1 /*
   2  * Copyright (c) 1996, 2007, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.  Oracle designates this
   8  * particular file as subject to the "Classpath" exception as provided
   9  * by Oracle in the LICENSE file that accompanied this code.
  10  *
  11  * This code is distributed in the hope that it will be useful, but WITHOUT
  12  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  13  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  14  * version 2 for more details (a copy is included in the LICENSE file that
  15  * accompanied this code).
  16  *
  17  * You should have received a copy of the GNU General Public License version
  18  * 2 along with this work; if not, write to the Free Software Foundation,
  19  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  20  *
  21  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  22  * or visit www.oracle.com if you need additional information or have any
  23  * questions.
  24  */
  25 
  26 /*
  27  * Portions Copyright (c) 1995  Colin Plumb.  All rights reserved.
  28  */
  29 
  30 package java.math;
  31 
  32 import java.util.Random;
  33 import java.io.*;
  34 
  35 /**
  36  * Immutable arbitrary-precision integers.  All operations behave as if
  37  * BigIntegers were represented in two's-complement notation (like Java's
  38  * primitive integer types).  BigInteger provides analogues to all of Java's
  39  * primitive integer operators, and all relevant methods from java.lang.Math.
  40  * Additionally, BigInteger provides operations for modular arithmetic, GCD
  41  * calculation, primality testing, prime generation, bit manipulation,
  42  * and a few other miscellaneous operations.
  43  *
  44  * <p>Semantics of arithmetic operations exactly mimic those of Java's integer
  45  * arithmetic operators, as defined in <i>The Java Language Specification</i>.
  46  * For example, division by zero throws an {@code ArithmeticException}, and
  47  * division of a negative by a positive yields a negative (or zero) remainder.
  48  * All of the details in the Spec concerning overflow are ignored, as
  49  * BigIntegers are made as large as necessary to accommodate the results of an
  50  * operation.
  51  *
  52  * <p>Semantics of shift operations extend those of Java's shift operators
  53  * to allow for negative shift distances.  A right-shift with a negative
  54  * shift distance results in a left shift, and vice-versa.  The unsigned
  55  * right shift operator ({@code >>>}) is omitted, as this operation makes
  56  * little sense in combination with the "infinite word size" abstraction
  57  * provided by this class.
  58  *
  59  * <p>Semantics of bitwise logical operations exactly mimic those of Java's
  60  * bitwise integer operators.  The binary operators ({@code and},
  61  * {@code or}, {@code xor}) implicitly perform sign extension on the shorter
  62  * of the two operands prior to performing the operation.
  63  *
  64  * <p>Comparison operations perform signed integer comparisons, analogous to
  65  * those performed by Java's relational and equality operators.
  66  *
  67  * <p>Modular arithmetic operations are provided to compute residues, perform
  68  * exponentiation, and compute multiplicative inverses.  These methods always
  69  * return a non-negative result, between {@code 0} and {@code (modulus - 1)},
  70  * inclusive.
  71  *
  72  * <p>Bit operations operate on a single bit of the two's-complement
  73  * representation of their operand.  If necessary, the operand is sign-
  74  * extended so that it contains the designated bit.  None of the single-bit
  75  * operations can produce a BigInteger with a different sign from the
  76  * BigInteger being operated on, as they affect only a single bit, and the
  77  * "infinite word size" abstraction provided by this class ensures that there
  78  * are infinitely many "virtual sign bits" preceding each BigInteger.
  79  *
  80  * <p>For the sake of brevity and clarity, pseudo-code is used throughout the
  81  * descriptions of BigInteger methods.  The pseudo-code expression
  82  * {@code (i + j)} is shorthand for "a BigInteger whose value is
  83  * that of the BigInteger {@code i} plus that of the BigInteger {@code j}."
  84  * The pseudo-code expression {@code (i == j)} is shorthand for
  85  * "{@code true} if and only if the BigInteger {@code i} represents the same
  86  * value as the BigInteger {@code j}."  Other pseudo-code expressions are
  87  * interpreted similarly.
  88  *
  89  * <p>All methods and constructors in this class throw
  90  * {@code NullPointerException} when passed
  91  * a null object reference for any input parameter.
  92  *
  93  * @see     BigDecimal
  94  * @author  Josh Bloch
  95  * @author  Michael McCloskey
  96  * @since JDK1.1
  97  */
  98 
  99 public class BigInteger extends Number implements Comparable<BigInteger> {
 100     /**
 101      * The signum of this BigInteger: -1 for negative, 0 for zero, or
 102      * 1 for positive.  Note that the BigInteger zero <i>must</i> have
 103      * a signum of 0.  This is necessary to ensures that there is exactly one
 104      * representation for each BigInteger value.
 105      *
 106      * @serial
 107      */
 108     final int signum;
 109 
 110     /**
 111      * The magnitude of this BigInteger, in <i>big-endian</i> order: the
 112      * zeroth element of this array is the most-significant int of the
 113      * magnitude.  The magnitude must be "minimal" in that the most-significant
 114      * int ({@code mag[0]}) must be non-zero.  This is necessary to
 115      * ensure that there is exactly one representation for each BigInteger
 116      * value.  Note that this implies that the BigInteger zero has a
 117      * zero-length mag array.
 118      */
 119     final int[] mag;
 120 
 121     // These "redundant fields" are initialized with recognizable nonsense
 122     // values, and cached the first time they are needed (or never, if they
 123     // aren't needed).
 124 
 125      /**
 126      * One plus the bitCount of this BigInteger. Zeros means unitialized.
 127      *
 128      * @serial
 129      * @see #bitCount
 130      * @deprecated Deprecated since logical value is offset from stored
 131      * value and correction factor is applied in accessor method.
 132      */
 133     @Deprecated
 134     private int bitCount;
 135 
 136     /**
 137      * One plus the bitLength of this BigInteger. Zeros means unitialized.
 138      * (either value is acceptable).
 139      *
 140      * @serial
 141      * @see #bitLength()
 142      * @deprecated Deprecated since logical value is offset from stored
 143      * value and correction factor is applied in accessor method.
 144      */
 145     @Deprecated
 146     private int bitLength;
 147 
 148     /**
 149      * Two plus the lowest set bit of this BigInteger, as returned by
 150      * getLowestSetBit().
 151      *
 152      * @serial
 153      * @see #getLowestSetBit
 154      * @deprecated Deprecated since logical value is offset from stored
 155      * value and correction factor is applied in accessor method.
 156      */
 157     @Deprecated
 158     private int lowestSetBit;
 159 
 160     /**
 161      * Two plus the index of the lowest-order int in the magnitude of this
 162      * BigInteger that contains a nonzero int, or -2 (either value is acceptable).
 163      * The least significant int has int-number 0, the next int in order of
 164      * increasing significance has int-number 1, and so forth.
 165      * @deprecated Deprecated since logical value is offset from stored
 166      * value and correction factor is applied in accessor method.
 167      */
 168     @Deprecated
 169     private int firstNonzeroIntNum;
 170 
 171     /**
 172      * This mask is used to obtain the value of an int as if it were unsigned.
 173      */
 174     final static long LONG_MASK = 0xffffffffL;
 175 
 176     //Constructors
 177 
 178     /**
 179      * Translates a byte array containing the two's-complement binary
 180      * representation of a BigInteger into a BigInteger.  The input array is
 181      * assumed to be in <i>big-endian</i> byte-order: the most significant
 182      * byte is in the zeroth element.
 183      *
 184      * @param  val big-endian two's-complement binary representation of
 185      *         BigInteger.
 186      * @throws NumberFormatException {@code val} is zero bytes long.
 187      */
 188     public BigInteger(byte[] val) {
 189         if (val.length == 0)
 190             throw new NumberFormatException("Zero length BigInteger");
 191 
 192         if (val[0] < 0) {
 193             mag = makePositive(val);
 194             signum = -1;
 195         } else {
 196             mag = stripLeadingZeroBytes(val);
 197             signum = (mag.length == 0 ? 0 : 1);
 198         }
 199     }
 200 
 201     /**
 202      * This private constructor translates an int array containing the
 203      * two's-complement binary representation of a BigInteger into a
 204      * BigInteger. The input array is assumed to be in <i>big-endian</i>
 205      * int-order: the most significant int is in the zeroth element.
 206      */
 207     private BigInteger(int[] val) {
 208         if (val.length == 0)
 209             throw new NumberFormatException("Zero length BigInteger");
 210 
 211         if (val[0] < 0) {
 212             mag = makePositive(val);
 213             signum = -1;
 214         } else {
 215             mag = trustedStripLeadingZeroInts(val);
 216             signum = (mag.length == 0 ? 0 : 1);
 217         }
 218     }
 219 
 220     /**
 221      * Translates the sign-magnitude representation of a BigInteger into a
 222      * BigInteger.  The sign is represented as an integer signum value: -1 for
 223      * negative, 0 for zero, or 1 for positive.  The magnitude is a byte array
 224      * in <i>big-endian</i> byte-order: the most significant byte is in the
 225      * zeroth element.  A zero-length magnitude array is permissible, and will
 226      * result in a BigInteger value of 0, whether signum is -1, 0 or 1.
 227      *
 228      * @param  signum signum of the number (-1 for negative, 0 for zero, 1
 229      *         for positive).
 230      * @param  magnitude big-endian binary representation of the magnitude of
 231      *         the number.
 232      * @throws NumberFormatException {@code signum} is not one of the three
 233      *         legal values (-1, 0, and 1), or {@code signum} is 0 and
 234      *         {@code magnitude} contains one or more non-zero bytes.
 235      */
 236     public BigInteger(int signum, byte[] magnitude) {
 237         this.mag = stripLeadingZeroBytes(magnitude);
 238 
 239         if (signum < -1 || signum > 1)
 240             throw(new NumberFormatException("Invalid signum value"));
 241 
 242         if (this.mag.length==0) {
 243             this.signum = 0;
 244         } else {
 245             if (signum == 0)
 246                 throw(new NumberFormatException("signum-magnitude mismatch"));
 247             this.signum = signum;
 248         }
 249     }
 250 
 251     /**
 252      * A constructor for internal use that translates the sign-magnitude
 253      * representation of a BigInteger into a BigInteger. It checks the
 254      * arguments and copies the magnitude so this constructor would be
 255      * safe for external use.
 256      */
 257     private BigInteger(int signum, int[] magnitude) {
 258         this.mag = stripLeadingZeroInts(magnitude);
 259 
 260         if (signum < -1 || signum > 1)
 261             throw(new NumberFormatException("Invalid signum value"));
 262 
 263         if (this.mag.length==0) {
 264             this.signum = 0;
 265         } else {
 266             if (signum == 0)
 267                 throw(new NumberFormatException("signum-magnitude mismatch"));
 268             this.signum = signum;
 269         }
 270     }
 271 
 272     /**
 273      * Translates the String representation of a BigInteger in the
 274      * specified radix into a BigInteger.  The String representation
 275      * consists of an optional minus or plus sign followed by a
 276      * sequence of one or more digits in the specified radix.  The
 277      * character-to-digit mapping is provided by {@code
 278      * Character.digit}.  The String may not contain any extraneous
 279      * characters (whitespace, for example).
 280      *
 281      * @param val String representation of BigInteger.
 282      * @param radix radix to be used in interpreting {@code val}.
 283      * @throws NumberFormatException {@code val} is not a valid representation
 284      *         of a BigInteger in the specified radix, or {@code radix} is
 285      *         outside the range from {@link Character#MIN_RADIX} to
 286      *         {@link Character#MAX_RADIX}, inclusive.
 287      * @see    Character#digit
 288      */
 289     public BigInteger(String val, int radix) {
 290         int cursor = 0, numDigits;
 291         final int len = val.length();
 292 
 293         if (radix < Character.MIN_RADIX || radix > Character.MAX_RADIX)
 294             throw new NumberFormatException("Radix out of range");
 295         if (len == 0)
 296             throw new NumberFormatException("Zero length BigInteger");
 297 
 298         // Check for at most one leading sign
 299         int sign = 1;
 300         int index1 = val.lastIndexOf('-');
 301         int index2 = val.lastIndexOf('+');
 302         if ((index1 + index2) <= -1) {
 303             // No leading sign character or at most one leading sign character
 304             if (index1 == 0 || index2 == 0) {
 305                 cursor = 1;
 306                 if (len == 1)
 307                     throw new NumberFormatException("Zero length BigInteger");
 308             }
 309             if (index1 == 0)
 310                 sign = -1;
 311         } else
 312             throw new NumberFormatException("Illegal embedded sign character");
 313 
 314         // Skip leading zeros and compute number of digits in magnitude
 315         while (cursor < len &&
 316                Character.digit(val.charAt(cursor), radix) == 0)
 317             cursor++;
 318         if (cursor == len) {
 319             signum = 0;
 320             mag = ZERO.mag;
 321             return;
 322         }
 323 
 324         numDigits = len - cursor;
 325         signum = sign;
 326 
 327         // Pre-allocate array of expected size. May be too large but can
 328         // never be too small. Typically exact.
 329         int numBits = (int)(((numDigits * bitsPerDigit[radix]) >>> 10) + 1);
 330         int numWords = (numBits + 31) >>> 5;
 331         int[] magnitude = new int[numWords];
 332 
 333         // Process first (potentially short) digit group
 334         int firstGroupLen = numDigits % digitsPerInt[radix];
 335         if (firstGroupLen == 0)
 336             firstGroupLen = digitsPerInt[radix];
 337         String group = val.substring(cursor, cursor += firstGroupLen);
 338         magnitude[numWords - 1] = Integer.parseInt(group, radix);
 339         if (magnitude[numWords - 1] < 0)
 340             throw new NumberFormatException("Illegal digit");
 341 
 342         // Process remaining digit groups
 343         int superRadix = intRadix[radix];
 344         int groupVal = 0;
 345         while (cursor < len) {
 346             group = val.substring(cursor, cursor += digitsPerInt[radix]);
 347             groupVal = Integer.parseInt(group, radix);
 348             if (groupVal < 0)
 349                 throw new NumberFormatException("Illegal digit");
 350             destructiveMulAdd(magnitude, superRadix, groupVal);
 351         }
 352         // Required for cases where the array was overallocated.
 353         mag = trustedStripLeadingZeroInts(magnitude);
 354     }
 355 
 356     // Constructs a new BigInteger using a char array with radix=10
 357     BigInteger(char[] val) {
 358         int cursor = 0, numDigits;
 359         int len = val.length;
 360 
 361         // Check for leading minus sign
 362         int sign = 1;
 363         if (val[0] == '-') {
 364             if (len == 1)
 365                 throw new NumberFormatException("Zero length BigInteger");
 366             sign = -1;
 367             cursor = 1;
 368         } else if (val[0] == '+') {
 369             if (len == 1)
 370                 throw new NumberFormatException("Zero length BigInteger");
 371             cursor = 1;
 372         }
 373 
 374         // Skip leading zeros and compute number of digits in magnitude
 375         while (cursor < len && Character.digit(val[cursor], 10) == 0)
 376             cursor++;
 377         if (cursor == len) {
 378             signum = 0;
 379             mag = ZERO.mag;
 380             return;
 381         }
 382 
 383         numDigits = len - cursor;
 384         signum = sign;
 385 
 386         // Pre-allocate array of expected size
 387         int numWords;
 388         if (len < 10) {
 389             numWords = 1;
 390         } else {
 391             int numBits = (int)(((numDigits * bitsPerDigit[10]) >>> 10) + 1);
 392             numWords = (numBits + 31) >>> 5;
 393         }
 394         int[] magnitude = new int[numWords];
 395 
 396         // Process first (potentially short) digit group
 397         int firstGroupLen = numDigits % digitsPerInt[10];
 398         if (firstGroupLen == 0)
 399             firstGroupLen = digitsPerInt[10];
 400         magnitude[numWords - 1] = parseInt(val, cursor,  cursor += firstGroupLen);
 401 
 402         // Process remaining digit groups
 403         while (cursor < len) {
 404             int groupVal = parseInt(val, cursor, cursor += digitsPerInt[10]);
 405             destructiveMulAdd(magnitude, intRadix[10], groupVal);
 406         }
 407         mag = trustedStripLeadingZeroInts(magnitude);
 408     }
 409 
 410     // Create an integer with the digits between the two indexes
 411     // Assumes start < end. The result may be negative, but it
 412     // is to be treated as an unsigned value.
