); i++) chars[i] = '?';
   2030   }
   2031 #endif
   2032 
   2033   // We assemble the result string in reverse order, and then reverse it.
   2034   // TODO(jkummerow): Consider building the string from the right, and
   2035   // left-shifting it if the length estimate was too large.
   2036   int pos = 0;
   2037 
   2038   digit_t last_digit;
   2039   if (length == 1) {
   2040     last_digit = x->digit(0);
   2041   } else {
   2042     int chunk_chars =
   2043         kDigitBits * kBitsPerCharTableMultiplier / max_bits_per_char;
   2044     digit_t chunk_divisor = digit_pow(radix, chunk_chars);
   2045     // By construction of chunk_chars, there can't have been overflow.
   2046     DCHECK_NE(chunk_divisor, 0);
   2047     int nonzero_digit = length - 1;
   2048     DCHECK_NE(x->digit(nonzero_digit), 0);
   2049     // {rest} holds the part of the BigInt that we haven't looked at yet.
   2050     // Not to be confused with "remainder"!
   2051     Handle<MutableBigInt> rest;
   2052     // In the first round, divide the input, allocating a new BigInt for
   2053     // the result == rest; from then on divide the rest in-place.
   2054     Handle<BigIntBase>* dividend = &x;
   2055     do {
   2056       digit_t chunk;
   2057       AbsoluteDivSmall(isolate, *dividend, chunk_divisor, &rest, &chunk);
   2058       DCHECK(!rest.is_null());
   2059       dividend = reinterpret_cast<Handle<BigIntBase>*>(&rest);
   2060       DisallowHeapAllocation no_gc;
   2061       uint8_t* chars = result->GetChars();
   2062       for (int i = 0; i < chunk_chars; i++) {
   2063         chars[pos++] = kConversionChars[chunk % radix];
   2064         chunk /= radix;
   2065       }
   2066       DCHECK_EQ(chunk, 0);
   2067       if (rest->digit(nonzero_digit) == 0) nonzero_digit--;
   2068       // We can never clear more than one digit per iteration, because
   2069       // chunk_divisor is smaller than max digit value.
   2070       DCHECK_GT(rest->digit(nonzero_digit), 0);
   2071     } while (nonzero_digit > 0);
   2072     last_digit = rest->digit(0);
   2073   }
   2074   DisallowHeapAllocation no_gc;
   2075   uint8_t* chars = result->GetChars();
   2076   do {
   2077     chars[pos++] = kConversionChars[last_digit % radix];
   2078     last_digit /= radix;
   2079   } while (last_digit > 0);
   2080   DCHECK_GE(pos, 1);
   2081   DCHECK(pos <= static_cast<int>(chars_required));
   2082   // Remove leading zeroes.
   2083   while (pos > 1 && chars[pos - 1] == '0') pos--;
   2084   if (sign) chars[pos++] = '-';
   2085   // Trim any over-allocation (which can happen due to conservative estimates).
   2086   if (pos < static_cast<int>(chars_required)) {
   2087     result->synchronized_set_length(pos);
   2088     int string_size =
   2089         SeqOneByteString::SizeFor(static_cast<int>(chars_required));
   2090     int needed_size = SeqOneByteString::SizeFor(pos);
   2091     if (needed_size < string_size) {
   2092       Address new_end = result->address() + needed_size;
   2093       heap->CreateFillerObjectAt(new_end, (string_size - needed_size),
   2094                                  ClearRecordedSlots::kNo);
   2095     }
   2096   }
   2097   // Reverse the string.
   2098   for (int i = 0, j = pos - 1; i < j; i++, j--) {
   2099     uint8_t tmp = chars[i];
   2100     chars[i] = chars[j];
   2101     chars[j] = tmp;
   2102   }
   2103 #if DEBUG
   2104   // Verify that all characters have been written.
