From d6fb0646531172f23648441c224cdcccd721b894 Mon Sep 17 00:00:00 2001 From: xm <1271024303@qq.com> Date: 星期一, 14 十二月 2020 09:59:01 +0800 Subject: [PATCH] 请合并代码,完成晾衣架最终功能。 --- ZigbeeApp/Shared/Phone/ZigBee/Common/BigInteger.cs | 3126 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 3,126 insertions(+), 0 deletions(-) diff --git a/ZigbeeApp/Shared/Phone/ZigBee/Common/BigInteger.cs b/ZigbeeApp/Shared/Phone/ZigBee/Common/BigInteger.cs new file mode 100755 index 0000000..6e96ac4 --- /dev/null +++ b/ZigbeeApp/Shared/Phone/ZigBee/Common/BigInteger.cs @@ -0,0 +1,3126 @@ +锘縰sing System; +using System.Collections.Generic; +using System.Linq; +using System.Text; +namespace ZigBee.Common +{ + public class BigInteger + { + // maximum length of the BigInteger in uint (4 bytes) + // change this to suit the required level of precision. + + private const int maxLength = 70; + + // primes smaller than 2000 to test the generated prime number + + public static readonly int[] primesBelow2000 = { + 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, + 101, 103, 107, 109, 113, 127, 131, 137, 139, 149, 151, 157, 163, 167, 173, 179, 181, 191, 193, 197, 199, + 211, 223, 227, 229, 233, 239, 241, 251, 257, 263, 269, 271, 277, 281, 283, 293, + 307, 311, 313, 317, 331, 337, 347, 349, 353, 359, 367, 373, 379, 383, 389, 397, + 401, 409, 419, 421, 431, 433, 439, 443, 449, 457, 461, 463, 467, 479, 487, 491, 499, + 503, 509, 521, 523, 541, 547, 557, 563, 569, 571, 577, 587, 593, 599, + 601, 607, 613, 617, 619, 631, 641, 643, 647, 653, 659, 661, 673, 677, 683, 691, + 701, 709, 719, 727, 733, 739, 743, 751, 757, 761, 769, 773, 787, 797, + 809, 811, 821, 823, 827, 829, 839, 853, 857, 859, 863, 877, 881, 883, 887, + 907, 911, 919, 929, 937, 941, 947, 953, 967, 971, 977, 983, 991, 997, + 1009, 1013, 1019, 1021, 1031, 1033, 1039, 1049, 1051, 1061, 1063, 1069, 1087, 1091, 1093, 1097, + 1103, 1109, 1117, 1123, 1129, 1151, 1153, 1163, 1171, 1181, 1187, 1193, + 1201, 1213, 1217, 1223, 1229, 1231, 1237, 1249, 1259, 1277, 1279, 1283, 1289, 1291, 1297, + 1301, 1303, 1307, 1319, 1321, 1327, 1361, 1367, 1373, 1381, 1399, + 1409, 1423, 1427, 1429, 1433, 1439, 1447, 1451, 1453, 1459, 1471, 1481, 1483, 1487, 1489, 1493, 1499, + 1511, 1523, 1531, 1543, 1549, 1553, 1559, 1567, 1571, 1579, 1583, 1597, + 1601, 1607, 1609, 1613, 1619, 1621, 1627, 1637, 1657, 1663, 1667, 1669, 1693, 1697, 1699, + 1709, 1721, 1723, 1733, 1741, 1747, 1753, 1759, 1777, 1783, 1787, 1789, + 1801, 1811, 1823, 1831, 1847, 1861, 1867, 1871, 1873, 1877, 1879, 1889, + 1901, 1907, 1913, 1931, 1933, 1949, 1951, 1973, 1979, 1987, 1993, 1997, 1999 }; + + + private uint[] data = null; // stores bytes from the Big Integer + public int dataLength; // number of actual chars used + + + //*********************************************************************** + // Constructor (Default value for BigInteger is 0 + //*********************************************************************** + + public BigInteger() + { + data = new uint[maxLength]; + dataLength = 1; + } + + + //*********************************************************************** + // Constructor (Default value provided by long) + //*********************************************************************** + + public BigInteger(long value) + { + data = new uint[maxLength]; + long tempVal = value; + + // copy bytes from long to BigInteger without any assumption of + // the length of the long datatype + + dataLength = 0; + while (value != 0 && dataLength < maxLength) + { + data[dataLength] = (uint)(value & 0xFFFFFFFF); + value >>= 32; + dataLength++; + } + + if (tempVal > 0) // overflow check for +ve value + { + if (value != 0 || (data[maxLength - 1] & 0x80000000) != 0) + throw (new ArithmeticException("Positive overflow in constructor.")); + } + else if (tempVal < 0) // underflow check for -ve value + { + if (value != -1 || (data[dataLength - 1] & 0x80000000) == 0) + throw (new ArithmeticException("Negative underflow in constructor.")); + } + + if (dataLength == 0) + dataLength = 1; + } + + + //*********************************************************************** + // Constructor (Default value provided by ulong) + //*********************************************************************** + + public BigInteger(ulong value) + { + data = new uint[maxLength]; + + // copy bytes from ulong to BigInteger without any assumption of + // the length of the ulong datatype + + dataLength = 0; + while (value != 0 && dataLength < maxLength) + { + data[dataLength] = (uint)(value & 0xFFFFFFFF); + value >>= 32; + dataLength++; + } + + if (value != 0 || (data[maxLength - 1] & 0x80000000) != 0) + throw (new ArithmeticException("Positive overflow in constructor.")); + + if (dataLength == 0) + dataLength = 1; + } + + //*********************************************************************** + // Constructor (Default value provided by BigInteger) + //*********************************************************************** + + public BigInteger(BigInteger bi) + { + data = new uint[maxLength]; + + dataLength = bi.dataLength; + + for (int i = 0; i < dataLength; i++) + data[i] = bi.data[i]; + } + + + //*********************************************************************** + // Constructor (Default value provided by a string of digits of the + // specified base) + // + // Example (base 10) + // ----------------- + // To initialize "a" with the default value of 1234 in base 10 + // BigInteger a = new BigInteger("1234", 10) + // + // To initialize "a" with the default value of -1234 + // BigInteger a = new BigInteger("-1234", 10) + // + // Example (base 16) + // ----------------- + // To initialize "a" with the default value of 0x1D4F in base 16 + // BigInteger a = new BigInteger("1D4F", 16) + // + // To initialize "a" with the default value of -0x1D4F + // BigInteger a = new BigInteger("-1D4F", 16) + // + // Note that string values are specified in the <sign><magnitude> + // format. + // + //*********************************************************************** + + public BigInteger(string value, int radix) + { + BigInteger multiplier = new BigInteger(1); + BigInteger result = new BigInteger(); + value = (value.ToUpper()).Trim(); + int limit = 0; + + if (value[0] == '-') + limit = 1; + + for (int i = value.Length - 1; i >= limit; i--) + { + int posVal = (int)value[i]; + + if (posVal >= '0' && posVal <= '9') + posVal -= '0'; + else if (posVal >= 'A' && posVal <= 'Z') + posVal = (posVal - 'A') + 10; + else + posVal = 9999999; // arbitrary large + + + if (posVal >= radix) + throw (new ArithmeticException("Invalid string in constructor.")); + else + { + if (value[0] == '-') + posVal = -posVal; + + result = result + (multiplier * posVal); + + if ((i - 1) >= limit) + multiplier = multiplier * radix; + } + } + + if (value[0] == '-') // negative values + { + if ((result.data[maxLength - 1] & 0x80000000) == 0) + throw (new ArithmeticException("Negative underflow in constructor.")); + } + else // positive values + { + if ((result.data[maxLength - 1] & 0x80000000) != 0) + throw (new ArithmeticException("Positive overflow in constructor.")); + } + + data = new uint[maxLength]; + for (int i = 0; i < result.dataLength; i++) + data[i] = result.data[i]; + + dataLength = result.dataLength; + } + + + //*********************************************************************** + // Constructor (Default value provided by an array of bytes) + // + // The lowest index of the input byte array (i.e [0]) should contain the + // most significant byte of the number, and the highest index should + // contain the least significant byte. + // + // E.g. + // To initialize "a" with the default value of 0x1D4F in base 16 + // byte[] temp = { 0x1D, 0x4F }; + // BigInteger a = new BigInteger(temp) + // + // Note that this method of initialization does not allow the + // sign to be specified. + // + //*********************************************************************** + + public BigInteger(byte[] inData) + { + dataLength = inData.Length >> 2; + + int leftOver = inData.Length & 0x3; + if (leftOver != 0) // length not multiples of 4 + dataLength++; + + + if (dataLength > maxLength) + throw (new ArithmeticException("Byte overflow in constructor.")); + + data = new uint[maxLength]; + + for (int i = inData.Length - 1, j = 0; i >= 3; i -= 4, j++) + { + data[j] = (uint)((inData[i - 3] << 24) + (inData[i - 2] << 16) + + (inData[i - 1] << 8) + inData[i]); + } + + if (leftOver == 1) + data[dataLength - 1] = (uint)inData[0]; + else if (leftOver == 2) + data[dataLength - 1] = (uint)((inData[0] << 8) + inData[1]); + else if (leftOver == 3) + data[dataLength - 1] = (uint)((inData[0] << 16) + (inData[1] << 8) + inData[2]); + + + while (dataLength > 1 && data[dataLength - 1] == 0) + dataLength--; + + //Console.WriteLine("Len = " + dataLength); + } + + + //*********************************************************************** + // Constructor (Default value provided by an array of bytes of the + // specified length.) + //*********************************************************************** + + public BigInteger(byte[] inData, int inLen) + { + dataLength = inLen >> 2; + + int leftOver = inLen & 0x3; + if (leftOver != 0) // length not multiples of 4 + dataLength++; + + if (dataLength > maxLength || inLen > inData.Length) + throw (new ArithmeticException("Byte overflow in constructor.")); + + + data = new uint[maxLength]; + + for (int i = inLen - 1, j = 0; i >= 3; i -= 4, j++) + { + data[j] = (uint)((inData[i - 3] << 24) + (inData[i - 2] << 16) + + (inData[i - 1] << 8) + inData[i]); + } + + if (leftOver == 1) + data[dataLength - 1] = (uint)inData[0]; + else if (leftOver == 2) + data[dataLength - 1] = (uint)((inData[0] << 8) + inData[1]); + else if (leftOver == 3) + data[dataLength - 1] = (uint)((inData[0] << 16) + (inData[1] << 8) + inData[2]); + + + if (dataLength == 0) + dataLength = 1; + + while (dataLength > 1 && data[dataLength - 1] == 0) + dataLength--; + + //Console.WriteLine("Len = " + dataLength); + } + + + //*********************************************************************** + // Constructor (Default value provided by an array of unsigned integers) + //********************************************************************* + + public BigInteger(uint[] inData) + { + dataLength = inData.Length; + + if (dataLength > maxLength) + throw (new ArithmeticException("Byte overflow in constructor.")); + + data = new uint[maxLength]; + + for (int i = dataLength - 1, j = 0; i >= 0; i--, j++) + data[j] = inData[i]; + + while (dataLength > 1 && data[dataLength - 1] == 0) + dataLength--; + + //Console.WriteLine("Len = " + dataLength); + } + + + //*********************************************************************** + // Overloading of the typecast operator. + // For BigInteger bi = 10; + //*********************************************************************** + + public static implicit operator BigInteger(long value) + { + return (new BigInteger(value)); + } + + public static implicit operator BigInteger(ulong value) + { + return (new BigInteger(value)); + } + + public static implicit operator BigInteger(int value) + { + return (new BigInteger((long)value)); + } + + public static implicit operator BigInteger(uint value) + { + return (new BigInteger((ulong)value)); + } + + + //*********************************************************************** + // Overloading of addition operator + //*********************************************************************** + + public static BigInteger operator +(BigInteger bi1, BigInteger bi2) + { + BigInteger result = new BigInteger(); + + result.dataLength = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + + long carry = 0; + for (int i = 0; i < result.dataLength; i++) + { + long sum = (long)bi1.data[i] + (long)bi2.data[i] + carry; + carry = sum >> 32; + result.data[i] = (uint)(sum & 0xFFFFFFFF); + } + + if (carry != 0 && result.dataLength < maxLength) + { + result.data[result.dataLength] = (uint)(carry); + result.dataLength++; + } + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + + // overflow check + int lastPos = maxLength - 1; + if ((bi1.data[lastPos] & 0x80000000) == (bi2.data[lastPos] & 0x80000000) && + (result.data[lastPos] & 0x80000000) != (bi1.data[lastPos] & 0x80000000)) + { + throw (new ArithmeticException()); + } + + return result; + } + + + //*********************************************************************** + // Overloading of the unary ++ operator + //*********************************************************************** + + public static BigInteger operator ++(BigInteger bi1) + { + BigInteger result = new BigInteger(bi1); + + long val, carry = 1; + int index = 0; + + while (carry != 0 && index < maxLength) + { + val = (long)(result.data[index]); + val++; + + result.data[index] = (uint)(val & 0xFFFFFFFF); + carry = val >> 32; + + index++; + } + + if (index > result.dataLength) + result.dataLength = index; + else + { + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + } + + // overflow check + int lastPos = maxLength - 1; + + // overflow if initial value was +ve but ++ caused a sign + // change to negative. + + if ((bi1.data[lastPos] & 0x80000000) == 0 && + (result.data[lastPos] & 0x80000000) != (bi1.data[lastPos] & 0x80000000)) + { + throw (new ArithmeticException("Overflow in ++.")); + } + return result; + } + + + //*********************************************************************** + // Overloading of subtraction operator + //*********************************************************************** + + public static BigInteger operator -(BigInteger bi1, BigInteger bi2) + { + BigInteger result = new BigInteger(); + + result.dataLength = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + + long carryIn = 0; + for (int i = 0; i < result.dataLength; i++) + { + long diff; + + diff = (long)bi1.data[i] - (long)bi2.data[i] - carryIn; + result.data[i] = (uint)(diff & 0xFFFFFFFF); + + if (diff < 0) + carryIn = 1; + else + carryIn = 0; + } + + // roll over to negative + if (carryIn != 0) + { + for (int i = result.dataLength; i < maxLength; i++) + result.data[i] = 0xFFFFFFFF; + result.dataLength = maxLength; + } + + // fixed in v1.03 to give correct datalength for a - (-b) + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + // overflow check + + int lastPos = maxLength - 1; + if ((bi1.data[lastPos] & 0x80000000) != (bi2.data[lastPos] & 0x80000000) && + (result.data[lastPos] & 0x80000000) != (bi1.data[lastPos] & 0x80000000)) + { + throw (new ArithmeticException()); + } + + return result; + } + + + //*********************************************************************** + // Overloading of the unary -- operator + //*********************************************************************** + + public static BigInteger operator --(BigInteger bi1) + { + BigInteger result = new BigInteger(bi1); + + long val; + bool carryIn = true; + int index = 0; + + while (carryIn && index < maxLength) + { + val = (long)(result.data[index]); + val--; + + result.data[index] = (uint)(val & 0xFFFFFFFF); + + if (val >= 0) + carryIn = false; + + index++; + } + + if (index > result.dataLength) + result.dataLength = index; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + // overflow check + int lastPos = maxLength - 1; + + // overflow if initial value was -ve but -- caused a sign + // change to positive. + + if ((bi1.data[lastPos] & 0x80000000) != 0 && + (result.data[lastPos] & 0x80000000) != (bi1.data[lastPos] & 0x80000000)) + { + throw (new ArithmeticException("Underflow in --.")); + } + + return result; + } + + + //*********************************************************************** + // Overloading of multiplication operator + //*********************************************************************** + + public static BigInteger operator *(BigInteger bi1, BigInteger bi2) + { + int lastPos = maxLength - 1; + bool bi1Neg = false, bi2Neg = false; + + // take the absolute value of the inputs + try + { + if ((bi1.data[lastPos] & 0x80000000) != 0) // bi1 negative + { + bi1Neg = true; bi1 = -bi1; + } + if ((bi2.data[lastPos] & 0x80000000) != 0) // bi2 negative + { + bi2Neg = true; bi2 = -bi2; + } + } + catch (Exception) { } + + BigInteger result = new