 413     private int parseInt(char[] source, int start, int end) {
 414         int result = Character.digit(source[start++], 10);
 415         if (result == -1)
 416             throw new NumberFormatException(new String(source));
 417 
 418         for (int index = start; index<end; index++) {
 419             int nextVal = Character.digit(source[index], 10);
 420             if (nextVal == -1)
 421                 throw new NumberFormatException(new String(source));
 422             result = 10*result + nextVal;
 423         }
 424 
 425         return result;
 426     }
 427 
 428     // bitsPerDigit in the given radix times 1024
 429     // Rounded up to avoid underallocation.
 430     private static long bitsPerDigit[] = { 0, 0,
 431         1024, 1624, 2048, 2378, 2648, 2875, 3072, 3247, 3402, 3543, 3672,
 432         3790, 3899, 4001, 4096, 4186, 4271, 4350, 4426, 4498, 4567, 4633,
 433         4696, 4756, 4814, 4870, 4923, 4975, 5025, 5074, 5120, 5166, 5210,
 434                                            5253, 5295};
 435 
 436     // Multiply x array times word y in place, and add word z
 437     private static void destructiveMulAdd(int[] x, int y, int z) {
 438         // Perform the multiplication word by word
 439         long ylong = y & LONG_MASK;
 440         long zlong = z & LONG_MASK;
 441         int len = x.length;
 442 
 443         long product = 0;
 444         long carry = 0;
 445         for (int i = len-1; i >= 0; i--) {
 446             product = ylong * (x[i] & LONG_MASK) + carry;
 447             x[i] = (int)product;
 448             carry = product >>> 32;
 449         }
 450 
 451         // Perform the addition
 452         long sum = (x[len-1] & LONG_MASK) + zlong;
 453         x[len-1] = (int)sum;
 454         carry = sum >>> 32;
 455         for (int i = len-2; i >= 0; i--) {
 456             sum = (x[i] & LONG_MASK) + carry;
 457             x[i] = (int)sum;
 458             carry = sum >>> 32;
 459         }
 460     }
 461 
 462     /**
 463      * Translates the decimal String representation of a BigInteger into a
 464      * BigInteger.  The String representation consists of an optional minus
 465      * sign followed by a sequence of one or more decimal digits.  The
 466      * character-to-digit mapping is provided by {@code Character.digit}.
 467      * The String may not contain any extraneous characters (whitespace, for
 468      * example).
 469      *
 470      * @param val decimal String representation of BigInteger.
 471      * @throws NumberFormatException {@code val} is not a valid representation
 472      *         of a BigInteger.
 473      * @see    Character#digit
 474      */
 475     public BigInteger(String val) {
 476         this(val, 10);
 477     }
 478 
 479     /**
 480      * Constructs a randomly generated BigInteger, uniformly distributed over
 481      * the range 0 to (2<sup>{@code numBits}</sup> - 1), inclusive.
 482      * The uniformity of the distribution assumes that a fair source of random
 483      * bits is provided in {@code rnd}.  Note that this constructor always
 484      * constructs a non-negative BigInteger.
 485      *
 486      * @param  numBits maximum bitLength of the new BigInteger.
 487      * @param  rnd source of randomness to be used in computing the new
 488      *         BigInteger.
 489      * @throws IllegalArgumentException {@code numBits} is negative.
 490      * @see #bitLength()
 491      */
 492     public BigInteger(int numBits, Random rnd) {
 493         this(1, randomBits(numBits, rnd));
 494     }
 495 
 496     private static byte[] randomBits(int numBits, Random rnd) {
 497         if (numBits < 0)
 498             throw new IllegalArgumentException("numBits must be non-negative");
 499         int numBytes = (int)(((long)numBits+7)/8); // avoid overflow
 500         byte[] randomBits = new byte[numBytes];
 501 
 502         // Generate random bytes and mask out any excess bits
 503         if (numBytes > 0) {
 504             rnd.nextBytes(randomBits);
 505             int excessBits = 8*numBytes - numBits;
 506             randomBits[0] &= (1 << (8-excessBits)) - 1;
 507         }
 508         return randomBits;
 509     }
 510 
 511     /**
 512      * Constructs a randomly generated positive BigInteger that is probably
 513      * prime, with the specified bitLength.
 514      *
 515      * <p>It is recommended that the {@link #probablePrime probablePrime}
 516      * method be used in preference to this constructor unless there
 517      * is a compelling need to specify a certainty.
 518      *
 519      * @param  bitLength bitLength of the returned BigInteger.
 520      * @param  certainty a measure of the uncertainty that the caller is
 521      *         willing to tolerate.  The probability that the new BigInteger
 522      *         represents a prime number will exceed
 523      *         (1 - 1/2<sup>{@code certainty}</sup>).  The execution time of
 524      *         this constructor is proportional to the value of this parameter.
 525      * @param  rnd source of random bits used to select candidates to be
 526      *         tested for primality.
 527      * @throws ArithmeticException {@code bitLength < 2}.
 528      * @see    #bitLength()
 529      */
 530     public BigInteger(int bitLength, int certainty, Random rnd) {
 531         BigInteger prime;
 532 
 533         if (bitLength < 2)
 534             throw new ArithmeticException("bitLength < 2");
 535         // The cutoff of 95 was chosen empirically for best performance
 536         prime = (bitLength < 95 ? smallPrime(bitLength, certainty, rnd)
 537                                 : largePrime(bitLength, certainty, rnd));
 538         signum = 1;
 539         mag = prime.mag;
 540     }
 541 
 542     // Minimum size in bits that the requested prime number has
 543     // before we use the large prime number generating algorithms
 544     private static final int SMALL_PRIME_THRESHOLD = 95;
 545 
 546     // Certainty required to meet the spec of probablePrime
 547     private static final int DEFAULT_PRIME_CERTAINTY = 100;
 548 
 549     /**
 550      * Returns a positive BigInteger that is probably prime, with the
 551      * specified bitLength. The probability that a BigInteger returned
 552      * by this method is composite does not exceed 2<sup>-100</sup>.
 553      *
 554      * @param  bitLength bitLength of the returned BigInteger.
 555      * @param  rnd source of random bits used to select candidates to be
 556      *         tested for primality.
 557      * @return a BigInteger of {@code bitLength} bits that is probably prime
 558      * @throws ArithmeticException {@code bitLength < 2}.
 559      * @see    #bitLength()
 560      * @since 1.4
 561      */
 562     public static BigInteger probablePrime(int bitLength, Random rnd) {
 563         if (bitLength < 2)
 564             throw new ArithmeticException("bitLength < 2");
 565 
 566         // The cutoff of 95 was chosen empirically for best performance
 567         return (bitLength < SMALL_PRIME_THRESHOLD ?
 568                 smallPrime(bitLength, DEFAULT_PRIME_CERTAINTY, rnd) :
 569                 largePrime(bitLength, DEFAULT_PRIME_CERTAINTY, rnd));
 570     }
 571 
 572     /**
 573      * Find a random number of the specified bitLength that is probably prime.
 574      * This method is used for smaller primes, its performance degrades on
 575      * larger bitlengths.
 576      *
 577      * This method assumes bitLength > 1.
 578      */
 579     private static BigInteger smallPrime(int bitLength, int certainty, Random rnd) {
 580         int magLen = (bitLength + 31) >>> 5;
 581         int temp[] = new int[magLen];
 582         int highBit = 1 << ((bitLength+31) & 0x1f);  // High bit of high int
 583         int highMask = (highBit << 1) - 1;  // Bits to keep in high int
 584 
 585         while(true) {
 586             // Construct a candidate
 587             for (int i=0; i<magLen; i++)
 588                 temp[i] = rnd.nextInt();
 589             temp[0] = (temp[0] & highMask) | highBit;  // Ensure exact length
 590             if (bitLength > 2)
 591                 temp[magLen-1] |= 1;  // Make odd if bitlen > 2
 592 
 593             BigInteger p = new BigInteger(temp, 1);
 594 
 595             // Do cheap "pre-test" if applicable
 596             if (bitLength > 6) {
 597                 long r = p.remainder(SMALL_PRIME_PRODUCT).longValue();
 598                 if ((r%3==0)  || (r%5==0)  || (r%7==0)  || (r%11==0) ||
 599                     (r%13==0) || (r%17==0) || (r%19==0) || (r%23==0) ||
 600                     (r%29==0) || (r%31==0) || (r%37==0) || (r%41==0))
 601                     continue; // Candidate is composite; try another
 602             }
 603 
 604             // All candidates of bitLength 2 and 3 are prime by this point
 605             if (bitLength < 4)
 606                 return p;
 607 
 608             // Do expensive test if we survive pre-test (or it's inapplicable)
 609             if (p.primeToCertainty(certainty, rnd))
 610                 return p;
 611         }
 612     }
 613 
 614     private static final BigInteger SMALL_PRIME_PRODUCT
 615                        = valueOf(3L*5*7*11*13*17*19*23*29*31*37*41);
 616 
 617     /**
 618      * Find a random number of the specified bitLength that is probably prime.
 619      * This method is more appropriate for larger bitlengths since it uses
 620      * a sieve to eliminate most composites before using a more expensive
 621      * test.
 622      */
 623     private static BigInteger largePrime(int bitLength, int certainty, Random rnd) {
 624         BigInteger p;
 625         p = new BigInteger(bitLength, rnd).setBit(bitLength-1);
 626         p.mag[p.mag.length-1] &= 0xfffffffe;
 627 
 628         // Use a sieve length likely to contain the next prime number
 629         int searchLen = (bitLength / 20) * 64;
 630         BitSieve searchSieve = new BitSieve(p, searchLen);
 631         BigInteger candidate = searchSieve.retrieve(p, certainty, rnd);
 632 
 633         while ((candidate == null) || (candidate.bitLength() != bitLength)) {
 634             p = p.add(BigInteger.valueOf(2*searchLen));
 635             if (p.bitLength() != bitLength)
 636                 p = new BigInteger(bitLength, rnd).setBit(bitLength-1);
 637             p.mag[p.mag.length-1] &= 0xfffffffe;
 638             searchSieve = new BitSieve(p, searchLen);
 639             candidate = searchSieve.retrieve(p, certainty, rnd);
 640         }
 641         return candidate;
 642     }
 643 
 644    /**
 645     * Returns the first integer greater than this {@code BigInteger} that
 646     * is probably prime.  The probability that the number returned by this
 647     * method is composite does not exceed 2<sup>-100</sup>. This method will
 648     * never skip over a prime when searching: if it returns {@code p}, there
 649     * is no prime {@code q} such that {@code this < q < p}.
 650     *
 651     * @return the first integer greater than this {@code BigInteger} that
 652     *         is probably prime.
 653     * @throws ArithmeticException {@code this < 0}.
 654     * @since 1.5
 655     */
 656     public BigInteger nextProbablePrime() {
 657         if (this.signum < 0)
 658             throw new ArithmeticException("start < 0: " + this);
 659 
 660         // Handle trivial cases
 661         if ((this.signum == 0) || this.equals(ONE))
 662             return TWO;
 663 
 664         BigInteger result = this.add(ONE);
 665 
 666         // Fastpath for small numbers
 667         if (result.bitLength() < SMALL_PRIME_THRESHOLD) {
 668 
 669             // Ensure an odd number
 670             if (!result.testBit(0))
 671                 result = result.add(ONE);
 672 
 673             while(true) {
 674                 // Do cheap "pre-test" if applicable
 675                 if (result.bitLength() > 6) {
 676                     long r = result.remainder(SMALL_PRIME_PRODUCT).longValue();
 677                     if ((r%3==0)  || (r%5==0)  || (r%7==0)  || (r%11==0) ||
 678                         (r%13==0) || (r%17==0) || (r%19==0) || (r%23==0) ||
 679                         (r%29==0) || (r%31==0) || (r%37==0) || (r%41==0)) {
 680                         result = result.add(TWO);
 681                         continue; // Candidate is composite; try another
 682                     }
 683                 }
 684 
 685                 // All candidates of bitLength 2 and 3 are prime by this point
 686                 if (result.bitLength() < 4)
 687                     return result;
 688 
 689                 // The expensive test
 690                 if (result.primeToCertainty(DEFAULT_PRIME_CERTAINTY, null))
 691                     return result;
 692 
 693                 result = result.add(TWO);
 694             }
 695         }
 696 
 697         // Start at previous even number
 698         if (result.testBit(0))
 699             result = result.subtract(ONE);
 700 
 701         // Looking for the next large prime
 702         int searchLen = (result.bitLength() / 20) * 64;
 703 
 704         while(true) {
 705            BitSieve searchSieve = new BitSieve(result, searchLen);
 706            BigInteger candidate = searchSieve.retrieve(result,
 707                                                  DEFAULT_PRIME_CERTAINTY, null);
 708            if (candidate != null)
 709                return candidate;
 710            result = result.add(BigInteger.valueOf(2 * searchLen));
 711         }
 712     }
 713 
 714     /**
 715      * Returns {@code true} if this BigInteger is probably prime,
 716      * {@code false} if it's definitely composite.
 717      *
 718      * This method assumes bitLength > 2.
 719      *
 720      * @param  certainty a measure of the uncertainty that the caller is
 721      *         willing to tolerate: if the call returns {@code true}
 722      *         the probability that this BigInteger is prime exceeds
 723      *         {@code (1 - 1/2<sup>certainty</sup>)}.  The execution time of
 724      *         this method is proportional to the value of this parameter.
 725      * @return {@code true} if this BigInteger is probably prime,
 726      *         {@code false} if it's definitely composite.
 727      */
 728     boolean primeToCertainty(int certainty, Random random) {
 729         int rounds = 0;
 730         int n = (Math.min(certainty, Integer.MAX_VALUE-1)+1)/2;
 731 
 732         // The relationship between the certainty and the number of rounds
 733         // we perform is given in the draft standard ANSI X9.80, "PRIME
 734         // NUMBER GENERATION, PRIMALITY TESTING, AND PRIMALITY CERTIFICATES".
 735         int sizeInBits = this.bitLength();
 736         if (sizeInBits < 100) {
 737             rounds = 50;
 738             rounds = n < rounds ? n : rounds;
 739             return passesMillerRabin(rounds, random);
 740         }
 741 
 742         if (sizeInBits < 256) {
 743             rounds = 27;
 744         } else if (sizeInBits < 512) {
 745             rounds = 15;
 746         } else if (sizeInBits < 768) {
 747             rounds = 8;
 748         } else if (sizeInBits < 1024) {
 749             rounds = 4;
 750         } else {
 751             rounds = 2;
 752         }
 753         rounds = n < rounds ? n : rounds;
 754 
 755         return passesMillerRabin(rounds, random) && passesLucasLehmer();
 756     }
 757 
 758     /**
 759      * Returns true iff this BigInteger is a Lucas-Lehmer probable prime.