   2105   DCHECK(result->length() == pos);
   2106   for (int i = 0; i < pos; i++) DCHECK_NE(chars[i], '?');
   2107 #endif
   2108   return result;
   2109 }
   2110 
   2111 Handle<BigInt> BigInt::AsIntN(Isolate* isolate, uint64_t n, Handle<BigInt> x) {
   2112   if (x->is_zero()) return x;
   2113   if (n == 0) return MutableBigInt::Zero(isolate);
   2114   uint64_t needed_length = (n + kDigitBits - 1) / kDigitBits;
   2115   uint64_t x_length = static_cast<uint64_t>(x->length());
   2116   // If {x} has less than {n} bits, return it directly.
   2117   if (x_length < needed_length) return x;
   2118   DCHECK_LE(needed_length, kMaxInt);
   2119   digit_t top_digit = x->digit(static_cast<int>(needed_length) - 1);
   2120   digit_t compare_digit = static_cast<digit_t>(1) << ((n - 1) % kDigitBits);
   2121   if (x_length == needed_length && top_digit < compare_digit) return x;
   2122   // Otherwise we have to truncate (which is a no-op in the special case
   2123   // of x == -2^(n-1)), and determine the right sign. We also might have
   2124   // to subtract from 2^n to simulate having two's complement representation.
   2125   // In most cases, the result's sign is x->sign() xor "(n-1)th bit present".
   2126   // The only exception is when x is negative, has the (n-1)th bit, and all
   2127   // its bits below (n-1) are zero. In that case, the result is the minimum
   2128   // n-bit integer (example: asIntN(3, -12n) => -4n).
   2129   bool has_bit = (top_digit & compare_digit) == compare_digit;
   2130   DCHECK_LE(n, kMaxInt);
   2131   int N = static_cast<int>(n);
   2132   if (!has_bit) {
   2133     return MutableBigInt::TruncateToNBits(isolate, N, x);
   2134   }
   2135   if (!x->sign()) {
   2136     return MutableBigInt::TruncateAndSubFromPowerOfTwo(isolate, N, x, true);
   2137   }
   2138   // Negative numbers must subtract from 2^n, except for the special case
   2139   // described above.
   2140   if ((top_digit & (compare_digit - 1)) == 0) {
   2141     for (int i = static_cast<int>(needed_length) - 2; i >= 0; i--) {
   2142       if (x->digit(i) != 0) {
   2143         return MutableBigInt::TruncateAndSubFromPowerOfTwo(isolate, N, x,
   2144                                                            false);
   2145       }
   2146     }
   2147     return MutableBigInt::TruncateToNBits(isolate, N, x);
   2148   }
   2149   return MutableBigInt::TruncateAndSubFromPowerOfTwo(isolate, N, x, false);
   2150 }
   2151 
   2152 MaybeHandle<BigInt> BigInt::AsUintN(Isolate* isolate, uint64_t n,
   2153                                     Handle<BigInt> x) {
   2154   if (x->is_zero()) return x;
   2155   if (n == 0) return MutableBigInt::Zero(isolate);
   2156   // If {x} is negative, simulate two's complement representation.
   2157   if (x->sign()) {
   2158     if (n > kMaxLengthBits) {
   2159       THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
   2160                       BigInt);
   2161     }
   2162     return MutableBigInt::TruncateAndSubFromPowerOfTwo(
   2163         isolate, static_cast<int>(n), x, false);
   2164   }
   2165   // If {x} is positive and has up to {n} bits, return it directly.
   2166   if (n >= kMaxLengthBits) return x;
   2167   STATIC_ASSERT(kMaxLengthBits < kMaxInt - kDigitBits);
   2168   int needed_length = static_cast<int>((n + kDigitBits - 1) / kDigitBits);
   2169   if (x->length() < needed_length) return x;
   2170   int bits_in_top_digit = n % kDigitBits;
   2171   if (bits_in_top_digit == 0) {
   2172     if (x->length() == needed_length) return x;
   2173   } else {
   2174     digit_t top_digit = x->digit(needed_length - 1);
   2175     if ((top_digit >> bits_in_top_digit) == 0) return x;
   2176   }
   2177   // Otherwise, truncate.