BigInteger(); + + // multiply the absolute values + try + { + for (int i = 0; i < bi1.dataLength; i++) + { + if (bi1.data[i] == 0) continue; + + ulong mcarry = 0; + for (int j = 0, k = i; j < bi2.dataLength; j++, k++) + { + // k = i + j + ulong val = ((ulong)bi1.data[i] * (ulong)bi2.data[j]) + + (ulong)result.data[k] + mcarry; + + result.data[k] = (uint)(val & 0xFFFFFFFF); + mcarry = (val >> 32); + } + + if (mcarry != 0) + result.data[i + bi2.dataLength] = (uint)mcarry; + } + } + catch (Exception) + { + throw (new ArithmeticException("Multiplication overflow.")); + } + + + result.dataLength = bi1.dataLength + bi2.dataLength; + if (result.dataLength > maxLength) + result.dataLength = maxLength; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + // overflow check (result is -ve) + if ((result.data[lastPos] & 0x80000000) != 0) + { + if (bi1Neg != bi2Neg && result.data[lastPos] == 0x80000000) // different sign + { + // handle the special case where multiplication produces + // a max negative number in 2's complement. + + if (result.dataLength == 1) + return result; + else + { + bool isMaxNeg = true; + for (int i = 0; i < result.dataLength - 1 && isMaxNeg; i++) + { + if (result.data[i] != 0) + isMaxNeg = false; + } + + if (isMaxNeg) + return result; + } + } + + throw (new ArithmeticException("Multiplication overflow.")); + } + + // if input has different signs, then result is -ve + if (bi1Neg != bi2Neg) + return -result; + + return result; + } + + + + //*********************************************************************** + // Overloading of unary << operators + //*********************************************************************** + + public static BigInteger operator <<(BigInteger bi1, int shiftVal) + { + BigInteger result = new BigInteger(bi1); + result.dataLength = shiftLeft(result.data, shiftVal); + + return result; + } + + + // least significant bits at lower part of buffer + + private static int shiftLeft(uint[] buffer, int shiftVal) + { + int shiftAmount = 32; + int bufLen = buffer.Length; + + while (bufLen > 1 && buffer[bufLen - 1] == 0) + bufLen--; + + for (int count = shiftVal; count > 0;) + { + if (count < shiftAmount) + shiftAmount = count; + + //Console.WriteLine("shiftAmount = {0}", shiftAmount); + + ulong carry = 0; + for (int i = 0; i < bufLen; i++) + { + ulong val = ((ulong)buffer[i]) << shiftAmount; + val |= carry; + + buffer[i] = (uint)(val & 0xFFFFFFFF); + carry = val >> 32; + } + + if (carry != 0) + { + if (bufLen + 1 <= buffer.Length) + { + buffer[bufLen] = (uint)carry; + bufLen++; + } + } + count -= shiftAmount; + } + return bufLen; + } + + + //*********************************************************************** + // Overloading of unary >> operators + //*********************************************************************** + + public static BigInteger operator >>(BigInteger bi1, int shiftVal) + { + BigInteger result = new BigInteger(bi1); + result.dataLength = shiftRight(result.data, shiftVal); + + + if ((bi1.data[maxLength - 1] & 0x80000000) != 0) // negative + { + for (int i = maxLength - 1; i >= result.dataLength; i--) + result.data[i] = 0xFFFFFFFF; + + uint mask = 0x80000000; + for (int i = 0; i < 32; i++) + { + if ((result.data[result.dataLength - 1] & mask) != 0) + break; + + result.data[result.dataLength - 1] |= mask; + mask >>= 1; + } + result.dataLength = maxLength; + } + + return result; + } + + + private static int shiftRight(uint[] buffer, int shiftVal) + { + int shiftAmount = 32; + int invShift = 0; + int bufLen = buffer.Length; + + while (bufLen > 1 && buffer[bufLen - 1] == 0) + bufLen--; + + //Console.WriteLine("bufLen = " + bufLen + " buffer.Length = " + buffer.Length); + + for (int count = shiftVal; count > 0;) + { + if (count < shiftAmount) + { + shiftAmount = count; + invShift = 32 - shiftAmount; + } + + //Console.WriteLine("shiftAmount = {0}", shiftAmount); + + ulong carry = 0; + for (int i = bufLen - 1; i >= 0; i--) + { + ulong val = ((ulong)buffer[i]) >> shiftAmount; + val |= carry; + + carry = ((ulong)buffer[i]) << invShift; + buffer[i] = (uint)(val); + } + + count -= shiftAmount; + } + + while (bufLen > 1 && buffer[bufLen - 1] == 0) + bufLen--; + + return bufLen; + } + + + //*********************************************************************** + // Overloading of the NOT operator (1's complement) + //*********************************************************************** + + public static BigInteger operator ~(BigInteger bi1) + { + BigInteger result = new BigInteger(bi1); + + for (int i = 0; i < maxLength; i++) + result.data[i] = (uint)(~(bi1.data[i])); + + result.dataLength = maxLength; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + return result; + } + + + //*********************************************************************** + // Overloading of the NEGATE operator (2's complement) + //*********************************************************************** + + public static BigInteger operator -(BigInteger bi1) + { + // handle neg of zero separately since it'll cause an overflow + // if we proceed. + + if (bi1.dataLength == 1 && bi1.data[0] == 0) + return (new BigInteger()); + + BigInteger result = new BigInteger(bi1); + + // 1's complement + for (int i = 0; i < maxLength; i++) + result.data[i] = (uint)(~(bi1.data[i])); + + // add one to result of 1's complement + long val, carry = 1; + int index = 0; + + while (carry != 0 && index < maxLength) + { + val = (long)(result.data[index]); + val++; + + result.data[index] = (uint)(val & 0xFFFFFFFF); + carry = val >> 32; + + index++; + } + + if ((bi1.data[maxLength - 1] & 0x80000000) == (result.data[maxLength - 1] & 0x80000000)) + throw (new ArithmeticException("Overflow in negation.\n")); + + result.dataLength = maxLength; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + return result; + } + + + //*********************************************************************** + // Overloading of equality operator + //*********************************************************************** + + public static bool operator ==(BigInteger bi1, BigInteger bi2) + { + return bi1.Equals(bi2); + } + + + public static bool operator !=(BigInteger bi1, BigInteger bi2) + { + return !(bi1.Equals(bi2)); + } + + + public override bool Equals(object o) + { + BigInteger bi = (BigInteger)o; + + if (this.dataLength != bi.dataLength) + return false; + + for (int i = 0; i < this.dataLength; i++) + { + if (this.data[i] != bi.data[i]) + return false; + } + return true; + } + + + public override int GetHashCode() + { + return this.ToString().GetHashCode(); + } + + + //*********************************************************************** + // Overloading of inequality operator + //*********************************************************************** + + public static bool operator >(BigInteger bi1, BigInteger bi2) + { + int pos = maxLength - 1; + + // bi1 is negative, bi2 is positive + if ((bi1.data[pos] & 0x80000000) != 0 && (bi2.data[pos] & 0x80000000) == 0) + return false; + + // bi1 is positive, bi2 is negative + else if ((bi1.data[pos] & 0x80000000) == 0 && (bi2.data[pos] & 0x80000000) != 0) + return true; + + // same sign + int len = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + for (pos = len - 1; pos >= 0 && bi1.data[pos] == bi2.data[pos]; pos--) ; + + if (pos >= 0) + { + if (bi1.data[pos] > bi2.data[pos]) + return true; + return false; + } + return false; + } + + + public static bool operator <(BigInteger bi1, BigInteger bi2) + { + int pos = maxLength - 1; + + // bi1 is negative, bi2 is positive + if ((bi1.data[pos] & 0x80000000) != 0 && (bi2.data[pos] & 0x80000000) == 0) + return true; + + // bi1 is positive, bi2 is negative + else if ((bi1.data[pos] & 0x80000000) == 0 && (bi2.data[pos] & 0x80000000) != 0) + return false; + + // same sign + int len = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + for (pos = len - 1; pos >= 0 && bi1.data[pos] == bi2.data[pos]; pos--) ; + + if (pos >= 0) + { + if (bi1.data[pos] < bi2.data[pos]) + return true; + return false; + } + return false; + } + + + public static bool operator >=(BigInteger bi1, BigInteger bi2) + { + return (bi1 == bi2 || bi1 > bi2); + } + + + public static bool operator <=(BigInteger bi1, BigInteger bi2) + { + return (bi1 == bi2 || bi1 < bi2); + } + + + //*********************************************************************** + // Private function that supports the division of two numbers with + // a divisor that has more than 1 digit. + // + // Algorithm taken from [1] + //*********************************************************************** + + private static void multiByteDivide(BigInteger bi1, BigInteger bi2, + BigInteger outQuotient, BigInteger outRemainder) + { + uint[] result = new uint[maxLength]; + + int remainderLen = bi1.dataLength + 1; + uint[] remainder = new uint[remainderLen]; + + uint mask = 0x80000000; + uint val = bi2.data[bi2.dataLength - 1]; + int shift = 0, resultPos = 0; + + while (mask != 0 && (val & mask) == 0) + { + shift++; mask >>= 1; + } + + //Console.WriteLine("shift = {0}", shift); + //Console.WriteLine("Before bi1 Len = {0}, bi2 Len = {1}", bi1.dataLength, bi2.dataLength); + + for (int i = 0; i < bi1.dataLength; i++) + remainder[i] = bi1.data[i]; + shiftLeft(remainder, shift); + bi2 = bi2 << shift; + + /* + Console.WriteLine("bi1 Len = {0}, bi2 Len = {1}", bi1.dataLength, bi2.dataLength); + Console.WriteLine("dividend = " + bi1 + "\ndivisor = " + bi2); + for(int q = remainderLen - 1; q >= 0; q--) + Console.Write("{0:x2}", remainder[q]); + Console.WriteLine(); + */ + + int j = remainderLen - bi2.dataLength; + int pos = remainderLen - 1; + + ulong firstDivisorByte = bi2.data[bi2.dataLength - 1]; + ulong secondDivisorByte = bi2.data[bi2.dataLength - 2]; + + int divisorLen = bi2.dataLength + 1; + uint[] dividendPart = new uint[divisorLen]; + + while (j > 0) + { + ulong dividend = ((ulong)remainder[pos] << 32) + (ulong)remainder[pos - 1]; + //Console.WriteLine("dividend = {0}", dividend); + + ulong q_hat = dividend / firstDivisorByte; + ulong r_hat = dividend % firstDivisorByte; + + //Console.WriteLine("q_hat = {0:X}, r_hat = {1:X}", q_hat, r_hat); + + bool done = false; + while (!done) + { + done = true; + + if (q_hat == 0x100000000 || + (q_hat * secondDivisorByte) > ((r_hat << 32) + remainder[pos - 2])) + { + q_hat--; + r_hat += firstDivisorByte; + + if (r_hat < 0x100000000) + done = false; + } + } + + for (int h = 0; h < divisorLen; h++) + dividendPart[h] = remainder[pos - h]; + + BigInteger kk = new BigInteger(dividendPart); + BigInteger ss = bi2 * (long)q_hat; + + //Console.WriteLine("ss before = " + ss); + while (ss > kk) + { + q_hat--; + ss -= bi2; + //Console.WriteLine(ss); + } + BigInteger yy = kk - ss; + + //Console.WriteLine("ss = " + ss); + //Console.WriteLine("kk = " + kk); + //Console.WriteLine("yy = " + yy); + + for (int h = 0; h < divisorLen; h++) + remainder[pos - h] = yy.data[bi2.dataLength - h]; + + /* + Console.WriteLine("dividend = "); + for(int q = remainderLen - 1; q >= 0; q--) + Console.Write("{0:x2}", remainder[q]); + Console.WriteLine("\n************ q_hat = {0:X}\n", q_hat); + */ + + result[resultPos++] = (uint)q_hat; + + pos--; + j--; + } + + outQuotient.dataLength = resultPos; + int y = 0; + for (int x = outQuotient.dataLength - 1; x >= 0; x--, y++) + outQuotient.data[y] = result[x]; + for (; y < maxLength; y++) + outQuotient.data[y] = 0; + + while (outQuotient.dataLength > 1 && outQuotient.data[outQuotient.dataLength - 1] == 0) + outQuotient.dataLength--; + + if (outQuotient.dataLength == 0) + outQuotient.dataLength = 1; + + outRemainder.dataLength = shiftRight(remainder, shift); + + for (y = 0; y < outRemainder.dataLength; y++) + outRemainder.data[y] = remainder[y]; + for (; y < maxLength; y++) + outRemainder.data[y] = 0; + } + + + //*********************************************************************** + // Private function that supports the division of two numbers with + // a divisor that has only 1 digit. + //*********************************************************************** + + private static void singleByteDivide(BigInteger bi1, BigInteger bi2, + BigInteger outQuotient, BigInteger outRemainder) + { + uint[] result = new uint[maxLength]; + int resultPos = 0; + + // copy dividend to reminder + for (int i = 0; i < maxLength; i++) + outRemainder.data[i] = bi1.data[i]; + outRemainder.dataLength = bi1.dataLength; + + while (outRemainder.dataLength > 1 && outRemainder.data[outRemainder.dataLength - 1] == 0) + outRemainder.dataLength--; + + ulong divisor = (ulong)bi2.data[0]; + int pos = outRemainder.dataLength - 1; + ulong dividend = (ulong)outRemainder.data[pos]; + + //Console.WriteLine("divisor = " + divisor + " dividend = " + dividend); + //Console.WriteLine("divisor = " + bi2 + "\ndividend = " + bi1); + + if (dividend >= divisor) + { + ulong quotient = dividend / divisor; + result[resultPos++] = (uint)quotient; + + outRemainder.data[pos] = (uint)(dividend % divisor); + } + pos--; + + while (pos >= 0) + { + //Console.WriteLine(pos); + + dividend = ((ulong)outRemainder.data[pos + 1] << 32) + (ulong)outRemainder.data[pos]; + ulong quotient = dividend / divisor; + result[resultPos++] = (uint)quotient; + + outRemainder.data[pos + 1] = 0; + outRemainder.data[pos--] = (uint)(dividend % divisor); + //Console.WriteLine(">>>> " + bi1); + } + + outQuotient.dataLength = resultPos; + int j = 0; + for (int i = outQuotient.dataLength - 1; i >= 0; i--, j++) + outQuotient.data[j] = result[i]; + for (; j < maxLength; j++) + outQuotient.data[j] = 0; + + while (outQuotient.dataLength > 1 && outQuotient.data[outQuotient.dataLength - 1] == 0) + outQuotient.dataLength--; + + if (outQuotient.dataLength == 0) + outQuotient.dataLength = 1; + + while (outRemainder.dataLength > 1 && outRemainder.data[outRemainder.dataLength - 1] == 0) + outRemainder.dataLength--; + } + + + //*********************************************************************** + // Overloading of division operator + //*********************************************************************** + + public static BigInteger operator /(BigInteger bi1, BigInteger bi2) + { + BigInteger quotient = new BigInteger(); + BigInteger remainder = new BigInteger(); + + int lastPos = maxLength - 1; + bool divisorNeg = false, dividendNeg = false; + + if ((bi1.data[lastPos] & 0x80000000) != 0) // bi1 negative + { + bi1 = -bi1; + dividendNeg = true; + } + if ((bi2.data[lastPos] & 0x80000000) != 0) // bi2 negative + { + bi2 = -bi2; + divisorNeg = true; + } + + if (bi1 < bi2) + { + return quotient; + } + + else + { + if (bi2.dataLength == 1) + singleByteDivide(bi1, bi2, quotient, remainder); + else + multiByteDivide(bi1, bi2, quotient, remainder); + + if (dividendNeg != divisorNeg) + return -quotient; + + return quotient; + } + } + + + //*********************************************************************** + // Overloading of modulus operator + //*********************************************************************** + + public static BigInteger operator %(BigInteger bi1, BigInteger bi2) + { + BigInteger quotient = new BigInteger(); + BigInteger remainder = new BigInteger(bi1); + + int lastPos = maxLength - 1; + bool dividendNeg = false; + + if ((bi1.data[lastPos] & 0x80000000) != 0) // bi1 negative + { + bi1 = -bi1; + dividendNeg = true; + } + if ((bi2.data[lastPos] & 0x80000000) != 0) // bi2 negative + bi2 = -bi2; + + if (bi1 < bi2) + { + return remainder; + } + + else + { + if (bi2.dataLength == 1) + singleByteDivide(bi1, bi2, quotient, remainder); + else + multiByteDivide(bi1, bi2, quotient, remainder); + + if (dividendNeg) + return -remainder; + + return remainder; + } + } + + + //*********************************************************************** + // Overloading of bitwise AND operator + //*********************************************************************** + + public static BigInteger operator &(BigInteger bi1, BigInteger bi2) + { + BigInteger result = new BigInteger(); + + int len = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + + for (int i = 0; i < len; i++) + { + uint sum = (uint)(bi1.data[i] & bi2.data[i]); + result.data[i] = sum; + } + + result.dataLength = maxLength; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + return result; + } + + + //*********************************************************************** + // Overloading of bitwise OR operator + //*********************************************************************** + + public static BigInteger operator |(BigInteger bi1, BigInteger bi2) + { + BigInteger result = new BigInteger(); + + int len = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + + for (int i = 0; i < len; i++) + { + uint sum = (uint)(bi1.data[i] | bi2.data[i]); + result.data[i] = sum; + } + + result.dataLength = maxLength; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + return result; + } + + + //*********************************************************************** + // Overloading of bitwise XOR operator + //*********************************************************************** + + public static BigInteger operator ^(BigInteger bi1, BigInteger bi2) + { + BigInteger result = new BigInteger(); + + int len = (bi1.dataLength > bi2.dataLength) ? bi1.dataLength : bi2.dataLength; + + for (int i = 0; i < len; i++) + { + uint sum = (uint)(bi1.data[i] ^ bi2.data[i]); + result.data[i] = sum; + } + + result.dataLength = maxLength; + + while (result.dataLength > 1 && result.data[result.dataLength - 1] == 0) + result.dataLength--; + + return result; + } + + + //*********************************************************************** + // Returns max(this, bi) + //*********************************************************************** + + public BigInteger max(BigInteger bi) + { + if (this > bi) + return (new BigInteger(this)); + else + return (new BigInteger(bi)); + } + + + //*********************************************************************** + // Returns min(this, bi) + //*********************************************************************** + + public BigInteger min(BigInteger bi) + { + if (this < bi) + return (new BigInteger(this)); + else + return (new BigInteger(bi)); + + } + + + //*********************************************************************** + // Returns the absolute value + //*********************************************************************** + + public BigInteger abs() + { + if ((this.data[maxLength - 1] & 0x80000000) != 0) + return (-this); + else + return (new BigInteger(this)); + } + + + //*********************************************************************** + // Returns a string representing the BigInteger in base 10. + //*********************************************************************** + + public override string ToString() + { + return ToString(10); + } + + + //*********************************************************************** + // Returns a string representing the BigInteger in sign-and-magnitude + // format in the specified radix. + // + // Example + // ------- + // If the value of BigInteger is -255 in base 10, then + // ToString(16) returns "-FF" + // + //*********************************************************************** + + public string ToString(int radix) + { + if (radix < 2 || radix > 36) + throw (new ArgumentException("Radix must be >= 2 and <= 36")); + + string charSet = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"; + string result = ""; + + BigInteger a = this; + + bool negative = false; + if ((a.data[maxLength - 1] & 0x80000000) != 0) + { + negative = true; + try + { + a = -a; + } + catch (Exception) { } + } + + BigInteger quotient = new BigInteger(); + BigInteger remainder = new BigInteger(); + BigInteger biRadix = new BigInteger(radix); + + if (a.dataLength == 1 && a.data[0] == 0) + result = "0"; + else + { + while (a.dataLength > 1 || (a.dataLength == 1 && a.data[0] != 0)) + { + singleByteDivide(a, biRadix, quotient, remainder); + + if (remainder.data[0] < 10) + result = remainder.data[0] + result; + else + result = charSet[(int)remainder.data[0] - 10] + result; + + a = quotient; + } + if (negative) + result = "-" + result; + } + + return result; + } + + + //*********************************************************************** + // Returns a hex string showing the contains of the BigInteger + // + // Examples + // ------- + // 1) If the value of BigInteger is 255 in base 10, then + // ToHexString() returns "FF" + // + // 2) If the value of BigInteger is -255 in base 10, then + // ToHexString() returns ".....FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF01", + // which is the 2's complement representation of -255. + // + //*********************************************************************** + + public string ToHexString() + { + string result = data[dataLength - 1].ToString("X"); + + for (int i = dataLength - 2; i >= 0; i--) + { + result += data[i].ToString("X8"); + } + + return result; + } + + + + //*********************************************************************** + // Modulo Exponentiation + //*********************************************************************** + + public BigInteger modPow(BigInteger exp, BigInteger n) + { + if ((exp.data[maxLength - 1] & 0x80000000) != 0) + throw (new ArithmeticException("Positive exponents only.")); + + BigInteger resultNum = 1; + BigInteger tempNum; + bool thisNegative = false; + + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative this + { + tempNum = -this % n; + thisNegative = true; + } + else + tempNum = this % n; // ensures (tempNum * tempNum) < b^(2k) + + if ((n.data[maxLength - 1] & 0x80000000) != 0) // negative n + n = -n; + + // calculate constant = b^(2k) / m + BigInteger constant = new BigInteger(); + + int i = n.dataLength << 1; + constant.data[i] = 0x00000001; + constant.dataLength = i + 1; + + constant = constant / n; + int totalBits = exp.bitCount(); + int count = 0; + + // perform squaring and multiply exponentiation + for (int pos = 0; pos < exp.dataLength; pos++) + { + uint mask = 0x01; + //Console.WriteLine("pos = " + pos); + + for (int index = 0; index < 32; index++) + { + if ((exp.data[pos] & mask) != 0) + resultNum = BarrettReduction(resultNum * tempNum, n, constant); + + mask <<= 1; + + tempNum = BarrettReduction(tempNum * tempNum, n, constant); + + + if (tempNum.dataLength == 1 && tempNum.data[0] == 1) + { + if (thisNegative && (exp.data[0] & 0x1) != 0) //odd exp + return -resultNum; + return resultNum; + } + count++; + if (count == totalBits) + break; + } + } + + if (thisNegative && (exp.data[0] & 0x1) != 0) //odd exp + return -resultNum; + + return resultNum; + } + + + + //*********************************************************************** + // Fast calculation of modular reduction using Barrett's reduction. + // Requires x < b^(2k), where b is the base. In this case, base is + // 2^32 (uint). + // + // Reference [4] + //*********************************************************************** + + private BigInteger BarrettReduction(BigInteger x, BigInteger n, BigInteger constant) + { + int k = n.dataLength, + kPlusOne = k + 1, + kMinusOne = k - 1; + + BigInteger q1 = new BigInteger(); + + // q1 = x / b^(k-1) + for (int i = kMinusOne, j = 0; i < x.dataLength; i++, j++) + q1.data[j] = x.data[i]; + q1.dataLength = x.dataLength - kMinusOne; + if (q1.dataLength <= 0) + q1.dataLength = 1; + + + BigInteger q2 = q1 * constant; + BigInteger q3 = new BigInteger(); + + // q3 = q2 / b^(k+1) + for (int i = kPlusOne, j = 0; i < q2.dataLength; i++, j++) + q3.data[j] = q2.data[i]; + q3.dataLength = q2.dataLength - kPlusOne; + if (q3.dataLength <= 0) + q3.dataLength = 1; + + + // r1 = x mod b^(k+1) + // i.e. keep the lowest (k+1) words + BigInteger r1 = new BigInteger(); + int lengthToCopy = (x.dataLength > kPlusOne) ? kPlusOne : x.dataLength; + for (int i = 0; i < lengthToCopy; i++) + r1.data[i] = x.data[i]; + r1.dataLength = lengthToCopy; + + + // r2 = (q3 * n) mod b^(k+1) + // partial multiplication of q3 and n + + BigInteger r2 = new BigInteger(); + for (int i = 0; i < q3.dataLength; i++) + { + if (q3.data[i] == 0) continue; + + ulong mcarry = 0; + int t = i; + for (int j = 0; j < n.dataLength && t < kPlusOne; j++, t++) + { + // t = i + j + ulong val = ((ulong)q3.data[i] * (ulong)n.data[j]) + + (ulong)r2.data[t] + mcarry; + + r2.data[t] = (uint)(val & 0xFFFFFFFF); + mcarry = (val >> 32); + } + + if (t < kPlusOne) + r2.data[t] = (uint)mcarry; + } + r2.dataLength = kPlusOne; + while (r2.dataLength > 1 && r2.data[r2.dataLength - 1] == 0) + r2.dataLength--; + + r1 -= r2; + if ((r1.data[maxLength - 1] & 0x80000000) != 0) // negative + { + BigInteger val = new BigInteger(); + val.data[kPlusOne] = 0x00000001; + val.dataLength = kPlusOne + 1; + r1 += val; + } + + while (r1 >= n) + r1 -= n; + + return r1; + } + + + //*********************************************************************** + // Returns gcd(this, bi) + //*********************************************************************** + + public BigInteger gcd(BigInteger bi) + { + BigInteger x; + BigInteger y; + + if ((data[maxLength - 1] & 0x80000000) != 0) // negative + x = -this; + else + x = this; + + if ((bi.data[maxLength - 1] & 0x80000000) != 0) // negative + y = -bi; + else + y = bi; + + BigInteger g = y; + + while (x.dataLength > 1 || (x.dataLength == 1 && x.data[0] != 0)) + { + g = x; + x = y % x; + y = g; + } + + return g; + } + + + //*********************************************************************** + // Populates "this" with the specified amount of random bits + //*********************************************************************** + + public void genRandomBits(int bits, Random rand) + { + int dwords = bits >> 5; + int remBits = bits & 0x1F; + + if (remBits != 0) + dwords++; + + if (dwords > maxLength) + throw (new ArithmeticException("Number of required bits > maxLength.")); + + for (int i = 0; i < dwords; i++) + data[i] = (uint)(rand.NextDouble() * 0x100000000); + + for (int i = dwords; i < maxLength; i++) + data[i] = 0; + + if (remBits != 0) + { + uint mask = (uint)(0x01 << (remBits - 1)); + data[dwords - 1] |= mask; + + mask = (uint)(0xFFFFFFFF >> (32 - remBits)); + data[dwords - 1] &= mask; + } + else + data[dwords - 1] |= 0x80000000; + + dataLength = dwords; + + if (dataLength == 0) + dataLength = 1; + } + + + //*********************************************************************** + // Returns the position of the most significant bit in the BigInteger. + // + // Eg. The result is 0, if the value of BigInteger is 0...0000 0000 + // The result is 1, if the value of BigInteger is 0...0000 0001 + // The result is 2, if the value of BigInteger is 0...0000 0010 + // The result is 2, if the value of BigInteger is 0...0000 0011 + // + //*********************************************************************** + + public int bitCount() + { + while (dataLength > 1 && data[dataLength - 1] == 0) + dataLength--; + + uint value = data[dataLength - 1]; + uint mask = 0x80000000; + int bits = 32; + + while (bits > 0 && (value & mask) == 0) + { + bits--; + mask >>= 1; + } + bits += ((dataLength - 1) << 5); + + return bits; + } + + + //*********************************************************************** + // Probabilistic prime test based on Fermat's little theorem + // + // for any a < p (p does not divide a) if + // a^(p-1) mod p != 1 then p is not prime. + // + // Otherwise, p is probably prime (pseudoprime to the chosen base). + // + // Returns + // ------- + // True if "this" is a pseudoprime to randomly chosen + // bases. The number of chosen bases is given by the "confidence" + // parameter. + // + // False if "this" is definitely NOT prime. + // + // Note - this method is fast but fails for Carmichael numbers except + // when the randomly chosen base is a factor of the number. + // + //*********************************************************************** + + public bool FermatLittleTest(int confidence) + { + BigInteger thisVal; + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative + thisVal = -this; + else + thisVal = this; + + if (thisVal.dataLength == 1) + { + // test small numbers + if (thisVal.data[0] == 0 || thisVal.data[0] == 1) + return false; + else if (thisVal.data[0] == 2 || thisVal.data[0] == 3) + return true; + } + + if ((thisVal.data[0] & 0x1) == 0) // even numbers + return false; + + int bits = thisVal.bitCount(); + BigInteger a = new BigInteger(); + BigInteger p_sub1 = thisVal - (new BigInteger(1)); + Random rand = new Random(); + + for (int round = 0; round < confidence; round++) + { + bool done = false; + + while (!done) // generate a < n + { + int testBits = 0; + + // make sure "a" has at least 2 bits + while (testBits < 2) + testBits = (int)(rand.NextDouble() * bits); + + a.genRandomBits(testBits, rand); + + int byteLen = a.dataLength; + + // make sure "a" is not 0 + if (byteLen > 1 || (byteLen == 1 && a.data[0] != 1)) + done = true; + } + + // check whether a factor exists (fix for version 1.03) + BigInteger gcdTest = a.gcd(thisVal); + if (gcdTest.dataLength == 1 && gcdTest.data[0] != 1) + return false; + + // calculate a^(p-1) mod p + BigInteger expResult = a.modPow(p_sub1, thisVal); + + int resultLen = expResult.dataLength; + + // is NOT prime is a^(p-1) mod p != 1 + + if (resultLen > 1 || (resultLen == 1 && expResult.data[0] != 1)) + { + //Console.WriteLine("a = " + a.ToString()); + return false; + } + } + + return true; + } + + + //*********************************************************************** + // Probabilistic prime test based on Rabin-Miller's + // + // for any p > 0 with p - 1 = 2^s * t + // + // p is probably prime (strong pseudoprime) if for any a < p, + // 1) a^t mod p = 1 or + // 2) a^((2^j)*t) mod p = p-1 for some 0 <= j <= s-1 + // + // Otherwise, p is composite. + // + // Returns + // ------- + // True if "this" is a strong pseudoprime to randomly chosen + // bases. The number of chosen bases is given by the "confidence" + // parameter. + // + // False if "this" is definitely NOT prime. + // + //*********************************************************************** + + public bool RabinMillerTest(int confidence) + { + BigInteger thisVal; + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative + thisVal = -this; + else + thisVal = this; + + if (thisVal.dataLength == 1) + { + // test small numbers + if (thisVal.data[0] == 0 || thisVal.data[0] == 1) + return false; + else if (thisVal.data[0] == 2 || thisVal.data[0] == 3) + return true; + } + + if ((thisVal.data[0] & 0x1) == 0) // even numbers + return false; + + + // calculate values of s and t + BigInteger p_sub1 = thisVal - (new BigInteger(1)); + int s = 0; + + for (int index = 0; index < p_sub1.dataLength; index++) + { + uint mask = 0x01; + + for (int i = 0; i < 32; i++) + { + if ((p_sub1.data[index] & mask) != 0) + { + index = p_sub1.dataLength; // to break the outer loop + break; + } + mask <<= 1; + s++; + } + } + + BigInteger t = p_sub1 >> s; + + int bits = thisVal.bitCount(); + BigInteger a = new BigInteger(); + Random rand = new Random(); + + for (int round = 0; round < confidence; round++) + { + bool done = false; + + while (!done) // generate a < n + { + int testBits = 0; + + // make sure "a" has at least 2 bits + while (testBits < 2) + testBits = (int)(rand.NextDouble() * bits); + + a.genRandomBits(testBits, rand); + + int byteLen = a.dataLength; + + // make sure "a" is not 0 + if (byteLen > 1 || (byteLen == 1 && a.data[0] != 1)) + done = true; + } + + // check whether a factor exists (fix for version 1.03) + BigInteger gcdTest = a.gcd(thisVal); + if (gcdTest.dataLength == 1 && gcdTest.data[0] != 1) + return false; + + BigInteger b = a.modPow(t, thisVal); + + /* + Console.WriteLine("a = " + a.ToString(10)); + Console.WriteLine("b = " + b.ToString(10)); + Console.WriteLine("t = " + t.ToString(10)); + Console.WriteLine("s = " + s); + */ + + bool result = false; + + if (b.dataLength == 1 && b.data[0] == 1) // a^t mod p = 1 + result = true; + + for (int j = 0; result == false && j < s; j++) + { + if (b == p_sub1) // a^((2^j)*t) mod p = p-1 for some 0 <= j <= s-1 + { + result = true; + break; + } + + b = (b * b) % thisVal; + } + + if (result == false) + return false; + } + return true; + } + + + //*********************************************************************** + // Probabilistic prime test based on Solovay-Strassen (Euler Criterion) + // + // p is probably prime if for any a < p (a is not multiple of p), + // a^((p-1)/2) mod p = J(a, p) + // + // where J is the Jacobi