 760      *
 761      * The following assumptions are made:
 762      * This BigInteger is a positive, odd number.
 763      */
 764     private boolean passesLucasLehmer() {
 765         BigInteger thisPlusOne = this.add(ONE);
 766 
 767         // Step 1
 768         int d = 5;
 769         while (jacobiSymbol(d, this) != -1) {
 770             // 5, -7, 9, -11, ...
 771             d = (d<0) ? Math.abs(d)+2 : -(d+2);
 772         }
 773 
 774         // Step 2
 775         BigInteger u = lucasLehmerSequence(d, thisPlusOne, this);
 776 
 777         // Step 3
 778         return u.mod(this).equals(ZERO);
 779     }
 780 
 781     /**
 782      * Computes Jacobi(p,n).
 783      * Assumes n positive, odd, n>=3.
 784      */
 785     private static int jacobiSymbol(int p, BigInteger n) {
 786         if (p == 0)
 787             return 0;
 788 
 789         // Algorithm and comments adapted from Colin Plumb's C library.
 790         int j = 1;
 791         int u = n.mag[n.mag.length-1];
 792 
 793         // Make p positive
 794         if (p < 0) {
 795             p = -p;
 796             int n8 = u & 7;
 797             if ((n8 == 3) || (n8 == 7))
 798                 j = -j; // 3 (011) or 7 (111) mod 8
 799         }
 800 
 801         // Get rid of factors of 2 in p
 802         while ((p & 3) == 0)
 803             p >>= 2;
 804         if ((p & 1) == 0) {
 805             p >>= 1;
 806             if (((u ^ (u>>1)) & 2) != 0)
 807                 j = -j; // 3 (011) or 5 (101) mod 8
 808         }
 809         if (p == 1)
 810             return j;
 811         // Then, apply quadratic reciprocity
 812         if ((p & u & 2) != 0)   // p = u = 3 (mod 4)?
 813             j = -j;
 814         // And reduce u mod p
 815         u = n.mod(BigInteger.valueOf(p)).intValue();
 816 
 817         // Now compute Jacobi(u,p), u < p
 818         while (u != 0) {
 819             while ((u & 3) == 0)
 820                 u >>= 2;
 821             if ((u & 1) == 0) {
 822                 u >>= 1;
 823                 if (((p ^ (p>>1)) & 2) != 0)
 824                     j = -j;     // 3 (011) or 5 (101) mod 8
 825             }
 826             if (u == 1)
 827                 return j;
 828             // Now both u and p are odd, so use quadratic reciprocity
 829             assert (u < p);
 830             int t = u; u = p; p = t;
 831             if ((u & p & 2) != 0) // u = p = 3 (mod 4)?
 832                 j = -j;
 833             // Now u >= p, so it can be reduced
 834             u %= p;
 835         }
 836         return 0;
 837     }
 838 
 839     private static BigInteger lucasLehmerSequence(int z, BigInteger k, BigInteger n) {
 840         BigInteger d = BigInteger.valueOf(z);
 841         BigInteger u = ONE; BigInteger u2;
 842         BigInteger v = ONE; BigInteger v2;
 843 
 844         for (int i=k.bitLength()-2; i>=0; i--) {
 845             u2 = u.multiply(v).mod(n);
 846 
 847             v2 = v.square().add(d.multiply(u.square())).mod(n);
 848             if (v2.testBit(0))
 849                 v2 = v2.subtract(n);
 850 
 851             v2 = v2.shiftRight(1);
 852 
 853             u = u2; v = v2;
 854             if (k.testBit(i)) {
 855                 u2 = u.add(v).mod(n);
 856                 if (u2.testBit(0))
 857                     u2 = u2.subtract(n);
 858 
 859                 u2 = u2.shiftRight(1);
 860                 v2 = v.add(d.multiply(u)).mod(n);
 861                 if (v2.testBit(0))
 862                     v2 = v2.subtract(n);
 863                 v2 = v2.shiftRight(1);
 864 
 865                 u = u2; v = v2;
 866             }
 867         }
 868         return u;
 869     }
 870 
 871     private static volatile Random staticRandom;
 872 
 873     private static Random getSecureRandom() {
 874         if (staticRandom == null) {
 875             staticRandom = new java.security.SecureRandom();
 876         }
 877         return staticRandom;
 878     }
 879 
 880     /**
 881      * Returns true iff this BigInteger passes the specified number of
 882      * Miller-Rabin tests. This test is taken from the DSA spec (NIST FIPS
 883      * 186-2).
 884      *
 885      * The following assumptions are made:
 886      * This BigInteger is a positive, odd number greater than 2.
 887      * iterations<=50.
 888      */
 889     private boolean passesMillerRabin(int iterations, Random rnd) {
 890         // Find a and m such that m is odd and this == 1 + 2**a * m
 891         BigInteger thisMinusOne = this.subtract(ONE);
 892         BigInteger m = thisMinusOne;
 893         int a = m.getLowestSetBit();
 894         m = m.shiftRight(a);
 895 
 896         // Do the tests
 897         if (rnd == null) {
 898             rnd = getSecureRandom();
 899         }
 900         for (int i=0; i<iterations; i++) {
 901             // Generate a uniform random on (1, this)
 902             BigInteger b;
 903             do {
 904                 b = new BigInteger(this.bitLength(), rnd);
 905             } while (b.compareTo(ONE) <= 0 || b.compareTo(this) >= 0);
 906 
 907             int j = 0;
 908             BigInteger z = b.modPow(m, this);
 909             while(!((j==0 && z.equals(ONE)) || z.equals(thisMinusOne))) {
 910                 if (j>0 && z.equals(ONE) || ++j==a)
 911                     return false;
 912                 z = z.modPow(TWO, this);
 913             }
 914         }
 915         return true;
 916     }
 917 
 918     /**
 919      * This internal constructor differs from its public cousin
 920      * with the arguments reversed in two ways: it assumes that its
 921      * arguments are correct, and it doesn't copy the magnitude array.
 922      */
 923     BigInteger(int[] magnitude, int signum) {
 924         this.signum = (magnitude.length==0 ? 0 : signum);
 925         this.mag = magnitude;
 926     }
 927 
 928     /**
 929      * This private constructor is for internal use and assumes that its
 930      * arguments are correct.
 931      */
 932     private BigInteger(byte[] magnitude, int signum) {
 933         this.signum = (magnitude.length==0 ? 0 : signum);
 934         this.mag = stripLeadingZeroBytes(magnitude);
 935     }
 936 
 937     //Static Factory Methods
 938 
 939     /**
 940      * Returns a BigInteger whose value is equal to that of the
 941      * specified {@code long}.  This "static factory method" is
 942      * provided in preference to a ({@code long}) constructor
 943      * because it allows for reuse of frequently used BigIntegers.
 944      *
 945      * @param  val value of the BigInteger to return.
 946      * @return a BigInteger with the specified value.
 947      */
 948     public static BigInteger valueOf(long val) {
 949         // If -MAX_CONSTANT < val < MAX_CONSTANT, return stashed constant
 950         if (val == 0)
 951             return ZERO;
 952         if (val > 0 && val <= MAX_CONSTANT)
 953             return posConst[(int) val];
 954         else if (val < 0 && val >= -MAX_CONSTANT)
 955             return negConst[(int) -val];
 956 
 957         return new BigInteger(val);
 958     }
 959 
 960     /**
 961      * Constructs a BigInteger with the specified value, which may not be zero.
 962      */
 963     private BigInteger(long val) {
 964         if (val < 0) {
 965             val = -val;
 966             signum = -1;
 967         } else {
 968             signum = 1;
 969         }
 970 
 971         int highWord = (int)(val >>> 32);
 972         if (highWord==0) {
 973             mag = new int[1];
 974             mag[0] = (int)val;
 975         } else {
 976             mag = new int[2];
 977             mag[0] = highWord;
 978             mag[1] = (int)val;
 979         }
 980     }
 981 
 982     /**
 983      * Returns a BigInteger with the given two's complement representation.
 984      * Assumes that the input array will not be modified (the returned
 985      * BigInteger will reference the input array if feasible).
 986      */
 987     private static BigInteger valueOf(int val[]) {
 988         return (val[0]>0 ? new BigInteger(val, 1) : new BigInteger(val));
 989     }
 990 
 991     // Constants
 992 
 993     /**
 994      * Initialize static constant array when class is loaded.
 995      */
 996     private final static int MAX_CONSTANT = 16;
 997     private static BigInteger posConst[] = new BigInteger[MAX_CONSTANT+1];
 998     private static BigInteger negConst[] = new BigInteger[MAX_CONSTANT+1];
 999     static {
1000         for (int i = 1; i <= MAX_CONSTANT; i++) {
1001             int[] magnitude = new int[1];
1002             magnitude[0] = i;
1003             posConst[i] = new BigInteger(magnitude,  1);
1004             negConst[i] = new BigInteger(magnitude, -1);
1005         }
1006     }
1007 
1008     /**
1009      * The BigInteger constant zero.
1010      *
1011      * @since   1.2
1012      */
1013     public static final BigInteger ZERO = new BigInteger(new int[0], 0);
1014 
1015     /**
1016      * The BigInteger constant one.
1017      *
1018      * @since   1.2
1019      */
1020     public static final BigInteger ONE = valueOf(1);
1021 
1022     /**
1023      * The BigInteger constant two.  (Not exported.)
1024      */
1025     private static final BigInteger TWO = valueOf(2);
1026 
1027     /**
1028      * The BigInteger constant ten.
1029      *
1030      * @since   1.5
1031      */
1032     public static final BigInteger TEN = valueOf(10);
1033 
1034     // Arithmetic Operations
1035 
1036     /**
1037      * Returns a BigInteger whose value is {@code (this + val)}.
1038      *
1039      * @param  val value to be added to this BigInteger.
1040      * @return {@code this + val}
1041      */
1042     public BigInteger add(BigInteger val) {
1043         if (val.signum == 0)
1044             return this;
1045         if (signum == 0)
1046             return val;
1047         if (val.signum == signum)
1048             return new BigInteger(add(mag, val.mag), signum);
1049 
1050         int cmp = compareMagnitude(val);
1051         if (cmp == 0)
1052             return ZERO;
1053         int[] resultMag = (cmp > 0 ? subtract(mag, val.mag)
1054                            : subtract(val.mag, mag));
1055         resultMag = trustedStripLeadingZeroInts(resultMag);
1056 
1057         return new BigInteger(resultMag, cmp == signum ? 1 : -1);
1058     }
1059 
1060     /**
1061      * Adds the contents of the int arrays x and y. This method allocates
1062      * a new int array to hold the answer and returns a reference to that
1063      * array.
1064      */
1065     private static int[] add(int[] x, int[] y) {
1066         // If x is shorter, swap the two arrays
1067         if (x.length < y.length) {
1068             int[] tmp = x;
1069             x = y;
1070             y = tmp;
1071         }
1072 
1073         int xIndex = x.length;
1074         int yIndex = y.length;
1075         int result[] = new int[xIndex];
1076         long sum = 0;
1077 
1078         // Add common parts of both numbers
1079         while(yIndex > 0) {
1080             sum = (x[--xIndex] & LONG_MASK) +
1081                   (y[--yIndex] & LONG_MASK) + (sum >>> 32);
1082             result[xIndex] = (int)sum;
1083         }
1084 
1085         // Copy remainder of longer number while carry propagation is required
1086         boolean carry = (sum >>> 32 != 0);
1087         while (xIndex > 0 && carry)
1088             carry = ((result[--xIndex] = x[xIndex] + 1) == 0);
1089 
1090         // Copy remainder of longer number
1091         while (xIndex > 0)
1092             result[--xIndex] = x[xIndex];
1093 
1094         // Grow result if necessary
1095         if (carry) {
1096             int bigger[] = new int[result.length + 1];
1097             System.arraycopy(result, 0, bigger, 1, result.length);
1098             bigger[0] = 0x01;
1099             return bigger;
1100         }
1101         return result;
1102     }
1103 
1104     /**
1105      * Returns a BigInteger whose value is {@code (this - val)}.
1106      *
1107      * @param  val value to be subtracted from this BigInteger.
1108      * @return {@code this - val}
1109      */
1110     public BigInteger subtract(BigInteger val) {
1111         if (val.signum == 0)
1112             return this;
1113         if (signum == 0)
1114             return val.negate();
1115         if (val.signum != signum)
1116             return new BigInteger(add(mag, val.mag), signum);
1117 
1118         int cmp = compareMagnitude(val);
1119         if (cmp == 0)
1120             return ZERO;
1121         int[] resultMag = (cmp > 0 ? subtract(mag, val.mag)
1122                            : subtract(val.mag, mag));
1123         resultMag = trustedStripLeadingZeroInts(resultMag);
1124         return new BigInteger(resultMag, cmp == signum ? 1 : -1);
1125     }
1126 
1127     /**
1128      * Subtracts the contents of the second int arrays (little) from the
1129      * first (big).  The first int array (big) must represent a larger number
1130      * than the second.  This method allocates the space necessary to hold the
1131      * answer.
1132      */
1133     private static int[] subtract(int[] big, int[] little) {
1134         int bigIndex = big.length;
1135         int result[] = new int[bigIndex];
1136         int littleIndex = little.length;
1137         long difference = 0;
1138 
1139         // Subtract common parts of both numbers
1140         while(littleIndex > 0) {
1141             difference = (big[--bigIndex] & LONG_MASK) -
1142                          (little[--littleIndex] & LONG_MASK) +
1143                          (difference >> 32);
1144             result[bigIndex] = (int)difference;
1145         }
1146 
1147         // Subtract remainder of longer number while borrow propagates
1148         boolean borrow = (difference >> 32 != 0);
1149         while (bigIndex > 0 && borrow)
1150             borrow = ((result[--bigIndex] = big[bigIndex] - 1) == -1);
1151 
1152         // Copy remainder of longer number
1153         while (bigIndex > 0)
1154             result[--bigIndex] = big[bigIndex];
1155 
1156         return result;
1157     }
1158 
1159     /**
1160      * Returns a BigInteger whose value is {@code (this * val)}.
1161      *
1162      * @param  val value to be multiplied by this BigInteger.
1163      * @return {@code this * val}
1164      */
1165     public BigInteger multiply(BigInteger val) {
1166         if (val.signum == 0 || signum == 0)
1167             return ZERO;
1168 
1169         int[] result = multiplyToLen(mag, mag.length,
1170                                      val.mag, val.mag.length, null);
1171         result = trustedStripLeadingZeroInts(result);
1172         return new BigInteger(result, signum == val.signum ? 1 : -1);
1173     }
1174 
1175     /**
1176      * Package private methods used by BigDecimal code to multiply a BigInteger
1177      * with a long. Assumes v is not equal to INFLATED.