   2178   DCHECK_LE(n, kMaxInt);
   2179   return MutableBigInt::TruncateToNBits(isolate, static_cast<int>(n), x);
   2180 }
   2181 
   2182 Handle<BigInt> MutableBigInt::TruncateToNBits(Isolate* isolate, int n,
   2183                                               Handle<BigInt> x) {
   2184   // Only call this when there's something to do.
   2185   DCHECK_NE(n, 0);
   2186   DCHECK_GT(x->length(), n / kDigitBits);
   2187 
   2188   int needed_digits = (n + (kDigitBits - 1)) / kDigitBits;
   2189   DCHECK_LE(needed_digits, x->length());
   2190   Handle<MutableBigInt> result = New(isolate, needed_digits).ToHandleChecked();
   2191 
   2192   // Copy all digits except the MSD.
   2193   int last = needed_digits - 1;
   2194   for (int i = 0; i < last; i++) {
   2195     result->set_digit(i, x->digit(i));
   2196   }
   2197 
   2198   // The MSD might contain extra bits that we don't want.
   2199   digit_t msd = x->digit(last);
   2200   if (n % kDigitBits != 0) {
   2201     int drop = kDigitBits - (n % kDigitBits);
   2202     msd = (msd << drop) >> drop;
   2203   }
   2204   result->set_digit(last, msd);
   2205   result->set_sign(x->sign());
   2206   return MakeImmutable(result);
   2207 }
   2208 
   2209 // Subtracts the least significant n bits of abs(x) from 2^n.
   2210 Handle<BigInt> MutableBigInt::TruncateAndSubFromPowerOfTwo(Isolate* isolate,
   2211                                                            int n,
   2212                                                            Handle<BigInt> x,
   2213                                                            bool result_sign) {
   2214   DCHECK_NE(n, 0);
   2215   DCHECK_LE(n, kMaxLengthBits);
   2216 
   2217   int needed_digits = (n + (kDigitBits - 1)) / kDigitBits;
   2218   DCHECK_LE(needed_digits, kMaxLength);  // Follows from n <= kMaxLengthBits.
   2219   Handle<MutableBigInt> result = New(isolate, needed_digits).ToHandleChecked();
   2220 
   2221   // Process all digits except the MSD.
   2222   int i = 0;
   2223   int last = needed_digits - 1;
   2224   int x_length = x->length();
   2225   digit_t borrow = 0;
   2226   // Take digits from {x} unless its length is exhausted.
   2227   int limit = Min(last, x_length);
   2228   for (; i < limit; i++) {
   2229     digit_t new_borrow = 0;
   2230     digit_t difference = digit_sub(0, x->digit(i), &new_borrow);
   2231     difference = digit_sub(difference, borrow, &new_borrow);
   2232     result->set_digit(i, difference);
   2233     borrow = new_borrow;
   2234   }
   2235   // Then simulate leading zeroes in {x} as needed.
   2236   for (; i < last; i++) {
   2237     digit_t new_borrow = 0;
   2238     digit_t difference = digit_sub(0, borrow, &new_borrow);
   2239     result->set_digit(i, difference);
   2240     borrow = new_borrow;
   2241   }
   2242 
   2243   // The MSD might contain extra bits that we don't want.
   2244   digit_t msd = last < x_length ? x->digit(last) : 0;
   2245   int msd_bits_consumed = n % kDigitBits;
   2246   digit_t result_msd;
   2247   if (msd_bits_consumed == 0) {
   2248     digit_t new_borrow = 0;
   2249     result_msd = digit_sub(0, msd, &new_borrow);
   2250     result_msd = digit_sub(result_msd, borrow, &new_borrow);
   2251   } else {
   2252     int drop = kDigitBits - msd_bits_consumed;
   2253     msd = (msd << drop) >> drop;
   2254     digit_t minuend_msd = static_cast<digit_t>(1) << (kDigitBits - drop);
   2255     digit_t new_borrow = 0;
   2256     result_msd = digit_sub(minuend_msd, msd, &new_borrow);
   2257     result_msd = digit_sub(result_msd, borrow, &new_borrow);
   2258     DCHECK_EQ(new_borrow, 0);  // result < 2^n.
   2259     // If all subtracted bits were zero, we have to get rid of the
   2260     // materialized minuend_msd again.