symbol. + // + // Otherwise, p is composite. + // + // Returns + // ------- + // True if "this" is a Euler pseudoprime to randomly chosen + // bases. The number of chosen bases is given by the "confidence" + // parameter. + // + // False if "this" is definitely NOT prime. + // + //*********************************************************************** + + public bool SolovayStrassenTest(int confidence) + { + BigInteger thisVal; + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative + thisVal = -this; + else + thisVal = this; + + if (thisVal.dataLength == 1) + { + // test small numbers + if (thisVal.data[0] == 0 || thisVal.data[0] == 1) + return false; + else if (thisVal.data[0] == 2 || thisVal.data[0] == 3) + return true; + } + + if ((thisVal.data[0] & 0x1) == 0) // even numbers + return false; + + + int bits = thisVal.bitCount(); + BigInteger a = new BigInteger(); + BigInteger p_sub1 = thisVal - 1; + BigInteger p_sub1_shift = p_sub1 >> 1; + + Random rand = new Random(); + + for (int round = 0; round < confidence; round++) + { + bool done = false; + + while (!done) // generate a < n + { + int testBits = 0; + + // make sure "a" has at least 2 bits + while (testBits < 2) + testBits = (int)(rand.NextDouble() * bits); + + a.genRandomBits(testBits, rand); + + int byteLen = a.dataLength; + + // make sure "a" is not 0 + if (byteLen > 1 || (byteLen == 1 && a.data[0] != 1)) + done = true; + } + + // check whether a factor exists (fix for version 1.03) + BigInteger gcdTest = a.gcd(thisVal); + if (gcdTest.dataLength == 1 && gcdTest.data[0] != 1) + return false; + + // calculate a^((p-1)/2) mod p + + BigInteger expResult = a.modPow(p_sub1_shift, thisVal); + if (expResult == p_sub1) + expResult = -1; + + // calculate Jacobi symbol + BigInteger jacob = Jacobi(a, thisVal); + + //Console.WriteLine("a = " + a.ToString(10) + " b = " + thisVal.ToString(10)); + //Console.WriteLine("expResult = " + expResult.ToString(10) + " Jacob = " + jacob.ToString(10)); + + // if they are different then it is not prime + if (expResult != jacob) + return false; + } + + return true; + } + + + //*********************************************************************** + // Implementation of the Lucas Strong Pseudo Prime test. + // + // Let n be an odd number with gcd(n,D) = 1, and n - J(D, n) = 2^s * d + // with d odd and s >= 0. + // + // If Ud mod n = 0 or V2^r*d mod n = 0 for some 0 <= r < s, then n + // is a strong Lucas pseudoprime with parameters (P, Q). We select + // P and Q based on Selfridge. + // + // Returns True if number is a strong Lucus pseudo prime. + // Otherwise, returns False indicating that number is composite. + //*********************************************************************** + + public bool LucasStrongTest() + { + BigInteger thisVal; + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative + thisVal = -this; + else + thisVal = this; + + if (thisVal.dataLength == 1) + { + // test small numbers + if (thisVal.data[0] == 0 || thisVal.data[0] == 1) + return false; + else if (thisVal.data[0] == 2 || thisVal.data[0] == 3) + return true; + } + + if ((thisVal.data[0] & 0x1) == 0) // even numbers + return false; + + return LucasStrongTestHelper(thisVal); + } + + + private bool LucasStrongTestHelper(BigInteger thisVal) + { + // Do the test (selects D based on Selfridge) + // Let D be the first element of the sequence + // 5, -7, 9, -11, 13, ... for which J(D,n) = -1 + // Let P = 1, Q = (1-D) / 4 + + long D = 5, sign = -1, dCount = 0; + bool done = false; + + while (!done) + { + int Jresult = BigInteger.Jacobi(D, thisVal); + + if (Jresult == -1) + done = true; // J(D, this) = 1 + else + { + if (Jresult == 0 && Math.Abs(D) < thisVal) // divisor found + return false; + + if (dCount == 20) + { + // check for square + BigInteger root = thisVal.sqrt(); + if (root * root == thisVal) + return false; + } + + //Console.WriteLine(D); + D = (Math.Abs(D) + 2) * sign; + sign = -sign; + } + dCount++; + } + + long Q = (1 - D) >> 2; + + /* + Console.WriteLine("D = " + D); + Console.WriteLine("Q = " + Q); + Console.WriteLine("(n,D) = " + thisVal.gcd(D)); + Console.WriteLine("(n,Q) = " + thisVal.gcd(Q)); + Console.WriteLine("J(D|n) = " + BigInteger.Jacobi(D, thisVal)); + */ + + BigInteger p_add1 = thisVal + 1; + int s = 0; + + for (int index = 0; index < p_add1.dataLength; index++) + { + uint mask = 0x01; + + for (int i = 0; i < 32; i++) + { + if ((p_add1.data[index] & mask) != 0) + { + index = p_add1.dataLength; // to break the outer loop + break; + } + mask <<= 1; + s++; + } + } + + BigInteger t = p_add1 >> s; + + // calculate constant = b^(2k) / m + // for Barrett Reduction + BigInteger constant = new BigInteger(); + + int nLen = thisVal.dataLength << 1; + constant.data[nLen] = 0x00000001; + constant.dataLength = nLen + 1; + + constant = constant / thisVal; + + BigInteger[] lucas = LucasSequenceHelper(1, Q, t, thisVal, constant, 0); + bool isPrime = false; + + if ((lucas[0].dataLength == 1 && lucas[0].data[0] == 0) || + (lucas[1].dataLength == 1 && lucas[1].data[0] == 0)) + { + // u(t) = 0 or V(t) = 0 + isPrime = true; + } + + for (int i = 1; i < s; i++) + { + if (!isPrime) + { + // doubling of index + lucas[1] = thisVal.BarrettReduction(lucas[1] * lucas[1], thisVal, constant); + lucas[1] = (lucas[1] - (lucas[2] << 1)) % thisVal; + + //lucas[1] = ((lucas[1] * lucas[1]) - (lucas[2] << 1)) % thisVal; + + if ((lucas[1].dataLength == 1 && lucas[1].data[0] == 0)) + isPrime = true; + } + + lucas[2] = thisVal.BarrettReduction(lucas[2] * lucas[2], thisVal, constant); //Q^k + } + + + if (isPrime) // additional checks for composite numbers + { + // If n is prime and gcd(n, Q) == 1, then + // Q^((n+1)/2) = Q * Q^((n-1)/2) is congruent to (Q * J(Q, n)) mod n + + BigInteger g = thisVal.gcd(Q); + if (g.dataLength == 1 && g.data[0] == 1) // gcd(this, Q) == 1 + { + if ((lucas[2].data[maxLength - 1] & 0x80000000) != 0) + lucas[2] += thisVal; + + BigInteger temp = (Q * BigInteger.Jacobi(Q, thisVal)) % thisVal; + if ((temp.data[maxLength - 1] & 0x80000000) != 0) + temp += thisVal; + + if (lucas[2] != temp) + isPrime = false; + } + } + + return isPrime; + } + + + //*********************************************************************** + // Determines whether a number is probably prime, using the Rabin-Miller's + // test. Before applying the test, the number is tested for divisibility + // by primes < 2000 + // + // Returns true if number is probably prime. + //*********************************************************************** + + public bool isProbablePrime(int confidence) + { + BigInteger thisVal; + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative + thisVal = -this; + else + thisVal = this; + + + // test for divisibility by primes < 2000 + for (int p = 0; p < primesBelow2000.Length; p++) + { + BigInteger divisor = primesBelow2000[p]; + + if (divisor >= thisVal) + break; + + BigInteger resultNum = thisVal % divisor; + if (resultNum.IntValue() == 0) + { + /* + Console.WriteLine("Not prime! Divisible by {0}\n", + primesBelow2000[p]); + */ + return false; + } + } + + if (thisVal.RabinMillerTest(confidence)) + return true; + else + { + //Console.WriteLine("Not prime! Failed primality test\n"); + return false; + } + } + + + //*********************************************************************** + // Determines whether this BigInteger is probably prime using a + // combination of base 2 strong pseudoprime test and Lucas strong + // pseudoprime test. + // + // The sequence of the primality test is as follows, + // + // 1) Trial divisions are carried out using prime numbers below 2000. + // if any of the primes divides this BigInteger, then it is not prime. + // + // 2) Perform base 2 strong pseudoprime test. If this BigInteger is a + // base 2 strong pseudoprime, proceed on to the next step. + // + // 3) Perform strong Lucas pseudoprime test. + // + // Returns True if this BigInteger is both a base 2 strong pseudoprime + // and a strong Lucas pseudoprime. + // + // For a detailed discussion of this primality test, see [6]. + // + //*********************************************************************** + + public