1178      */
1179     BigInteger multiply(long v) {
1180         if (v == 0 || signum == 0)
1181           return ZERO;
1182         if (v == BigDecimal.INFLATED)
1183             return multiply(BigInteger.valueOf(v));
1184         int rsign = (v > 0 ? signum : -signum);
1185         if (v < 0)
1186             v = -v;
1187         long dh = v >>> 32;      // higher order bits
1188         long dl = v & LONG_MASK; // lower order bits
1189 
1190         int xlen = mag.length;
1191         int[] value = mag;
1192         int[] rmag = (dh == 0L) ? (new int[xlen + 1]) : (new int[xlen + 2]);
1193         long carry = 0;
1194         int rstart = rmag.length - 1;
1195         for (int i = xlen - 1; i >= 0; i--) {
1196             long product = (value[i] & LONG_MASK) * dl + carry;
1197             rmag[rstart--] = (int)product;
1198             carry = product >>> 32;
1199         }
1200         rmag[rstart] = (int)carry;
1201         if (dh != 0L) {
1202             carry = 0;
1203             rstart = rmag.length - 2;
1204             for (int i = xlen - 1; i >= 0; i--) {
1205                 long product = (value[i] & LONG_MASK) * dh +
1206                     (rmag[rstart] & LONG_MASK) + carry;
1207                 rmag[rstart--] = (int)product;
1208                 carry = product >>> 32;
1209             }
1210             rmag[0] = (int)carry;
1211         }
1212         if (carry == 0L)
1213             rmag = java.util.Arrays.copyOfRange(rmag, 1, rmag.length);
1214         return new BigInteger(rmag, rsign);
1215     }
1216 
1217     /**
1218      * Multiplies int arrays x and y to the specified lengths and places
1219      * the result into z. There will be no leading zeros in the resultant array.
1220      */
1221     private int[] multiplyToLen(int[] x, int xlen, int[] y, int ylen, int[] z) {
1222         int xstart = xlen - 1;
1223         int ystart = ylen - 1;
1224 
1225         if (z == null || z.length < (xlen+ ylen))
1226             z = new int[xlen+ylen];
1227 
1228         long carry = 0;
1229         for (int j=ystart, k=ystart+1+xstart; j>=0; j--, k--) {
1230             long product = (y[j] & LONG_MASK) *
1231                            (x[xstart] & LONG_MASK) + carry;
1232             z[k] = (int)product;
1233             carry = product >>> 32;
1234         }
1235         z[xstart] = (int)carry;
1236 
1237         for (int i = xstart-1; i >= 0; i--) {
1238             carry = 0;
1239             for (int j=ystart, k=ystart+1+i; j>=0; j--, k--) {
1240                 long product = (y[j] & LONG_MASK) *
1241                                (x[i] & LONG_MASK) +
1242                                (z[k] & LONG_MASK) + carry;
1243                 z[k] = (int)product;
1244                 carry = product >>> 32;
1245             }
1246             z[i] = (int)carry;
1247         }
1248         return z;
1249     }
1250 
1251     /**
1252      * Returns a BigInteger whose value is {@code (this<sup>2</sup>)}.
1253      *
1254      * @return {@code this<sup>2</sup>}
1255      */
1256     private BigInteger square() {
1257         if (signum == 0)
1258             return ZERO;
1259         int[] z = squareToLen(mag, mag.length, null);
1260         return new BigInteger(trustedStripLeadingZeroInts(z), 1);
1261     }
1262 
1263     /**
1264      * Squares the contents of the int array x. The result is placed into the
1265      * int array z.  The contents of x are not changed.
1266      */
1267     private static final int[] squareToLen(int[] x, int len, int[] z) {
1268         /*
1269          * The algorithm used here is adapted from Colin Plumb's C library.
1270          * Technique: Consider the partial products in the multiplication
1271          * of "abcde" by itself:
1272          *
1273          *               a  b  c  d  e
1274          *            *  a  b  c  d  e
1275          *          ==================
1276          *              ae be ce de ee
1277          *           ad bd cd dd de
1278          *        ac bc cc cd ce
1279          *     ab bb bc bd be
1280          *  aa ab ac ad ae
1281          *
1282          * Note that everything above the main diagonal:
1283          *              ae be ce de = (abcd) * e
1284          *           ad bd cd       = (abc) * d
1285          *        ac bc             = (ab) * c
1286          *     ab                   = (a) * b
1287          *
1288          * is a copy of everything below the main diagonal:
1289          *                       de
1290          *                 cd ce
1291          *           bc bd be
1292          *     ab ac ad ae
1293          *
1294          * Thus, the sum is 2 * (off the diagonal) + diagonal.
1295          *
1296          * This is accumulated beginning with the diagonal (which
1297          * consist of the squares of the digits of the input), which is then
1298          * divided by two, the off-diagonal added, and multiplied by two
1299          * again.  The low bit is simply a copy of the low bit of the
1300          * input, so it doesn't need special care.
1301          */
1302         int zlen = len << 1;
1303         if (z == null || z.length < zlen)
1304             z = new int[zlen];
1305 
1306         // Store the squares, right shifted one bit (i.e., divided by 2)
1307         int lastProductLowWord = 0;
1308         for (int j=0, i=0; j<len; j++) {
1309             long piece = (x[j] & LONG_MASK);
1310             long product = piece * piece;
1311             z[i++] = (lastProductLowWord << 31) | (int)(product >>> 33);
1312             z[i++] = (int)(product >>> 1);
1313             lastProductLowWord = (int)product;
1314         }
1315 
1316         // Add in off-diagonal sums
1317         for (int i=len, offset=1; i>0; i--, offset+=2) {
1318             int t = x[i-1];
1319             t = mulAdd(z, x, offset, i-1, t);
1320             addOne(z, offset-1, i, t);
1321         }
1322 
1323         // Shift back up and set low bit
1324         primitiveLeftShift(z, zlen, 1);
1325         z[zlen-1] |= x[len-1] & 1;
1326 
1327         return z;
1328     }
1329 
1330     /**
1331      * Returns a BigInteger whose value is {@code (this / val)}.
1332      *
1333      * @param  val value by which this BigInteger is to be divided.
1334      * @return {@code this / val}
1335      * @throws ArithmeticException if {@code val} is zero.
1336      */
1337     public BigInteger divide(BigInteger val) {
1338         MutableBigInteger q = new MutableBigInteger(),
1339                           a = new MutableBigInteger(this.mag),
1340                           b = new MutableBigInteger(val.mag);
1341 
1342         a.divide(b, q);
1343         return q.toBigInteger(this.signum == val.signum ? 1 : -1);
1344     }
1345 
1346     /**
1347      * Returns an array of two BigIntegers containing {@code (this / val)}
1348      * followed by {@code (this % val)}.
1349      *
1350      * @param  val value by which this BigInteger is to be divided, and the
1351      *         remainder computed.
1352      * @return an array of two BigIntegers: the quotient {@code (this / val)}
1353      *         is the initial element, and the remainder {@code (this % val)}
1354      *         is the final element.
1355      * @throws ArithmeticException if {@code val} is zero.
1356      */
1357     public BigInteger[] divideAndRemainder(BigInteger val) {
1358         BigInteger[] result = new BigInteger[2];
1359         MutableBigInteger q = new MutableBigInteger(),
1360                           a = new MutableBigInteger(this.mag),
1361                           b = new MutableBigInteger(val.mag);
1362         MutableBigInteger r = a.divide(b, q);
1363         result[0] = q.toBigInteger(this.signum == val.signum ? 1 : -1);
1364         result[1] = r.toBigInteger(this.signum);
1365         return result;
1366     }
1367 
1368     /**
1369      * Returns a BigInteger whose value is {@code (this % val)}.
1370      *
1371      * @param  val value by which this BigInteger is to be divided, and the
1372      *         remainder computed.
1373      * @return {@code this % val}
1374      * @throws ArithmeticException if {@code val} is zero.
1375      */
1376     public BigInteger remainder(BigInteger val) {
1377         MutableBigInteger q = new MutableBigInteger(),
1378                           a = new MutableBigInteger(this.mag),
1379                           b = new MutableBigInteger(val.mag);
1380 
1381         return a.divide(b, q).toBigInteger(this.signum);
1382     }
1383 
1384     /**
1385      * Returns a BigInteger whose value is <tt>(this<sup>exponent</sup>)</tt>.
1386      * Note that {@code exponent} is an integer rather than a BigInteger.
1387      *
1388      * @param  exponent exponent to which this BigInteger is to be raised.
1389      * @return <tt>this<sup>exponent</sup></tt>
1390      * @throws ArithmeticException {@code exponent} is negative.  (This would
1391      *         cause the operation to yield a non-integer value.)
1392      */
1393     public BigInteger pow(int exponent) {
1394         if (exponent < 0)
1395             throw new ArithmeticException("Negative exponent");
1396         if (signum==0)
1397             return (exponent==0 ? ONE : this);
1398 
1399         // Perform exponentiation using repeated squaring trick
1400         int newSign = (signum<0 && (exponent&1)==1 ? -1 : 1);
1401         int[] baseToPow2 = this.mag;
1402         int[] result = {1};
1403 
1404         while (exponent != 0) {
1405             if ((exponent & 1)==1) {
1406                 result = multiplyToLen(result, result.length,
1407                                        baseToPow2, baseToPow2.length, null);
1408                 result = trustedStripLeadingZeroInts(result);
1409             }
1410             if ((exponent >>>= 1) != 0) {
1411                 baseToPow2 = squareToLen(baseToPow2, baseToPow2.length, null);
1412                 baseToPow2 = trustedStripLeadingZeroInts(baseToPow2);
1413             }
1414         }
1415         return new BigInteger(result, newSign);
1416     }
1417 
1418     /**
1419      * Returns a BigInteger whose value is the greatest common divisor of
1420      * {@code abs(this)} and {@code abs(val)}.  Returns 0 if
1421      * {@code this==0 && val==0}.
1422      *
1423      * @param  val value with which the GCD is to be computed.
1424      * @return {@code GCD(abs(this), abs(val))}
1425      */
1426     public BigInteger gcd(BigInteger val) {
1427         if (val.signum == 0)
1428             return this.abs();
1429         else if (this.signum == 0)
1430             return val.abs();
1431 
1432         MutableBigInteger a = new MutableBigInteger(this);
1433         MutableBigInteger b = new MutableBigInteger(val);
1434 
1435         MutableBigInteger result = a.hybridGCD(b);
1436 
1437         return result.toBigInteger(1);
1438     }
1439 
1440     /**
1441      * Package private method to return bit length for an integer.
1442      */
1443     static int bitLengthForInt(int n) {
1444         return 32 - Integer.numberOfLeadingZeros(n);
1445     }
1446 
1447     /**
1448      * Left shift int array a up to len by n bits. Returns the array that
1449      * results from the shift since space may have to be reallocated.
1450      */
1451     private static int[] leftShift(int[] a, int len, int n) {
1452         int nInts = n >>> 5;
1453         int nBits = n&0x1F;
1454         int bitsInHighWord = bitLengthForInt(a[0]);
1455 
1456         // If shift can be done without recopy, do so
1457         if (n <= (32-bitsInHighWord)) {
1458             primitiveLeftShift(a, len, nBits);
1459             return a;
1460         } else { // Array must be resized
1461             if (nBits <= (32-bitsInHighWord)) {
1462                 int result[] = new int[nInts+len];
1463                 for (int i=0; i<len; i++)
1464                     result[i] = a[i];
1465                 primitiveLeftShift(result, result.length, nBits);
1466                 return result;
1467             } else {
1468                 int result[] = new int[nInts+len+1];
1469                 for (int i=0; i<len; i++)
1470                     result[i] = a[i];
1471                 primitiveRightShift(result, result.length, 32 - nBits);
1472                 return result;
1473             }
1474         }
1475     }
1476 
1477     // shifts a up to len right n bits assumes no leading zeros, 0<n<32
1478     static void primitiveRightShift(int[] a, int len, int n) {
1479         int n2 = 32 - n;
1480         for (int i=len-1, c=a[i]; i>0; i--) {
1481             int b = c;
1482             c = a[i-1];
1483             a[i] = (c << n2) | (b >>> n);
1484         }
1485         a[0] >>>= n;
1486     }
1487 
1488     // shifts a up to len left n bits assumes no leading zeros, 0<=n<32
1489     static void primitiveLeftShift(int[] a, int len, int n) {
1490         if (len == 0 || n == 0)
1491             return;
1492 
1493         int n2 = 32 - n;
1494         for (int i=0, c=a[i], m=i+len-1; i<m; i++) {
1495             int b = c;
1496             c = a[i+1];
1497             a[i] = (b << n) | (c >>> n2);
1498         }
1499         a[len-1] <<= n;
1500     }
1501 
1502     /**
1503      * Calculate bitlength of contents of the first len elements an int array,
1504      * assuming there are no leading zero ints.
1505      */
1506     private static int bitLength(int[] val, int len) {
1507         if (len == 0)
1508             return 0;
1509         return ((len - 1) << 5) + bitLengthForInt(val[0]);
1510     }
1511 
1512     /**
1513      * Returns a BigInteger whose value is the absolute value of this
1514      * BigInteger.
1515      *
1516      * @return {@code abs(this)}
1517      */
1518     public BigInteger abs() {
1519         return (signum >= 0 ? this : this.negate());
1520     }
1521 
1522     /**
1523      * Returns a BigInteger whose value is {@code (-this)}.
1524      *
1525      * @return {@code -this}
1526      */
1527     public BigInteger negate() {
1528         return new BigInteger(this.mag, -this.signum);
1529     }
1530 
1531     /**
1532      * Returns the signum function of this BigInteger.
1533      *
1534      * @return -1, 0 or 1 as the value of this BigInteger is negative, zero or
1535      *         positive.
1536      */
1537     public int signum() {
1538         return this.signum;
1539     }
1540 
1541     // Modular Arithmetic Operations
1542 
1543     /**
1544      * Returns a BigInteger whose value is {@code (this mod m}).  This method
1545      * differs from {@code remainder} in that it always returns a
1546      * <i>non-negative</i> BigInteger.
1547      *
1548      * @param  m the modulus.
1549      * @return {@code this mod m}
1550      * @throws ArithmeticException {@code m} &le; 0
1551      * @see    #remainder
1552      */
1553     public BigInteger mod(BigInteger m) {
1554         if (m.signum <= 0)
1555             throw new ArithmeticException("BigInteger: modulus not positive");
1556 
1557         BigInteger result = this.remainder(m);
1558         return (result.signum >= 0 ? result : result.add(m));
1559     }
1560 
1561     /**
1562      * Returns a BigInteger whose value is
1563      * <tt>(this<sup>exponent</sup> mod m)</tt>.  (Unlike {@code pow}, this
1564      * method permits negative exponents.)
1565      *
1566      * @param  exponent the exponent.
1567      * @param  m the modulus.
1568      * @return <tt>this<sup>exponent</sup> mod m</tt>
1569      * @throws ArithmeticException {@code m} &le; 0 or the exponent is
1570      *         negative and this BigInteger is not <i>relatively
1571      *         prime</i> to {@code m}.
1572      * @see    #modInverse
1573      */
1574     public BigInteger modPow(BigInteger exponent, BigInteger m) {
1575         if (m.signum <= 0)
1576             throw new ArithmeticException("BigInteger: modulus not positive");
1577 
1578         // Trivial cases
1579         if (exponent.signum == 0)
1580             return (m.equals(ONE) ? ZERO : ONE);
1581 
1582         if (this.equals(ONE))
1583             return (m.equals(ONE) ? ZERO : ONE);
1584 
1585         if (this.equals(ZERO) && exponent.signum >= 0)
1586             return ZERO;
1587 
1588         if (this.equals(negConst[1]) && (!exponent.testBit(0)))
1589             return (m.equals(ONE) ? ZERO : ONE);
1590 
1591         boolean invertResult;
1592         if ((invertResult = (exponent.signum < 0)))
1593             exponent = exponent.negate();
1594 
1595         BigInteger base = (this.signum < 0 || this.compareTo(m) >= 0
1596                            ? this.mod(m) : this);
1597         BigInteger result;
1598         if (m.testBit(0)) { // odd modulus
1599             result = base.oddModPow(exponent, m);
1600         } else {
1601             /*
1602              * Even modulus.  Tear it into an "odd part" (m1) and power of two
1603              * (m2), exponentiate mod m1, manually exponentiate mod m2, and
1604              * use Chinese Remainder Theorem to combine results.