   2261     result_msd &= (minuend_msd - 1);
   2262   }
   2263   result->set_digit(last, result_msd);
   2264   result->set_sign(result_sign);
   2265   return MakeImmutable(result);
   2266 }
   2267 
   2268 Handle<BigInt> BigInt::FromInt64(Isolate* isolate, int64_t n) {
   2269   if (n == 0) return MutableBigInt::Zero(isolate);
   2270   STATIC_ASSERT(kDigitBits == 64 || kDigitBits == 32);
   2271   int length = 64 / kDigitBits;
   2272   Handle<MutableBigInt> result =
   2273       MutableBigInt::Cast(isolate->factory()->NewBigInt(length));
   2274   bool sign = n < 0;
   2275   result->initialize_bitfield(sign, length);
   2276   uint64_t absolute;
   2277   if (!sign) {
   2278     absolute = static_cast<uint64_t>(n);
   2279   } else {
   2280     if (n == std::numeric_limits<int64_t>::min()) {
   2281       absolute = static_cast<uint64_t>(std::numeric_limits<int64_t>::max()) + 1;
   2282     } else {
   2283       absolute = static_cast<uint64_t>(-n);
   2284     }
   2285   }
   2286   result->set_64_bits(absolute);
   2287   return MutableBigInt::MakeImmutable(result);
   2288 }
   2289 
   2290 Handle<BigInt> BigInt::FromUint64(Isolate* isolate, uint64_t n) {
   2291   if (n == 0) return MutableBigInt::Zero(isolate);
   2292   STATIC_ASSERT(kDigitBits == 64 || kDigitBits == 32);
   2293   int length = 64 / kDigitBits;
   2294   Handle<MutableBigInt> result =
   2295       MutableBigInt::Cast(isolate->factory()->NewBigInt(length));
   2296   result->initialize_bitfield(false, length);
   2297   result->set_64_bits(n);
   2298   return MutableBigInt::MakeImmutable(result);
   2299 }
   2300 
   2301 MaybeHandle<BigInt> BigInt::FromWords64(Isolate* isolate, int sign_bit,
   2302                                         int words64_count,
   2303                                         const uint64_t* words) {
   2304   if (words64_count < 0 || words64_count > kMaxLength / (64 / kDigitBits)) {
   2305     THROW_NEW_ERROR(isolate, NewRangeError(MessageTemplate::kBigIntTooBig),
   2306                     BigInt);
   2307   }
   2308   if (words64_count == 0) return MutableBigInt::Zero(isolate);
   2309   STATIC_ASSERT(kDigitBits == 64 || kDigitBits == 32);
   2310   int length = (64 / kDigitBits) * words64_count;
   2311   DCHECK_GT(length, 0);
   2312   if (kDigitBits == 32 && words[words64_count - 1] <= (1ULL << 32)) length--;
   2313 
   2314   Handle<MutableBigInt> result;
   2315   if (!MutableBigInt::New(isolate, length).ToHandle(&result)) {
   2316     return MaybeHandle<BigInt>();
   2317   }
   2318 
   2319   result->set_sign(sign_bit);
   2320   if (kDigitBits == 64) {
   2321     for (int i = 0; i < length; ++i) {
   2322       result->set_digit(i, static_cast<digit_t>(words[i]));
   2323     }
   2324   } else {
   2325     for (int i = 0; i < length; i += 2) {
   2326       digit_t lo = static_cast<digit_t>(words[i / 2]);
   2327       digit_t hi = static_cast<digit_t>(words[i / 2] >> 32);
   2328       result->set_digit(i, lo);
   2329       if (i + 1 < length) result->set_digit(i + 1, hi);
   2330     }
   2331   }
   2332 
   2333   return MutableBigInt::MakeImmutable(result);
   2334 }
   2335 
   2336 int BigInt::Words64Count() {
   2337   STATIC_ASSERT(kDigitBits == 64 || kDigitBits == 32);