1605              */
1606 
1607             // Tear m apart into odd part (m1) and power of 2 (m2)
1608             int p = m.getLowestSetBit();   // Max pow of 2 that divides m
1609 
1610             BigInteger m1 = m.shiftRight(p);  // m/2**p
1611             BigInteger m2 = ONE.shiftLeft(p); // 2**p
1612 
1613             // Calculate new base from m1
1614             BigInteger base2 = (this.signum < 0 || this.compareTo(m1) >= 0
1615                                 ? this.mod(m1) : this);
1616 
1617             // Caculate (base ** exponent) mod m1.
1618             BigInteger a1 = (m1.equals(ONE) ? ZERO :
1619                              base2.oddModPow(exponent, m1));
1620 
1621             // Calculate (this ** exponent) mod m2
1622             BigInteger a2 = base.modPow2(exponent, p);
1623 
1624             // Combine results using Chinese Remainder Theorem
1625             BigInteger y1 = m2.modInverse(m1);
1626             BigInteger y2 = m1.modInverse(m2);
1627 
1628             result = a1.multiply(m2).multiply(y1).add
1629                      (a2.multiply(m1).multiply(y2)).mod(m);
1630         }
1631 
1632         return (invertResult ? result.modInverse(m) : result);
1633     }
1634 
1635     static int[] bnExpModThreshTable = {7, 25, 81, 241, 673, 1793,
1636                                                 Integer.MAX_VALUE}; // Sentinel
1637 
1638     /**
1639      * Returns a BigInteger whose value is x to the power of y mod z.
1640      * Assumes: z is odd && x < z.
1641      */
1642     private BigInteger oddModPow(BigInteger y, BigInteger z) {
1643     /*
1644      * The algorithm is adapted from Colin Plumb's C library.
1645      *
1646      * The window algorithm:
1647      * The idea is to keep a running product of b1 = n^(high-order bits of exp)
1648      * and then keep appending exponent bits to it.  The following patterns
1649      * apply to a 3-bit window (k = 3):
1650      * To append   0: square
1651      * To append   1: square, multiply by n^1
1652      * To append  10: square, multiply by n^1, square
1653      * To append  11: square, square, multiply by n^3
1654      * To append 100: square, multiply by n^1, square, square
1655      * To append 101: square, square, square, multiply by n^5
1656      * To append 110: square, square, multiply by n^3, square
1657      * To append 111: square, square, square, multiply by n^7
1658      *
1659      * Since each pattern involves only one multiply, the longer the pattern
1660      * the better, except that a 0 (no multiplies) can be appended directly.
1661      * We precompute a table of odd powers of n, up to 2^k, and can then
1662      * multiply k bits of exponent at a time.  Actually, assuming random
1663      * exponents, there is on average one zero bit between needs to
1664      * multiply (1/2 of the time there's none, 1/4 of the time there's 1,
1665      * 1/8 of the time, there's 2, 1/32 of the time, there's 3, etc.), so
1666      * you have to do one multiply per k+1 bits of exponent.
1667      *
1668      * The loop walks down the exponent, squaring the result buffer as
1669      * it goes.  There is a wbits+1 bit lookahead buffer, buf, that is
1670      * filled with the upcoming exponent bits.  (What is read after the
1671      * end of the exponent is unimportant, but it is filled with zero here.)
1672      * When the most-significant bit of this buffer becomes set, i.e.
1673      * (buf & tblmask) != 0, we have to decide what pattern to multiply
1674      * by, and when to do it.  We decide, remember to do it in future
1675      * after a suitable number of squarings have passed (e.g. a pattern
1676      * of "100" in the buffer requires that we multiply by n^1 immediately;
1677      * a pattern of "110" calls for multiplying by n^3 after one more
1678      * squaring), clear the buffer, and continue.
1679      *
1680      * When we start, there is one more optimization: the result buffer
1681      * is implcitly one, so squaring it or multiplying by it can be
1682      * optimized away.  Further, if we start with a pattern like "100"
1683      * in the lookahead window, rather than placing n into the buffer
1684      * and then starting to square it, we have already computed n^2
1685      * to compute the odd-powers table, so we can place that into
1686      * the buffer and save a squaring.
1687      *
1688      * This means that if you have a k-bit window, to compute n^z,
1689      * where z is the high k bits of the exponent, 1/2 of the time
1690      * it requires no squarings.  1/4 of the time, it requires 1
1691      * squaring, ... 1/2^(k-1) of the time, it reqires k-2 squarings.
1692      * And the remaining 1/2^(k-1) of the time, the top k bits are a
1693      * 1 followed by k-1 0 bits, so it again only requires k-2
1694      * squarings, not k-1.  The average of these is 1.  Add that
1695      * to the one squaring we have to do to compute the table,
1696      * and you'll see that a k-bit window saves k-2 squarings
1697      * as well as reducing the multiplies.  (It actually doesn't
1698      * hurt in the case k = 1, either.)
1699      */
1700         // Special case for exponent of one
1701         if (y.equals(ONE))
1702             return this;
1703 
1704         // Special case for base of zero
1705         if (signum==0)
1706             return ZERO;
1707 
1708         int[] base = mag.clone();
1709         int[] exp = y.mag;
1710         int[] mod = z.mag;
1711         int modLen = mod.length;
1712 
1713         // Select an appropriate window size
1714         int wbits = 0;
1715         int ebits = bitLength(exp, exp.length);
1716         // if exponent is 65537 (0x10001), use minimum window size
1717         if ((ebits != 17) || (exp[0] != 65537)) {
1718             while (ebits > bnExpModThreshTable[wbits]) {
1719                 wbits++;
1720             }
1721         }
1722 
1723         // Calculate appropriate table size
1724         int tblmask = 1 << wbits;
1725 
1726         // Allocate table for precomputed odd powers of base in Montgomery form
1727         int[][] table = new int[tblmask][];
1728         for (int i=0; i<tblmask; i++)
1729             table[i] = new int[modLen];
1730 
1731         // Compute the modular inverse
1732         int inv = -MutableBigInteger.inverseMod32(mod[modLen-1]);
1733 
1734         // Convert base to Montgomery form
1735         int[] a = leftShift(base, base.length, modLen << 5);
1736 
1737         MutableBigInteger q = new MutableBigInteger(),
1738                           a2 = new MutableBigInteger(a),
1739                           b2 = new MutableBigInteger(mod);
1740 
1741         MutableBigInteger r= a2.divide(b2, q);
1742         table[0] = r.toIntArray();
1743 
1744         // Pad table[0] with leading zeros so its length is at least modLen
1745         if (table[0].length < modLen) {
1746            int offset = modLen - table[0].length;
1747            int[] t2 = new int[modLen];
1748            for (int i=0; i<table[0].length; i++)
1749                t2[i+offset] = table[0][i];
1750            table[0] = t2;
1751         }
1752 
1753         // Set b to the square of the base
1754         int[] b = squareToLen(table[0], modLen, null);
1755         b = montReduce(b, mod, modLen, inv);
1756 
1757         // Set t to high half of b
1758         int[] t = new int[modLen];
1759         for(int i=0; i<modLen; i++)
1760             t[i] = b[i];
1761 
1762         // Fill in the table with odd powers of the base
1763         for (int i=1; i<tblmask; i++) {
1764             int[] prod = multiplyToLen(t, modLen, table[i-1], modLen, null);
1765             table[i] = montReduce(prod, mod, modLen, inv);
1766         }
1767 
1768         // Pre load the window that slides over the exponent
1769         int bitpos = 1 << ((ebits-1) & (32-1));
1770 
1771         int buf = 0;
1772         int elen = exp.length;
1773         int eIndex = 0;
1774         for (int i = 0; i <= wbits; i++) {
1775             buf = (buf << 1) | (((exp[eIndex] & bitpos) != 0)?1:0);
1776             bitpos >>>= 1;
1777             if (bitpos == 0) {
1778                 eIndex++;
1779                 bitpos = 1 << (32-1);
1780                 elen--;
1781             }
1782         }
1783 
1784         int multpos = ebits;
1785 
1786         // The first iteration, which is hoisted out of the main loop
1787         ebits--;
1788         boolean isone = true;
1789 
1790         multpos = ebits - wbits;
1791         while ((buf & 1) == 0) {
1792             buf >>>= 1;
1793             multpos++;
1794         }
1795 
1796         int[] mult = table[buf >>> 1];
1797 
1798         buf = 0;
1799         if (multpos == ebits)
1800             isone = false;
1801 
1802         // The main loop
1803         while(true) {
1804             ebits--;
1805             // Advance the window
1806             buf <<= 1;
1807 
1808             if (elen != 0) {
1809                 buf |= ((exp[eIndex] & bitpos) != 0) ? 1 : 0;
1810                 bitpos >>>= 1;
1811                 if (bitpos == 0) {
1812                     eIndex++;
1813                     bitpos = 1 << (32-1);
1814                     elen--;
1815                 }
1816             }
1817 
1818             // Examine the window for pending multiplies
1819             if ((buf & tblmask) != 0) {
1820                 multpos = ebits - wbits;
1821                 while ((buf & 1) == 0) {
1822                     buf >>>= 1;
1823                     multpos++;
1824                 }
1825                 mult = table[buf >>> 1];
1826                 buf = 0;
1827             }
1828 
1829             // Perform multiply
1830             if (ebits == multpos) {
1831                 if (isone) {
1832                     b = mult.clone();
1833                     isone = false;
1834                 } else {
1835                     t = b;
1836                     a = multiplyToLen(t, modLen, mult, modLen, a);
1837                     a = montReduce(a, mod, modLen, inv);
1838                     t = a; a = b; b = t;
1839                 }
1840             }
1841 
1842             // Check if done
1843             if (ebits == 0)
1844                 break;
1845 
1846             // Square the input
1847             if (!isone) {
1848                 t = b;
1849                 a = squareToLen(t, modLen, a);
1850                 a = montReduce(a, mod, modLen, inv);
1851                 t = a; a = b; b = t;
1852             }
1853         }
1854 
1855         // Convert result out of Montgomery form and return
1856         int[] t2 = new int[2*modLen];
1857         for(int i=0; i<modLen; i++)
1858             t2[i+modLen] = b[i];
1859 
1860         b = montReduce(t2, mod, modLen, inv);
1861 
1862         t2 = new int[modLen];
1863         for(int i=0; i<modLen; i++)
1864             t2[i] = b[i];
1865 
1866         return new BigInteger(1, t2);
1867     }
1868 
1869     /**
1870      * Montgomery reduce n, modulo mod.  This reduces modulo mod and divides
1871      * by 2^(32*mlen). Adapted from Colin Plumb's C library.
1872      */
1873     private static int[] montReduce(int[] n, int[] mod, int mlen, int inv) {
1874         int c=0;
1875         int len = mlen;
1876         int offset=0;
1877 
1878         do {
1879             int nEnd = n[n.length-1-offset];
1880             int carry = mulAdd(n, mod, offset, mlen, inv * nEnd);
1881             c += addOne(n, offset, mlen, carry);
1882             offset++;
1883         } while(--len > 0);
1884 
1885         while(c>0)
1886             c += subN(n, mod, mlen);
1887 
1888         while (intArrayCmpToLen(n, mod, mlen) >= 0)
1889             subN(n, mod, mlen);
1890 
1891         return n;
1892     }
1893 
1894 
1895     /*
1896      * Returns -1, 0 or +1 as big-endian unsigned int array arg1 is less than,
1897      * equal to, or greater than arg2 up to length len.
1898      */
1899     private static int intArrayCmpToLen(int[] arg1, int[] arg2, int len) {
1900         for (int i=0; i<len; i++) {
1901             long b1 = arg1[i] & LONG_MASK;
1902             long b2 = arg2[i] & LONG_MASK;
1903             if (b1 < b2)
1904                 return -1;
1905             if (b1 > b2)
1906                 return 1;
1907         }
1908         return 0;
1909     }
1910 
1911     /**
1912      * Subtracts two numbers of same length, returning borrow.
1913      */
1914     private static int subN(int[] a, int[] b, int len) {
1915         long sum = 0;
1916 
1917         while(--len >= 0) {
1918             sum = (a[len] & LONG_MASK) -
1919                  (b[len] & LONG_MASK) + (sum >> 32);
1920             a[len] = (int)sum;
1921         }
1922 
1923         return (int)(sum >> 32);
1924     }
1925 
1926     /**
1927      * Multiply an array by one word k and add to result, return the carry
1928      */
1929     static int mulAdd(int[] out, int[] in, int offset, int len, int k) {
1930         long kLong = k & LONG_MASK;
1931         long carry = 0;
1932 
1933         offset = out.length-offset - 1;
1934         for (int j=len-1; j >= 0; j--) {
1935             long product = (in[j] & LONG_MASK) * kLong +
1936                            (out[offset] & LONG_MASK) + carry;
1937             out[offset--] = (int)product;
1938             carry = product >>> 32;
1939         }
1940         return (int)carry;
1941     }
1942 
1943     /**
1944      * Add one word to the number a mlen words into a. Return the resulting
1945      * carry.
1946      */
1947     static int addOne(int[] a, int offset, int mlen, int carry) {
1948         offset = a.length-1-mlen-offset;
1949         long t = (a[offset] & LONG_MASK) + (carry & LONG_MASK);
1950 
1951         a[offset] = (int)t;
1952         if ((t >>> 32) == 0)
1953             return 0;
1954         while (--mlen >= 0) {
1955             if (--offset < 0) { // Carry out of number
1956                 return 1;
1957             } else {
1958                 a[offset]++;
1959                 if (a[offset] != 0)
1960                     return 0;
1961             }
1962         }
1963         return 1;
1964     }
1965 
1966     /**
1967      * Returns a BigInteger whose value is (this ** exponent) mod (2**p)
1968      */
1969     private BigInteger modPow2(BigInteger exponent, int p) {
1970         /*
1971          * Perform exponentiation using repeated squaring trick, chopping off
1972          * high order bits as indicated by modulus.
1973          */
1974         BigInteger result = valueOf(1);
1975         BigInteger baseToPow2 = this.mod2(p);
1976         int expOffset = 0;
1977 
1978         int limit = exponent.bitLength();
1979 
1980         if (this.testBit(0))
1981            limit = (p-1) < limit ? (p-1) : limit;
1982 
1983         while (expOffset < limit) {
1984             if (exponent.testBit(expOffset))
1985                 result = result.multiply(baseToPow2).mod2(p);
1986             expOffset++;
1987             if (expOffset < limit)
1988                 baseToPow2 = baseToPow2.square().mod2(p);
1989         }
1990 
1991         return result;
1992     }
1993 
1994     /**
1995      * Returns a BigInteger whose value is this mod(2**p).
1996      * Assumes that this {@code BigInteger >= 0} and {@code p > 0}.
1997      */
1998     private BigInteger mod2(int p) {
1999         if (bitLength() <= p)
2000             return this;
2001 
2002         // Copy remaining ints of mag
2003         int numInts = (p + 31) >>> 5;
2004         int[] mag = new int[numInts];
2005         for (int i=0; i<numInts; i++)
2006             mag[i] = this.mag[i + (this.mag.length - numInts)];
2007 
2008         // Mask out any excess bits
2009         int excessBits = (numInts << 5) - p;
2010         mag[0] &= (1L << (32-excessBits)) - 1;
2011 
2012         return (mag[0]==0 ? new BigInteger(1, mag) : new BigInteger(mag, 1));
2013     }
2014 
2015     /**
2016      * Returns a BigInteger whose value is {@code (this}<sup>-1</sup> {@code mod m)}.