   2338   return length() / (64 / kDigitBits) +
   2339          (kDigitBits == 32 && length() % 2 == 1 ? 1 : 0);
   2340 }
   2341 
   2342 void BigInt::ToWordsArray64(int* sign_bit, int* words64_count,
   2343                             uint64_t* words) {
   2344   DCHECK_NE(sign_bit, nullptr);
   2345   DCHECK_NE(words64_count, nullptr);
   2346   *sign_bit = sign();
   2347   int available_words = *words64_count;
   2348   *words64_count = Words64Count();
   2349   if (available_words == 0) return;
   2350   DCHECK_NE(words, nullptr);
   2351 
   2352   int len = length();
   2353   if (kDigitBits == 64) {
   2354     for (int i = 0; i < len && i < available_words; ++i) words[i] = digit(i);
   2355   } else {
   2356     for (int i = 0; i < len && available_words > 0; i += 2) {
   2357       uint64_t lo = digit(i);
   2358       uint64_t hi = (i + 1) < len ? digit(i + 1) : 0;
   2359       words[i / 2] = lo | (hi << 32);
   2360       available_words--;
   2361     }
   2362   }
   2363 }
   2364 
   2365 uint64_t MutableBigInt::GetRawBits(BigIntBase* x, bool* lossless) {
   2366   if (lossless != nullptr) *lossless = true;
   2367   if (x->is_zero()) return 0;
   2368   int len = x->length();
   2369   STATIC_ASSERT(kDigitBits == 64 || kDigitBits == 32);
   2370   if (lossless != nullptr && len > 64 / kDigitBits) *lossless = false;
   2371   uint64_t raw = static_cast<uint64_t>(x->digit(0));
   2372   if (kDigitBits == 32 && len > 1) {
   2373     raw |= static_cast<uint64_t>(x->digit(1)) << 32;
   2374   }
   2375   // Simulate two's complement. MSVC dislikes "-raw".
   2376   return x->sign() ? ((~raw) + 1u) : raw;
   2377 }
   2378 
   2379 int64_t BigInt::AsInt64(bool* lossless) {
   2380   uint64_t raw = MutableBigInt::GetRawBits(this, lossless);
   2381   int64_t result = static_cast<int64_t>(raw);
   2382   if (lossless != nullptr && (result < 0) != sign()) *lossless = false;
   2383   return result;
   2384 }
   2385 
   2386 uint64_t BigInt::AsUint64(bool* lossless) {
   2387   uint64_t result = MutableBigInt::GetRawBits(this, lossless);
   2388   if (lossless != nullptr && sign()) *lossless = false;
   2389   return result;
   2390 }
   2391 
   2392 // Digit arithmetic helpers.
   2393 
   2394 #if V8_TARGET_ARCH_32_BIT
   2395 #define HAVE_TWODIGIT_T 1
   2396 typedef uint64_t twodigit_t;
   2397 #elif defined(__SIZEOF_INT128__)
   2398 // Both Clang and GCC support this on x64.
   2399 #define HAVE_TWODIGIT_T 1
   2400 typedef __uint128_t twodigit_t;
   2401 #endif
   2402 
   2403 // {carry} must point to an initialized digit_t and will either be incremented
   2404 // by one or left alone.
   2405 inline BigInt::digit_t MutableBigInt::digit_add(digit_t a, digit_t b,
   2406                                                 digit_t* carry) {
   2407 #if HAVE_TWODIGIT_T
   2408   twodigit_t result = static_cast<twodigit_t>(a) + static_cast<twodigit_t>(b);
   2409   *carry += result >> kDigitBits;
   2410   return static_cast<digit_t>(result);
   2411 #else
   2412   digit_t result = a + b;
   2413   if (result < a) *carry += 1;
   2414   return result;
   2415 #endif
   2416 }
   2417 
   2418 // {borrow} must point to an initialized digit_t and will either be incremented
   2419 // by one or left alone.
   2420 inline BigInt::digit_t MutableBigInt::digit_sub(digit_t a, digit_t b,
   2421                                                 digit_t* borrow) {
   2422 #if HAVE_TWODIGIT_T
   2423   twodigit_t result = static_cast<twodigit_t>(a) - static_cast<twodigit_t>(b);
   2424   *borrow += (result >> kDigitBits) & 1;
   2425   return static_cast<digit_t>(result);
   2426 #else
   2427   digit_t result = a - b;
   2428   if (result > a) *borrow += 1;
   2429   return static_cast<digit_t>(result);
   2430 #endif
   2431 }
   2432 
   2433 // Returns the low half of the result. High half is in {high}.
   2434 inline BigInt::digit_t MutableBigInt::digit_mul(digit_t a, digit_t b,
   2435                                                 digit_t* high) {
   2436 #if HAVE_TWODIGIT_T
   2437   twodigit_t result = static_cast<twodigit_t>(a) * static_cast<twodigit_t>(b);
   2438   *high = result >> kDigitBits;
   2439   return static_cast<digit_t>(result);
   2440 #else
   2441   // Multiply in half-pointer-sized chunks.
   2442   // For inputs [AH AL]*[BH BL], the result is:
   2443   //
   2444   //            [AL*BL]  // r_low
   2445   //    +    [AL*BH]     // r_mid1
   2446   //    +    [AH*BL]     // r_mid2
   2447   //    + [AH*BH]        // r_high
   2448   //    = [R4 R3 R2 R1]  // high = [R4 R3], low = [R2 R1]
   2449   //
   2450   // Where of course we must be careful with carries between the columns.
   2451   digit_t a_low = a & kHalfDigitMask;
   2452   digit_t a_high = a >> kHalfDigitBits;
   2453   digit_t b_low = b & kHalfDigitMask;
   2454   digit_t b_high = b >> kHalfDigitBits;
   2455 
   2456   digit_t r_low = a_low * b_low;
   2457   digit_t r_mid1 = a_low * b_high;
   2458   digit_t r_mid2 = a_high * b_low;
   2459   digit_t r_high = a_high * b_high;
   2460 
   2461   digit_t carry = 0;
   2462   digit_t low = digit_add(r_low, r_mid1 << kHalfDigitBits, &carry);
   2463   low = digit_add(low, r_mid2 << kHalfDigitBits, &carry);
   2464   *high =
   2465       (r_mid1 >> kHalfDigitBits) + (r_mid2 >> kHalfDigitBits) + r_high + carry;
   2466   return low;
   2467 #endif
   2468 }
   2469 
   2470 // Returns the quotient.
   2471 // quotient = (high << kDigitBits + low - remainder) / divisor
   2472 BigInt::digit_t MutableBigInt::digit_div(digit_t high, digit_t low,
   2473                                          digit_t divisor, digit_t* remainder) {
   2474   DCHECK(high < divisor);
   2475 #if V8_TARGET_ARCH_X64 && (__GNUC__ || __clang__)
   2476   digit_t quotient;
   2477   digit_t rem;
   2478   __asm__("divq  %[divisor]"
   2479           // Outputs: {quotient} will be in rax, {rem} in rdx.
   2480           : "=a"(quotient), "=d"(rem)
   2481           // Inputs: put {high} into rdx, {low} into rax, and {divisor} into
   2482           // any register or stack slot.
   2483           : "d"(high), "a"(low), [divisor] "rm"(divisor));
   2484   *remainder = rem;
   2485   return quotient;
   2486 #elif V8_TARGET_ARCH_IA32 && (__GNUC__ || __clang__)
   2487   digit_t quotient;
   2488   digit_t rem;
   2489   __asm__("divl  %[divisor]"
   2490           // Outputs: {quotient} will be in eax, {rem} in edx.
   2491           : "=a"(quotient), "=d"(rem)
   2492           // Inputs: put {high} into edx, {low} into eax, and {divisor} into
   2493           // any register or stack slot.
   2494           : "d"(high), "a"(low), [divisor] "rm"(divisor));
   2495   *remainder = rem;
   2496   return quotient;
   2497 #else
   2498   static const digit_t kHalfDigitBase = 1ull << kHalfDigitBits;
   2499   // Adapted from Warren, Hacker's Delight, p. 152.