2017      *
2018      * @param  m the modulus.
2019      * @return {@code this}<sup>-1</sup> {@code mod m}.
2020      * @throws ArithmeticException {@code  m} &le; 0, or this BigInteger
2021      *         has no multiplicative inverse mod m (that is, this BigInteger
2022      *         is not <i>relatively prime</i> to m).
2023      */
2024     public BigInteger modInverse(BigInteger m) {
2025         if (m.signum != 1)
2026             throw new ArithmeticException("BigInteger: modulus not positive");
2027 
2028         if (m.equals(ONE))
2029             return ZERO;
2030 
2031         // Calculate (this mod m)
2032         BigInteger modVal = this;
2033         if (signum < 0 || (this.compareMagnitude(m) >= 0))
2034             modVal = this.mod(m);
2035 
2036         if (modVal.equals(ONE))
2037             return ONE;
2038 
2039         MutableBigInteger a = new MutableBigInteger(modVal);
2040         MutableBigInteger b = new MutableBigInteger(m);
2041 
2042         MutableBigInteger result = a.mutableModInverse(b);
2043         return result.toBigInteger(1);
2044     }
2045 
2046     // Shift Operations
2047 
2048     /**
2049      * Returns a BigInteger whose value is {@code (this << n)}.
2050      * The shift distance, {@code n}, may be negative, in which case
2051      * this method performs a right shift.
2052      * (Computes <tt>floor(this * 2<sup>n</sup>)</tt>.)
2053      *
2054      * @param  n shift distance, in bits.
2055      * @return {@code this << n}
2056      * @throws ArithmeticException if the shift distance is {@code
2057      *         Integer.MIN_VALUE}.
2058      * @see #shiftRight
2059      */
2060     public BigInteger shiftLeft(int n) {
2061         if (signum == 0)
2062             return ZERO;
2063         if (n==0)
2064             return this;
2065         if (n<0) {
2066             if (n == Integer.MIN_VALUE) {
2067                 throw new ArithmeticException("Shift distance of Integer.MIN_VALUE not supported.");
2068             } else {
2069                 return shiftRight(-n);
2070             }
2071         }
2072 
2073         int nInts = n >>> 5;
2074         int nBits = n & 0x1f;
2075         int magLen = mag.length;
2076         int newMag[] = null;
2077 
2078         if (nBits == 0) {
2079             newMag = new int[magLen + nInts];
2080             for (int i=0; i<magLen; i++)
2081                 newMag[i] = mag[i];
2082         } else {
2083             int i = 0;
2084             int nBits2 = 32 - nBits;
2085             int highBits = mag[0] >>> nBits2;
2086             if (highBits != 0) {
2087                 newMag = new int[magLen + nInts + 1];
2088                 newMag[i++] = highBits;
2089             } else {
2090                 newMag = new int[magLen + nInts];
2091             }
2092             int j=0;
2093             while (j < magLen-1)
2094                 newMag[i++] = mag[j++] << nBits | mag[j] >>> nBits2;
2095             newMag[i] = mag[j] << nBits;
2096         }
2097 
2098         return new BigInteger(newMag, signum);
2099     }
2100 
2101     /**
2102      * Returns a BigInteger whose value is {@code (this >> n)}.  Sign
2103      * extension is performed.  The shift distance, {@code n}, may be
2104      * negative, in which case this method performs a left shift.
2105      * (Computes <tt>floor(this / 2<sup>n</sup>)</tt>.)
2106      *
2107      * @param  n shift distance, in bits.
2108      * @return {@code this >> n}
2109      * @throws ArithmeticException if the shift distance is {@code
2110      *         Integer.MIN_VALUE}.
2111      * @see #shiftLeft
2112      */
2113     public BigInteger shiftRight(int n) {
2114         if (n==0)
2115             return this;
2116         if (n<0) {
2117             if (n == Integer.MIN_VALUE) {
2118                 throw new ArithmeticException("Shift distance of Integer.MIN_VALUE not supported.");
2119             } else {
2120                 return shiftLeft(-n);
2121             }
2122         }
2123 
2124         int nInts = n >>> 5;
2125         int nBits = n & 0x1f;
2126         int magLen = mag.length;
2127         int newMag[] = null;
2128 
2129         // Special case: entire contents shifted off the end
2130         if (nInts >= magLen)
2131             return (signum >= 0 ? ZERO : negConst[1]);
2132 
2133         if (nBits == 0) {
2134             int newMagLen = magLen - nInts;
2135             newMag = new int[newMagLen];
2136             for (int i=0; i<newMagLen; i++)
2137                 newMag[i] = mag[i];
2138         } else {
2139             int i = 0;
2140             int highBits = mag[0] >>> nBits;
2141             if (highBits != 0) {
2142                 newMag = new int[magLen - nInts];
2143                 newMag[i++] = highBits;
2144             } else {
2145                 newMag = new int[magLen - nInts -1];
2146             }
2147 
2148             int nBits2 = 32 - nBits;
2149             int j=0;
2150             while (j < magLen - nInts - 1)
2151                 newMag[i++] = (mag[j++] << nBits2) | (mag[j] >>> nBits);
2152         }
2153 
2154         if (signum < 0) {
2155             // Find out whether any one-bits were shifted off the end.
2156             boolean onesLost = false;
2157             for (int i=magLen-1, j=magLen-nInts; i>=j && !onesLost; i--)
2158                 onesLost = (mag[i] != 0);
2159             if (!onesLost && nBits != 0)
2160                 onesLost = (mag[magLen - nInts - 1] << (32 - nBits) != 0);
2161 
2162             if (onesLost)
2163                 newMag = javaIncrement(newMag);
2164         }
2165 
2166         return new BigInteger(newMag, signum);
2167     }
2168 
2169     int[] javaIncrement(int[] val) {
2170         int lastSum = 0;
2171         for (int i=val.length-1;  i >= 0 && lastSum == 0; i--)
2172             lastSum = (val[i] += 1);
2173         if (lastSum == 0) {
2174             val = new int[val.length+1];
2175             val[0] = 1;
2176         }
2177         return val;
2178     }
2179 
2180     // Bitwise Operations
2181 
2182     /**
2183      * Returns a BigInteger whose value is {@code (this & val)}.  (This
2184      * method returns a negative BigInteger if and only if this and val are
2185      * both negative.)
2186      *
2187      * @param val value to be AND'ed with this BigInteger.
2188      * @return {@code this & val}
2189      */
2190     public BigInteger and(BigInteger val) {
2191         int[] result = new int[Math.max(intLength(), val.intLength())];
2192         for (int i=0; i<result.length; i++)
2193             result[i] = (getInt(result.length-i-1)
2194                          & val.getInt(result.length-i-1));
2195 
2196         return valueOf(result);
2197     }
2198 
2199     /**
2200      * Returns a BigInteger whose value is {@code (this | val)}.  (This method
2201      * returns a negative BigInteger if and only if either this or val is
2202      * negative.)
2203      *
2204      * @param val value to be OR'ed with this BigInteger.
2205      * @return {@code this | val}
2206      */
2207     public BigInteger or(BigInteger val) {
2208         int[] result = new int[Math.max(intLength(), val.intLength())];
2209         for (int i=0; i<result.length; i++)
2210             result[i] = (getInt(result.length-i-1)
2211                          | val.getInt(result.length-i-1));
2212 
2213         return valueOf(result);
2214     }
2215 
2216     /**
2217      * Returns a BigInteger whose value is {@code (this ^ val)}.  (This method
2218      * returns a negative BigInteger if and only if exactly one of this and
2219      * val are negative.)
2220      *
2221      * @param val value to be XOR'ed with this BigInteger.
2222      * @return {@code this ^ val}
2223      */
2224     public BigInteger xor(BigInteger val) {
2225         int[] result = new int[Math.max(intLength(), val.intLength())];
2226         for (int i=0; i<result.length; i++)
2227             result[i] = (getInt(result.length-i-1)
2228                          ^ val.getInt(result.length-i-1));
2229 
2230         return valueOf(result);
2231     }
2232 
2233     /**
2234      * Returns a BigInteger whose value is {@code (~this)}.  (This method
2235      * returns a negative value if and only if this BigInteger is
2236      * non-negative.)
2237      *
2238      * @return {@code ~this}
2239      */
2240     public BigInteger not() {
2241         int[] result = new int[intLength()];
2242         for (int i=0; i<result.length; i++)
2243             result[i] = ~getInt(result.length-i-1);
2244 
2245         return valueOf(result);
2246     }
2247 
2248     /**
2249      * Returns a BigInteger whose value is {@code (this & ~val)}.  This
2250      * method, which is equivalent to {@code and(val.not())}, is provided as
2251      * a convenience for masking operations.  (This method returns a negative
2252      * BigInteger if and only if {@code this} is negative and {@code val} is
2253      * positive.)
2254      *
2255      * @param val value to be complemented and AND'ed with this BigInteger.
2256      * @return {@code this & ~val}
2257      */
2258     public BigInteger andNot(BigInteger val) {
2259         int[] result = new int[Math.max(intLength(), val.intLength())];
2260         for (int i=0; i<result.length; i++)
2261             result[i] = (getInt(result.length-i-1)
2262                          & ~val.getInt(result.length-i-1));
2263 
2264         return valueOf(result);
2265     }
2266 
2267 
2268     // Single Bit Operations
2269 
2270     /**
2271      * Returns {@code true} if and only if the designated bit is set.
2272      * (Computes {@code ((this & (1<<n)) != 0)}.)
2273      *
2274      * @param  n index of bit to test.
2275      * @return {@code true} if and only if the designated bit is set.
2276      * @throws ArithmeticException {@code n} is negative.
2277      */
2278     public boolean testBit(int n) {
2279         if (n<0)
2280             throw new ArithmeticException("Negative bit address");
2281 
2282         return (getInt(n >>> 5) & (1 << (n & 31))) != 0;
2283     }
2284 
2285     /**
2286      * Returns a BigInteger whose value is equivalent to this BigInteger
2287      * with the designated bit set.  (Computes {@code (this | (1<<n))}.)
2288      *
2289      * @param  n index of bit to set.
2290      * @return {@code this | (1<<n)}
2291      * @throws ArithmeticException {@code n} is negative.
2292      */
2293     public BigInteger setBit(int n) {
2294         if (n<0)
2295             throw new ArithmeticException("Negative bit address");
2296 
2297         int intNum = n >>> 5;
2298         int[] result = new int[Math.max(intLength(), intNum+2)];
2299 
2300         for (int i=0; i<result.length; i++)
2301             result[result.length-i-1] = getInt(i);
2302 
2303         result[result.length-intNum-1] |= (1 << (n & 31));
2304 
2305         return valueOf(result);
2306     }
2307 
2308     /**
2309      * Returns a BigInteger whose value is equivalent to this BigInteger
2310      * with the designated bit cleared.
2311      * (Computes {@code (this & ~(1<<n))}.)
2312      *
2313      * @param  n index of bit to clear.
2314      * @return {@code this & ~(1<<n)}
2315      * @throws ArithmeticException {@code n} is negative.
2316      */
2317     public BigInteger clearBit(int n) {
2318         if (n<0)
2319             throw new ArithmeticException("Negative bit address");
2320 
2321         int intNum = n >>> 5;
2322         int[] result = new int[Math.max(intLength(), ((n + 1) >>> 5) + 1)];
2323 
2324         for (int i=0; i<result.length; i++)
2325             result[result.length-i-1] = getInt(i);
2326 
2327         result[result.length-intNum-1] &= ~(1 << (n & 31));
2328 
2329         return valueOf(result);
2330     }
2331 
2332     /**
2333      * Returns a BigInteger whose value is equivalent to this BigInteger
2334      * with the designated bit flipped.
2335      * (Computes {@code (this ^ (1<<n))}.)
2336      *
2337      * @param  n index of bit to flip.
2338      * @return {@code this ^ (1<<n)}
2339      * @throws ArithmeticException {@code n} is negative.
2340      */
2341     public BigInteger flipBit(int n) {
2342         if (n<0)
2343             throw new ArithmeticException("Negative bit address");
2344 
2345         int intNum = n >>> 5;
2346         int[] result = new int[Math.max(intLength(), intNum+2)];
2347 
2348         for (int i=0; i<result.length; i++)
2349             result[result.length-i-1] = getInt(i);
2350 
2351         result[result.length-intNum-1] ^= (1 << (n & 31));
2352 
2353         return valueOf(result);
2354     }
2355 
2356     /**
2357      * Returns the index of the rightmost (lowest-order) one bit in this
2358      * BigInteger (the number of zero bits to the right of the rightmost
2359      * one bit).  Returns -1 if this BigInteger contains no one bits.
2360      * (Computes {@code (this==0? -1 : log2(this & -this))}.)
2361      *
2362      * @return index of the rightmost one bit in this BigInteger.
2363      */
2364     public int getLowestSetBit() {
2365         @SuppressWarnings("deprecation") int lsb = lowestSetBit - 2;
2366         if (lsb == -2) {  // lowestSetBit not initialized yet
2367             lsb = 0;
2368             if (signum == 0) {
2369                 lsb -= 1;
2370             } else {
2371                 // Search for lowest order nonzero int
2372                 int i,b;
2373                 for (i=0; (b = getInt(i))==0; i++)
2374                     ;
2375                 lsb += (i << 5) + Integer.numberOfTrailingZeros(b);
2376             }
2377             lowestSetBit = lsb + 2;
2378         }
2379         return lsb;
2380     }
2381 
2382 
2383     // Miscellaneous Bit Operations
2384 
2385     /**
2386      * Returns the number of bits in the minimal two's-complement
2387      * representation of this BigInteger, <i>excluding</i> a sign bit.
2388      * For positive BigIntegers, this is equivalent to the number of bits in
2389      * the ordinary binary representation.  (Computes
2390      * {@code (ceil(log2(this < 0 ? -this : this+1)))}.)
2391      *
2392      * @return number of bits in the minimal two's-complement
2393      *         representation of this BigInteger, <i>excluding</i> a sign bit.
2394      */
2395     public int bitLength() {
2396         @SuppressWarnings("deprecation") int n = bitLength - 1;
2397         if (n == -1) { // bitLength not initialized yet
2398             int[] m = mag;
2399             int len = m.length;
2400             if (len == 0) {
2401                 n = 0; // offset by one to initialize
2402             }  else {
2403                 // Calculate the bit length of the magnitude
2404                 int magBitLength = ((len - 1) << 5) + bitLengthForInt(mag[0]);
2405                  if (signum < 0) {
2406                      // Check if magnitude is a power of two
2407                      boolean pow2 = (Integer.bitCount(mag[0]) == 1);
2408                      for(int i=1; i< len && pow2; i++)
2409                          pow2 = (mag[i] == 0);
2410 
2411                      n = (pow2 ? magBitLength -1 : magBitLength);
2412                  } else {
2413                      n = magBitLength;
2414                  }
2415             }
2416             bitLength = n + 1;
2417         }
2418         return n;
2419     }
2420 
2421     /**
2422      * Returns the number of bits in the two's complement representation
2423      * of this BigInteger that differ from its sign bit.  This method is
2424      * useful when implementing bit-vector style sets atop BigIntegers.
2425      *
2426      * @return number of bits in the two's complement representation
2427      *         of this BigInteger that differ from its sign bit.
2428      */
2429     public int bitCount() {
2430         @SuppressWarnings("deprecation") int bc = bitCount - 1;
2431         if (bc == -1) {  // bitCount not initialized yet
2432             bc = 0;      // offset by one to initialize
2433             // Count the bits in the magnitude
2434             for (int i=0; i<mag.length; i++)
2435                 bc += Integer.bitCount(mag[i]);
2436             if (signum < 0) {
2437                 // Count the trailing zeros in the magnitude
2438                 int magTrailingZeroCount = 0, j;
2439                 for (j=mag.length-1; mag[j]==0; j--)
2440                     magTrailingZeroCount += 32;
2441                 magTrailingZeroCount += Integer.numberOfTrailingZeros(mag[j]);
2442                 bc += magTrailingZeroCount - 1;
2443             }
2444             bitCount = bc + 1;
2445         }
2446         return bc;
2447     }
2448 
2449     // Primality Testing
2450 
2451     /**
2452      * Returns {@code true} if this BigInteger is probably prime,
2453      * {@code false} if it's definitely composite.  If
2454      * {@code certainty} is &le; 0, {@code true} is
2455      * returned.
2456      *
2457      * @param  certainty a measure of the uncertainty that the caller is
2458      *         willing to tolerate: if the call returns {@code true}
2459      *         the probability that this BigInteger is prime exceeds
2460      *         (1 - 1/2<sup>{@code certainty}</sup>).  The execution time of
2461      *         this method is proportional to the value of this parameter.
2462      * @return {@code true} if this BigInteger is probably prime,
2463      *         {@code false} if it's definitely composite.
2464      */
2465     public boolean isProbablePrime(int certainty) {
2466         if (certainty <= 0)
2467             return true;
2468         BigInteger w = this.abs();
2469         if (w.equals(TWO))
2470             return true;
2471         if (!w.testBit(0) || w.equals(ONE))
2472             return false;
2473 
2474         return w.primeToCertainty(certainty, null);
2475     }
2476 
2477     // Comparison Operations
2478 
2479     /**
2480      * Compares this BigInteger with the specified BigInteger.  This
2481      * method is provided in preference to individual methods for each
2482      * of the six boolean comparison operators ({@literal <}, ==,
2483      * {@literal >}, {@literal >=}, !=, {@literal <=}).  The suggested
2484      * idiom for performing these comparisons is: {@code
2485      * (x.compareTo(y)} &lt;<i>op</i>&gt; {@code 0)}, where
2486      * &lt;<i>op</i>&gt; is one of the six comparison operators.
2487      *
2488      * @param  val BigInteger to which this BigInteger is to be compared.
2489      * @return -1, 0 or 1 as this BigInteger is numerically less than, equal
2490      *         to, or greater than {@code val}.
2491      */
2492     public int compareTo(BigInteger val) {
2493         if (signum == val.signum) {
2494             switch (signum) {
2495             case 1:
2496                 return compareMagnitude(val);
2497             case -1:
2498                 return val.compareMagnitude(this);
2499             default:
2500                 return 0;
2501             }
2502         }
2503         return signum > val.signum ? 1 : -1;
2504     }
2505 
2506     /**
2507      * Compares the magnitude array of this BigInteger with the specified
2508      * BigInteger's. This is the version of compareTo ignoring sign.
2509      *
2510      * @param val BigInteger whose magnitude array to be compared.
2511      * @return -1, 0 or 1 as this magnitude array is less than, equal to or
2512      *         greater than the magnitude aray for the specified BigInteger's.
2513      */
2514     final int compareMagnitude(BigInteger val) {
2515         int[] m1 = mag;
2516         int len1 = m1.length;
2517         int[] m2 = val.mag;
2518         int len2 = m2.length;
2519         if (len1 < len2)
2520             return -1;
2521         if (len1 > len2)
2522             return 1;
2523         for (int i = 0; i < len1; i++) {
2524             int a = m1[i];
2525             int b = m2[i];
2526             if (a != b)
2527                 return ((a & LONG_MASK) < (b & LONG_MASK)) ? -1 : 1;
2528         }
2529         return 0;
2530     }
2531 
2532     /**
2533      * Compares this BigInteger with the specified Object for equality.
2534      *
2535      * @param  x Object to which this BigInteger is to be compared.
2536      * @return {@code true} if and only if the specified Object is a
2537      *         BigInteger whose value is numerically equal to this BigInteger.
2538      */
2539     public boolean equals(Object x) {
2540         // This test is just an optimization, which may or may not help
2541         if (x == this)
2542             return true;
2543 
2544         if (!(x instanceof BigInteger))
2545             return false;
2546 
2547         BigInteger xInt = (BigInteger) x;
2548         if (xInt.signum != signum)
2549             return false;
2550 
2551         int[] m = mag;
2552         int len = m.length;
2553         int[] xm = xInt.mag;
2554         if (len != xm.length)
2555             return false;
2556 
2557         for (int i = 0; i < len; i++)
2558             if (xm[i] != m[i])
2559                 return false;
2560 
2561         return true;
2562     }
2563 
2564     /**
2565      * Returns the minimum of this BigInteger and {@code val}.
2566      *
2567      * @param  val value with which the minimum is to be computed.
2568      * @return the BigInteger whose value is the lesser of this BigInteger and
2569      *         {@code val}.  If they are equal, either may be returned.
2570      */
2571     public BigInteger min(BigInteger val) {
2572         return (compareTo(val)<0 ? this : val);
2573     }
2574 
2575     /**
2576      * Returns the maximum of this BigInteger and {@code val}.
2577      *
2578      * @param  val value with which the maximum is to be computed.
2579      * @return the BigInteger whose value is the greater of this and
2580      *         {@code val}.  If they are equal, either may be returned.
2581      */
2582     public BigInteger max(BigInteger val) {
2583         return (compareTo(val)>0 ? this : val);
2584     }
2585 
2586 
2587     // Hash Function
2588 
2589     /**
2590      * Returns the hash code for this BigInteger.
2591      *
2592      * @return hash code for this BigInteger.
2593      */
2594     public int hashCode() {
2595         int hashCode = 0;
2596 
2597         for (int i=0; i<mag.length; i++)
2598             hashCode = (int)(31*hashCode + (mag[i] & LONG_MASK));
2599 
2600         return hashCode * signum;
2601     }
2602 
2603     /**
2604      * Returns the String representation of this BigInteger in the
2605      * given radix.  If the radix is outside the range from {@link
2606      * Character#MIN_RADIX} to {@link Character#MAX_RADIX} inclusive,
2607      * it will default to 10 (as is the case for
2608      * {@code Integer.toString}).  The digit-to-character mapping
2609      * provided by {@code Character.forDigit} is used, and a minus
2610      * sign is prepended if appropriate.  (This representation is
2611      * compatible with the {@link #BigInteger(String, int) (String,
2612      * int)} constructor.)
2613      *
2614      * @param  radix  radix of the String representation.
2615      * @return String representation of this BigInteger in the given radix.
2616      * @see    Integer#toString
2617      * @see    Character#forDigit
2618      * @see    #BigInteger(java.lang.String, int)
2619      */
2620     public String toString(int radix) {
2621         if (signum == 0)
2622             return "0";
2623         if (radix < Character.MIN_RADIX || radix > Character.MAX_RADIX)
2624             radix = 10;
2625 
2626         // Compute upper bound on number of digit groups and allocate space
2627         int maxNumDigitGroups = (4*mag.length + 6)/7;
2628         String digitGroup[] = new String[maxNumDigitGroups];
2629 
2630         // Translate number to string, a digit group at a time
2631         BigInteger tmp = this.abs();
2632         int numGroups = 0;
2633         while (tmp.signum != 0) {
2634             BigInteger d = longRadix[radix];
2635 
2636             MutableBigInteger q = new MutableBigInteger(),
2637                               a = new MutableBigInteger(tmp.mag),
2638                               b = new MutableBigInteger(d.mag);
2639             MutableBigInteger r = a.divide(b, q);
2640             BigInteger q2 = q.toBigInteger(tmp.signum * d.signum);
2641             BigInteger r2 = r.toBigInteger(tmp.signum * d.signum);
2642 
2643             digitGroup[numGroups++] = Long.toString(r2.longValue(), radix);
2644             tmp = q2;
2645         }
2646 
2647         // Put sign (if any) and first digit group into result buffer
2648         StringBuilder buf = new StringBuilder(numGroups*digitsPerLong[radix]+1);
2649         if (signum<0)
2650             buf.append('-');
2651         buf.append(digitGroup[numGroups-1]);
2652 
2653         // Append remaining digit groups padded with leading zeros
2654         for (int i=numGroups-2; i>=0; i--) {
2655             // Prepend (any) leading zeros for this digit group
2656             int numLeadingZeros = digitsPerLong[radix]-digitGroup[i].length();
2657             if (numLeadingZeros != 0)
2658                 buf.append(zeros[numLeadingZeros]);
2659             buf.append(digitGroup[i]);
2660         }
2661         return buf.toString();
2662     }
2663 
2664     /* zero[i] is a string of i consecutive zeros. */
2665     private static String zeros[] = new String[64];
2666     static {
2667         zeros[63] =
2668             "000000000000000000000000000000000000000000000000000000000000000";
2669         for (int i=0; i<63; i++)
2670             zeros[i] = zeros[63].substring(0, i);
2671     }
2672 
2673     /**
2674      * Returns the decimal String representation of this BigInteger.
2675      * The digit-to-character mapping provided by
2676      * {@code Character.forDigit} is used, and a minus sign is
2677      * prepended if appropriate.  (This representation is compatible
2678      * with the {@link #BigInteger(String) (String)} constructor, and
2679      * allows for String concatenation with Java's + operator.)
2680      *
2681      * @return decimal String representation of this BigInteger.
2682      * @see    Character#forDigit
2683      * @see    #BigInteger(java.lang.String)
2684      */
2685     public String toString() {
2686         return toString(10);
2687     }
2688 
2689     /**
2690      * Returns a byte array containing the two's-complement
2691      * representation of this BigInteger.  The byte array will be in
2692      * <i>big-endian</i> byte-order: the most significant byte is in
2693      * the zeroth element.  The array will contain the minimum number
2694      * of bytes required to represent this BigInteger, including at
2695      * least one sign bit, which is {@code (ceil((this.bitLength() +
2696      * 1)/8))}.  (This representation is compatible with the
2697      * {@link #BigInteger(byte[]) (byte[])} constructor.)
2698      *
2699      * @return a byte array containing the two's-complement representation of
2700      *         this BigInteger.
2701      * @see    #BigInteger(byte[])
2702      */
2703     public byte[] toByteArray() {
2704         int byteLen = bitLength()/8 + 1;
2705         byte[] byteArray = new byte[byteLen];
2706 
2707         for (int i=byteLen-1, bytesCopied=4, nextInt=0, intIndex=0; i>=0; i--) {
2708             if (bytesCopied == 4) {
2709                 nextInt = getInt(intIndex++);
2710                 bytesCopied = 1;
2711             } else {
2712                 nextInt >>>= 8;
2713                 bytesCopied++;
2714             }
2715             byteArray[i] = (byte)nextInt;
2716         }
2717         return byteArray;
2718     }
2719 
2720     /**
2721      * Converts this BigInteger to an {@code int}.  This
2722      * conversion is analogous to a
2723      * <i>narrowing primitive conversion</i> from {@code long} to
2724      * {@code int} as defined in section 5.1.3 of
2725      * <cite>The Java&trade; Language Specification</cite>:
2726      * if this BigInteger is too big to fit in an
2727      * {@code int}, only the low-order 32 bits are returned.
2728      * Note that this conversion can lose information about the
2729      * overall magnitude of the BigInteger value as well as return a
2730      * result with the opposite sign.
2731      *
2732      * @return this BigInteger converted to an {@code int}.
2733      */
2734     public int intValue() {
2735         int result = 0;
2736         result = getInt(0);
2737         return result;
2738     }
2739 
2740     /**
2741      * Converts this BigInteger to a {@code long}.  This
2742      * conversion is analogous to a
2743      * <i>narrowing primitive conversion</i> from {@code long} to
2744      * {@code int} as defined in section 5.1.3 of
2745      * <cite>The Java&trade; Language Specification</cite>:
2746      * if this BigInteger is too big to fit in a
2747      * {@code long}, only the low-order 64 bits are returned.
2748      * Note that this conversion can lose information about the
2749      * overall magnitude of the BigInteger value as well as return a
2750      * result with the opposite sign.
2751      *
2752      * @return this BigInteger converted to a {@code long}.
2753      */
2754     public long longValue() {
2755         long result = 0;
2756 
2757         for (int i=1; i>=0; i--)
2758             result = (result << 32) + (getInt(i) & LONG_MASK);
2759         return result;
2760     }
2761 
2762     /**
2763      * Converts this BigInteger to a {@code float}.  This
2764      * conversion is similar to the
2765      * <i>narrowing primitive conversion</i> from {@code double} to
2766      * {@code float} as defined in section 5.1.3 of
2767      * <cite>The Java&trade; Language Specification</cite>:
2768      * if this BigInteger has too great a magnitude
2769      * to represent as a {@code float}, it will be converted to
2770      * {@link Float#NEGATIVE_INFINITY} or {@link
2771      * Float#POSITIVE_INFINITY} as appropriate.  Note that even when
2772      * the return value is finite, this conversion can lose
2773      * information about the precision of the BigInteger value.
2774      *
2775      * @return this BigInteger converted to a {@code float}.
2776      */
2777     public float floatValue() {
2778         // Somewhat inefficient, but guaranteed to work.
2779         return Float.parseFloat(this.toString());
2780     }
2781 
2782     /**
2783      * Converts this BigInteger to a {@code double}.  This
2784      * conversion is similar to the
2785      * <i>narrowing primitive conversion</i> from {@code double} to
2786      * {@code float} as defined in section 5.1.3 of
2787      * <cite>The Java&trade; Language Specification</cite>:
2788      * if this BigInteger has too great a magnitude
2789      * to represent as a {@code double}, it will be converted to
2790      * {@link Double#NEGATIVE_INFINITY} or {@link
2791      * Double#POSITIVE_INFINITY} as appropriate.  Note that even when
2792      * the return value is finite, this conversion can lose
2793      * information about the precision of the BigInteger value.
2794      *
2795      * @return this BigInteger converted to a {@code double}.
2796      */
2797     public double doubleValue() {
2798         // Somewhat inefficient, but guaranteed to work.
2799         return Double.parseDouble(this.toString());
2800     }
2801 
2802     /**
2803      * Returns a copy of the input array stripped of any leading zero bytes.
2804      */
2805     private static int[] stripLeadingZeroInts(int val[]) {
2806         int vlen = val.length;
2807         int keep;
2808 
2809         // Find first nonzero byte
2810         for (keep = 0; keep < vlen && val[keep] == 0; keep++)
2811             ;
2812         return java.util.Arrays.copyOfRange(val, keep, vlen);
2813     }
2814 
2815     /**
2816      * Returns the input array stripped of any leading zero bytes.
2817      * Since the source is trusted the copying may be skipped.
2818      */
2819     private static int[] trustedStripLeadingZeroInts(int val[]) {
2820         int vlen = val.length;
2821         int keep;
2822 
2823         // Find first nonzero byte
2824         for (keep = 0; keep < vlen && val[keep] == 0; keep++)
2825             ;
2826         return keep == 0 ? val : java.util.Arrays.copyOfRange(val, keep, vlen);
2827     }
2828 
2829     /**
2830      * Returns a copy of the input array stripped of any leading zero bytes.