   2500   int s = base::bits::CountLeadingZeros(divisor);
   2501   DCHECK_NE(s, kDigitBits);  // {divisor} is not 0.
   2502   divisor <<= s;
   2503 
   2504   digit_t vn1 = divisor >> kHalfDigitBits;
   2505   digit_t vn0 = divisor & kHalfDigitMask;
   2506   // {s} can be 0. {low >> kDigitBits} would be undefined behavior, so
   2507   // we mask the shift amount with {kShiftMask}, and the result with
   2508   // {s_zero_mask} which is 0 if s == 0 and all 1-bits otherwise.
   2509   STATIC_ASSERT(sizeof(intptr_t) == sizeof(digit_t));
   2510   const int kShiftMask = kDigitBits - 1;
   2511   digit_t s_zero_mask =
   2512       static_cast<digit_t>(static_cast<intptr_t>(-s) >> (kDigitBits - 1));
   2513   digit_t un32 =
   2514       (high << s) | ((low >> ((kDigitBits - s) & kShiftMask)) & s_zero_mask);
   2515   digit_t un10 = low << s;
   2516   digit_t un1 = un10 >> kHalfDigitBits;
   2517   digit_t un0 = un10 & kHalfDigitMask;
   2518   digit_t q1 = un32 / vn1;
   2519   digit_t rhat = un32 - q1 * vn1;
   2520 
   2521   while (q1 >= kHalfDigitBase || q1 * vn0 > rhat * kHalfDigitBase + un1) {
   2522     q1--;
   2523     rhat += vn1;
   2524     if (rhat >= kHalfDigitBase) break;
   2525   }
   2526 
   2527   digit_t un21 = un32 * kHalfDigitBase + un1 - q1 * divisor;
   2528   digit_t q0 = un21 / vn1;
   2529   rhat = un21 - q0 * vn1;
   2530 
   2531   while (q0 >= kHalfDigitBase || q0 * vn0 > rhat * kHalfDigitBase + un0) {
   2532     q0--;
   2533     rhat += vn1;
   2534     if (rhat >= kHalfDigitBase) break;
   2535   }
   2536 
   2537   *remainder = (un21 * kHalfDigitBase + un0 - q0 * divisor) >> s;
   2538   return q1 * kHalfDigitBase + q0;
   2539 #endif
   2540 }
   2541 
   2542 // Raises {base} to the power of {exponent}. Does not check for overflow.
   2543 BigInt::digit_t MutableBigInt::digit_pow(digit_t base, digit_t exponent) {
   2544   digit_t result = 1ull;
   2545   while (exponent > 0) {
   2546     if (exponent & 1) {
   2547       result *= base;
   2548     }
   2549     exponent >>= 1;
   2550     base *= base;
   2551   }
   2552   return result;
   2553 }
   2554 
   2555 #undef HAVE_TWODIGIT_T
   2556 
   2557 void MutableBigInt::set_64_bits(uint64_t bits) {
   2558   STATIC_ASSERT(kDigitBits == 64 || kDigitBits == 32);
   2559   if (kDigitBits == 64) {
   2560     set_digit(0, static_cast<digit_t>(bits));
   2561   } else {
   2562     set_digit(0, static_cast<digit_t>(bits & 0xFFFFFFFFu));
   2563     set_digit(1, static_cast<digit_t>(bits >> 32));
   2564   }
   2565 }
   2566 
   2567 #ifdef OBJECT_PRINT
   2568 void BigInt::BigIntPrint(std::ostream& os) {
   2569   DisallowHeapAllocation no_gc;
   2570   HeapObject::PrintHeader(os, "BigInt");
   2571   int len = length();
   2572   os << "\n- length: " << len;
   2573   os << "\n- sign: " << sign();
   2574   if (len > 0) {
   2575     os << "\n- digits:";
   2576     for (int i = 0; i < len; i++) {
   2577       os << "\n    0x" << std::hex << digit(i);
   2578     }
   2579   }
   2580   os << std::dec << "\n";
   2581 }
   2582 #endif  // OBJECT_PRINT
   2583 
   2584 }  // namespace internal
   2585 }  // namespace v8
   2586