2831      */
2832     private static int[] stripLeadingZeroBytes(byte a[]) {
2833         int byteLength = a.length;
2834         int keep;
2835 
2836         // Find first nonzero byte
2837         for (keep = 0; keep < byteLength && a[keep]==0; keep++)
2838             ;
2839 
2840         // Allocate new array and copy relevant part of input array
2841         int intLength = ((byteLength - keep) + 3) >>> 2;
2842         int[] result = new int[intLength];
2843         int b = byteLength - 1;
2844         for (int i = intLength-1; i >= 0; i--) {
2845             result[i] = a[b--] & 0xff;
2846             int bytesRemaining = b - keep + 1;
2847             int bytesToTransfer = Math.min(3, bytesRemaining);
2848             for (int j=8; j <= (bytesToTransfer << 3); j += 8)
2849                 result[i] |= ((a[b--] & 0xff) << j);
2850         }
2851         return result;
2852     }
2853 
2854     /**
2855      * Takes an array a representing a negative 2's-complement number and
2856      * returns the minimal (no leading zero bytes) unsigned whose value is -a.
2857      */
2858     private static int[] makePositive(byte a[]) {
2859         int keep, k;
2860         int byteLength = a.length;
2861 
2862         // Find first non-sign (0xff) byte of input
2863         for (keep=0; keep<byteLength && a[keep]==-1; keep++)
2864             ;
2865 
2866 
2867         /* Allocate output array.  If all non-sign bytes are 0x00, we must
2868          * allocate space for one extra output byte. */
2869         for (k=keep; k<byteLength && a[k]==0; k++)
2870             ;
2871 
2872         int extraByte = (k==byteLength) ? 1 : 0;
2873         int intLength = ((byteLength - keep + extraByte) + 3)/4;
2874         int result[] = new int[intLength];
2875 
2876         /* Copy one's complement of input into output, leaving extra
2877          * byte (if it exists) == 0x00 */
2878         int b = byteLength - 1;
2879         for (int i = intLength-1; i >= 0; i--) {
2880             result[i] = a[b--] & 0xff;
2881             int numBytesToTransfer = Math.min(3, b-keep+1);
2882             if (numBytesToTransfer < 0)
2883                 numBytesToTransfer = 0;
2884             for (int j=8; j <= 8*numBytesToTransfer; j += 8)
2885                 result[i] |= ((a[b--] & 0xff) << j);
2886 
2887             // Mask indicates which bits must be complemented
2888             int mask = -1 >>> (8*(3-numBytesToTransfer));
2889             result[i] = ~result[i] & mask;
2890         }
2891 
2892         // Add one to one's complement to generate two's complement
2893         for (int i=result.length-1; i>=0; i--) {
2894             result[i] = (int)((result[i] & LONG_MASK) + 1);
2895             if (result[i] != 0)
2896                 break;
2897         }
2898 
2899         return result;
2900     }
2901 
2902     /**
2903      * Takes an array a representing a negative 2's-complement number and
2904      * returns the minimal (no leading zero ints) unsigned whose value is -a.
2905      */
2906     private static int[] makePositive(int a[]) {
2907         int keep, j;
2908 
2909         // Find first non-sign (0xffffffff) int of input
2910         for (keep=0; keep<a.length && a[keep]==-1; keep++)
2911             ;
2912 
2913         /* Allocate output array.  If all non-sign ints are 0x00, we must
2914          * allocate space for one extra output int. */
2915         for (j=keep; j<a.length && a[j]==0; j++)
2916             ;
2917         int extraInt = (j==a.length ? 1 : 0);
2918         int result[] = new int[a.length - keep + extraInt];
2919 
2920         /* Copy one's complement of input into output, leaving extra
2921          * int (if it exists) == 0x00 */
2922         for (int i = keep; i<a.length; i++)
2923             result[i - keep + extraInt] = ~a[i];
2924 
2925         // Add one to one's complement to generate two's complement
2926         for (int i=result.length-1; ++result[i]==0; i--)
2927             ;
2928 
2929         return result;
2930     }
2931 
2932     /*
2933      * The following two arrays are used for fast String conversions.  Both
2934      * are indexed by radix.  The first is the number of digits of the given
2935      * radix that can fit in a Java long without "going negative", i.e., the
2936      * highest integer n such that radix**n < 2**63.  The second is the
2937      * "long radix" that tears each number into "long digits", each of which
2938      * consists of the number of digits in the corresponding element in
2939      * digitsPerLong (longRadix[i] = i**digitPerLong[i]).  Both arrays have
2940      * nonsense values in their 0 and 1 elements, as radixes 0 and 1 are not
2941      * used.
2942      */
2943     private static int digitsPerLong[] = {0, 0,
2944         62, 39, 31, 27, 24, 22, 20, 19, 18, 18, 17, 17, 16, 16, 15, 15, 15, 14,
2945         14, 14, 14, 13, 13, 13, 13, 13, 13, 12, 12, 12, 12, 12, 12, 12, 12};
2946 
2947     private static BigInteger longRadix[] = {null, null,
2948         valueOf(0x4000000000000000L), valueOf(0x383d9170b85ff80bL),
2949         valueOf(0x4000000000000000L), valueOf(0x6765c793fa10079dL),
2950         valueOf(0x41c21cb8e1000000L), valueOf(0x3642798750226111L),
2951         valueOf(0x1000000000000000L), valueOf(0x12bf307ae81ffd59L),
2952         valueOf( 0xde0b6b3a7640000L), valueOf(0x4d28cb56c33fa539L),
2953         valueOf(0x1eca170c00000000L), valueOf(0x780c7372621bd74dL),
2954         valueOf(0x1e39a5057d810000L), valueOf(0x5b27ac993df97701L),
2955         valueOf(0x1000000000000000L), valueOf(0x27b95e997e21d9f1L),
2956         valueOf(0x5da0e1e53c5c8000L), valueOf( 0xb16a458ef403f19L),
2957         valueOf(0x16bcc41e90000000L), valueOf(0x2d04b7fdd9c0ef49L),
2958         valueOf(0x5658597bcaa24000L), valueOf( 0x6feb266931a75b7L),
2959         valueOf( 0xc29e98000000000L), valueOf(0x14adf4b7320334b9L),
2960         valueOf(0x226ed36478bfa000L), valueOf(0x383d9170b85ff80bL),
2961         valueOf(0x5a3c23e39c000000L), valueOf( 0x4e900abb53e6b71L),
2962         valueOf( 0x7600ec618141000L), valueOf( 0xaee5720ee830681L),
2963         valueOf(0x1000000000000000L), valueOf(0x172588ad4f5f0981L),
2964         valueOf(0x211e44f7d02c1000L), valueOf(0x2ee56725f06e5c71L),
2965         valueOf(0x41c21cb8e1000000L)};
2966 
2967     /*
2968      * These two arrays are the integer analogue of above.
2969      */
2970     private static int digitsPerInt[] = {0, 0, 30, 19, 15, 13, 11,
2971         11, 10, 9, 9, 8, 8, 8, 8, 7, 7, 7, 7, 7, 7, 7, 6, 6, 6, 6,
2972         6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5};
2973 
2974     private static int intRadix[] = {0, 0,
2975         0x40000000, 0x4546b3db, 0x40000000, 0x48c27395, 0x159fd800,
2976         0x75db9c97, 0x40000000, 0x17179149, 0x3b9aca00, 0xcc6db61,
2977         0x19a10000, 0x309f1021, 0x57f6c100, 0xa2f1b6f,  0x10000000,
2978         0x18754571, 0x247dbc80, 0x3547667b, 0x4c4b4000, 0x6b5a6e1d,
2979         0x6c20a40,  0x8d2d931,  0xb640000,  0xe8d4a51,  0x1269ae40,
2980         0x17179149, 0x1cb91000, 0x23744899, 0x2b73a840, 0x34e63b41,
2981         0x40000000, 0x4cfa3cc1, 0x5c13d840, 0x6d91b519, 0x39aa400
2982     };
2983 
2984     /**
2985      * These routines provide access to the two's complement representation
2986      * of BigIntegers.
2987      */
2988 
2989     /**
2990      * Returns the length of the two's complement representation in ints,
2991      * including space for at least one sign bit.
2992      */
2993     private int intLength() {
2994         return (bitLength() >>> 5) + 1;
2995     }
2996 
2997     /* Returns sign bit */
2998     private int signBit() {
2999         return signum < 0 ? 1 : 0;
3000     }
3001 
3002     /* Returns an int of sign bits */
3003     private int signInt() {
3004         return signum < 0 ? -1 : 0;
3005     }
3006 
3007     /**
3008      * Returns the specified int of the little-endian two's complement
3009      * representation (int 0 is the least significant).  The int number can
3010      * be arbitrarily high (values are logically preceded by infinitely many
3011      * sign ints).
3012      */
3013     private int getInt(int n) {
3014         if (n < 0)
3015             return 0;
3016         if (n >= mag.length)
3017             return signInt();
3018 
3019         int magInt = mag[mag.length-n-1];
3020 
3021         return (signum >= 0 ? magInt :
3022                 (n <= firstNonzeroIntNum() ? -magInt : ~magInt));
3023     }
3024 
3025     /**
3026      * Returns the index of the int that contains the first nonzero int in the
3027      * little-endian binary representation of the magnitude (int 0 is the
3028      * least significant). If the magnitude is zero, return value is undefined.
3029      */
3030      private int firstNonzeroIntNum() {
3031          int fn = firstNonzeroIntNum - 2;
3032          if (fn == -2) { // firstNonzeroIntNum not initialized yet
3033              fn = 0;
3034 
3035              // Search for the first nonzero int
3036              int i;
3037              int mlen = mag.length;
3038              for (i = mlen - 1; i >= 0 && mag[i] == 0; i--)
3039                  ;
3040              fn = mlen - i - 1;
3041              firstNonzeroIntNum = fn + 2; // offset by two to initialize
3042          }
3043          return fn;
3044      }
3045 
3046     /** use serialVersionUID from JDK 1.1. for interoperability */
3047     private static final long serialVersionUID = -8287574255936472291L;
3048 
3049     /**
3050      * Serializable fields for BigInteger.
3051      *
3052      * @serialField signum  int
3053      *              signum of this BigInteger.
3054      * @serialField magnitude int[]
3055      *              magnitude array of this BigInteger.
3056      * @serialField bitCount  int
3057      *              number of bits in this BigInteger
3058      * @serialField bitLength int
3059      *              the number of bits in the minimal two's-complement
3060      *              representation of this BigInteger
3061      * @serialField lowestSetBit int
3062      *              lowest set bit in the twos complement representation
3063      */
3064     private static final ObjectStreamField[] serialPersistentFields = {
3065         new ObjectStreamField("signum", Integer.TYPE),
3066         new ObjectStreamField("magnitude", byte[].class),
3067         new ObjectStreamField("bitCount", Integer.TYPE),
3068         new ObjectStreamField("bitLength", Integer.TYPE),
3069         new ObjectStreamField("firstNonzeroByteNum", Integer.TYPE),
3070         new ObjectStreamField("lowestSetBit", Integer.TYPE)
3071         };
3072 
3073     /**
3074      * Reconstitute the {@code BigInteger} instance from a stream (that is,
3075      * deserialize it). The magnitude is read in as an array of bytes
3076      * for historical reasons, but it is converted to an array of ints
3077      * and the byte array is discarded.
3078      * Note:
3079      * The current convention is to initialize the cache fields, bitCount,
3080      * bitLength and lowestSetBit, to 0 rather than some other marker value.
3081      * Therefore, no explicit action to set these fields needs to be taken in
3082      * readObject because those fields already have a 0 value be default since
3083      * defaultReadObject is not being used.
3084      */
3085     private void readObject(java.io.ObjectInputStream s)
3086         throws java.io.IOException, ClassNotFoundException {
3087         /*
3088          * In order to maintain compatibility with previous serialized forms,
3089          * the magnitude of a BigInteger is serialized as an array of bytes.
3090          * The magnitude field is used as a temporary store for the byte array
3091          * that is deserialized. The cached computation fields should be
3092          * transient but are serialized for compatibility reasons.
3093          */
3094 
3095         // prepare to read the alternate persistent fields
3096         ObjectInputStream.GetField fields = s.readFields();
3097 
3098         // Read the alternate persistent fields that we care about
3099         int sign = fields.get("signum", -2);
3100         byte[] magnitude = (byte[])fields.get("magnitude", null);
3101 
3102         // Validate signum
3103         if (sign < -1 || sign > 1) {
3104             String message = "BigInteger: Invalid signum value";
3105             if (fields.defaulted("signum"))
3106                 message = "BigInteger: Signum not present in stream";
3107             throw new java.io.StreamCorruptedException(message);
3108         }
3109         if ((magnitude.length == 0) != (sign == 0)) {
3110             String message = "BigInteger: signum-magnitude mismatch";
3111             if (fields.defaulted("magnitude"))
3112                 message = "BigInteger: Magnitude not present in stream";
3113             throw new java.io.StreamCorruptedException(message);
3114         }
3115 
3116         // Commit final fields via Unsafe
3117         unsafe.putIntVolatile(this, signumOffset, sign);
3118 
3119         // Calculate mag field from magnitude and discard magnitude
3120         unsafe.putObjectVolatile(this, magOffset,
3121                                  stripLeadingZeroBytes(magnitude));
3122     }
3123 
3124     // Support for resetting final fields while deserializing
3125     private static final sun.misc.Unsafe unsafe = sun.misc.Unsafe.getUnsafe();
3126     private static final long signumOffset;
3127     private static final long magOffset;
3128     static {
3129         try {
3130             signumOffset = unsafe.objectFieldOffset
3131                 (BigInteger.class.getDeclaredField("signum"));
3132             magOffset = unsafe.objectFieldOffset
3133                 (BigInteger.class.getDeclaredField("mag"));
3134         } catch (Exception ex) {
3135             throw new Error(ex);
3136         }
3137     }
3138 
3139     /**
3140      * Save the {@code BigInteger} instance to a stream.
3141      * The magnitude of a BigInteger is serialized as a byte array for
3142      * historical reasons.
3143      *
3144      * @serialData two necessary fields are written as well as obsolete
3145      *             fields for compatibility with older versions.
3146      */
3147     private void writeObject(ObjectOutputStream s) throws IOException {
3148         // set the values of the Serializable fields
3149         ObjectOutputStream.PutField fields = s.putFields();
3150         fields.put("signum", signum);
3151         fields.put("magnitude", magSerializedForm());
3152         // The values written for cached fields are compatible with older
3153         // versions, but are ignored in readObject so don't otherwise matter.
3154         fields.put("bitCount", -1);
3155         fields.put("bitLength", -1);
3156         fields.put("lowestSetBit", -2);
3157         fields.put("firstNonzeroByteNum", -2);
3158 
3159         // save them
3160         s.writeFields();
3161 }
3162 
3163     /**
3164      * Returns the mag array as an array of bytes.
3165      */
3166     private byte[] magSerializedForm() {
3167         int len = mag.length;
3168 
3169         int bitLen = (len == 0 ? 0 : ((len - 1) << 5) + bitLengthForInt(mag[0]));
3170         int byteLen = (bitLen + 7) >>> 3;
3171         byte[] result = new byte[byteLen];
3172 
3173         for (int i = byteLen - 1, bytesCopied = 4, intIndex = len - 1, nextInt = 0;
3174              i>=0; i--) {
3175             if (bytesCopied == 4) {
3176                 nextInt = mag[intIndex--];
3177                 bytesCopied = 1;
3178             } else {
3179                 nextInt >>>= 8;
3180                 bytesCopied++;
3181             }
3182             result[i] = (byte)nextInt;
3183         }
3184         return result;
3185     }
3186 }