bool isProbablePrime() + { + BigInteger thisVal; + if ((this.data[maxLength - 1] & 0x80000000) != 0) // negative + thisVal = -this; + else + thisVal = this; + + if (thisVal.dataLength == 1) + { + // test small numbers + if (thisVal.data[0] == 0 || thisVal.data[0] == 1) + return false; + else if (thisVal.data[0] == 2 || thisVal.data[0] == 3) + return true; + } + + if ((thisVal.data[0] & 0x1) == 0) // even numbers + return false; + + + // test for divisibility by primes < 2000 + for (int p = 0; p < primesBelow2000.Length; p++) + { + BigInteger divisor = primesBelow2000[p]; + + if (divisor >= thisVal) + break; + + BigInteger resultNum = thisVal % divisor; + if (resultNum.IntValue() == 0) + { + //Console.WriteLine("Not prime! Divisible by {0}\n", + // primesBelow2000[p]); + + return false; + } + } + + // Perform BASE 2 Rabin-Miller Test + + // calculate values of s and t + BigInteger p_sub1 = thisVal - (new BigInteger(1)); + int s = 0; + + for (int index = 0; index < p_sub1.dataLength; index++) + { + uint mask = 0x01; + + for (int i = 0; i < 32; i++) + { + if ((p_sub1.data[index] & mask) != 0) + { + index = p_sub1.dataLength; // to break the outer loop + break; + } + mask <<= 1; + s++; + } + } + + BigInteger t = p_sub1 >> s; + + int bits = thisVal.bitCount(); + BigInteger a = 2; + + // b = a^t mod p + BigInteger b = a.modPow(t, thisVal); + bool result = false; + + if (b.dataLength == 1 && b.data[0] == 1) // a^t mod p = 1 + result = true; + + for (int j = 0; result == false && j < s; j++) + { + if (b == p_sub1) // a^((2^j)*t) mod p = p-1 for some 0 <= j <= s-1 + { + result = true; + break; + } + + b = (b * b) % thisVal; + } + + // if number is strong pseudoprime to base 2, then do a strong lucas test + if (result) + result = LucasStrongTestHelper(thisVal); + + return result; + } + + + + //*********************************************************************** + // Returns the lowest 4 bytes of the BigInteger as an int. + //*********************************************************************** + + public int IntValue() + { + return (int)data[0]; + } + + + //*********************************************************************** + // Returns the lowest 8 bytes of the BigInteger as a long. + //*********************************************************************** + + public long LongValue() + { + long val = 0; + + val = (long)data[0]; + try + { // exception if maxLength = 1 + val |= (long)data[1] << 32; + } + catch (Exception) + { + if ((data[0] & 0x80000000) != 0) // negative + val = (int)data[0]; + } + + return val; + } + + + //*********************************************************************** + // Computes the Jacobi Symbol for a and b. + // Algorithm adapted from [3] and [4] with some optimizations + //*********************************************************************** + + public static int Jacobi(BigInteger a, BigInteger b) + { + // Jacobi defined only for odd integers + if ((b.data[0] & 0x1) == 0) + throw (new ArgumentException("Jacobi defined only for odd integers.")); + + if (a >= b) a %= b; + if (a.dataLength == 1 && a.data[0] == 0) return 0; // a == 0 + if (a.dataLength == 1 && a.data[0] == 1) return 1; // a == 1 + + if (a < 0) + { + if ((((b - 1).data[0]) & 0x2) == 0) //if( (((b-1) >> 1).data[0] & 0x1) == 0) + return Jacobi(-a, b); + else + return -Jacobi(-a, b); + } + + int e = 0; + for (int index = 0; index < a.dataLength; index++) + { + uint mask = 0x01; + + for (int i = 0; i < 32; i++) + { + if ((a.data[index] & mask) != 0) + { + index = a.dataLength; // to break the outer loop + break; + } + mask <<= 1; + e++; + } + } + + BigInteger a1 = a >> e; + + int s = 1; + if ((e & 0x1) != 0 && ((b.data[0] & 0x7) == 3 || (b.data[0] & 0x7) == 5)) + s = -1; + + if ((b.data[0] & 0x3) == 3 && (a1.data[0] & 0x3) == 3) + s = -s; + + if (a1.dataLength == 1 && a1.data[0] == 1) + return s; + else + return (s * Jacobi(b % a1, a1)); + } + + + + //*********************************************************************** + // Generates a positive BigInteger that is probably prime. + //*********************************************************************** + + public static BigInteger genPseudoPrime(int bits, int confidence, Random rand) + { + BigInteger result = new BigInteger(); + bool done = false; + + while (!done) + { + result.genRandomBits(bits, rand); + result.data[0] |= 0x01; // make it odd + + // prime test + done = result.isProbablePrime(confidence); + } + return result; + } + + + //*********************************************************************** + // Generates a random number with the specified number of bits such + // that gcd(number, this) = 1 + //*********************************************************************** + + public BigInteger genCoPrime(int bits, Random rand) + { + bool done = false; + BigInteger result = new BigInteger(); + + while (!done) + { + result.genRandomBits(bits, rand); + //Console.WriteLine(result.ToString(16)); + + // gcd test + BigInteger g = result.gcd(this); + if (g.dataLength == 1 && g.data[0] == 1) + done = true; + } + + return result; + } + + + //*********************************************************************** + // Returns the modulo inverse of this. Throws ArithmeticException if + // the inverse does not exist. (i.e. gcd(this, modulus) != 1) + //*********************************************************************** + + public BigInteger modInverse(BigInteger modulus) + { + BigInteger[] p = { 0, 1 }; + BigInteger[] q = new BigInteger[2]; // quotients + BigInteger[] r = { 0, 0 }; // remainders + + int step = 0; + + BigInteger a = modulus; + BigInteger b = this; + + while (b.dataLength > 1 || (b.dataLength == 1 && b.data[0] != 0)) + { + BigInteger quotient = new BigInteger(); + BigInteger remainder = new BigInteger(); + + if (step > 1) + { + BigInteger pval = (p[0] - (p[1] * q[0])) % modulus; + p[0] = p[1]; + p[1] = pval; + } + + if (b.dataLength == 1) + singleByteDivide(a, b, quotient, remainder); + else + multiByteDivide(a, b, quotient, remainder); + + /* + Console.WriteLine(quotient.dataLength); + Console.WriteLine("{0} = {1}({2}) + {3} p = {4}", a.ToString(10), + b.ToString(10), quotient.ToString(10), remainder.ToString(10), + p[1].ToString(10)); + */ + + q[0] = q[1]; + r[0] = r[1]; + q[1] = quotient; r[1] = remainder; + + a = b; + b = remainder; + + step++; + } + + if (r[0].dataLength > 1 || (r[0].dataLength == 1 && r[0].data[0] != 1)) + throw (new ArithmeticException("No inverse!")); + + BigInteger result = ((p[0] - (p[1] * q[0])) % modulus); + + if ((result.data[maxLength - 1] & 0x80000000) != 0) + result += modulus; // get the least positive modulus + + return result; + } + + + //*********************************************************************** + // Returns the value of the BigInteger as a byte array. The lowest + // index contains the MSB. + //*********************************************************************** + + public byte[] getBytes() + { + int numBits = bitCount(); + + int numBytes = numBits >> 3; + if ((numBits & 0x7) != 0) + numBytes++; + + byte[] result = new byte[numBytes]; + + //Console.WriteLine(result.Length); + + int pos = 0; + uint tempVal, val = data[dataLength - 1]; + + if ((tempVal = (val >> 24 & 0xFF)) != 0) + result[pos++] = (byte)tempVal; + if ((tempVal = (val >> 16 & 0xFF)) != 0) + result[pos++] = (byte)tempVal; + if ((tempVal = (val >> 8 & 0xFF)) != 0) + result[pos++] = (byte)tempVal; + if ((tempVal = (val & 0xFF)) != 0) + result[pos++] = (byte)tempVal; + + for (int i = dataLength - 2; i >= 0; i--, pos += 4) + { + val = data[i]; + result[pos + 3] = (byte)(val & 0xFF); + val >>= 8; + result[pos + 2] = (byte)(val & 0xFF); + val >>= 8; + result[pos + 1] = (byte)(val & 0xFF); + val >>= 8; + result[pos] = (byte)(val & 0xFF); + } + + return result; + } + + + //*********************************************************************** + // Sets the value of the specified bit to 1 + // The Least Significant Bit position is 0. + //*********************************************************************** + + public void setBit(uint bitNum) + { + uint bytePos = bitNum >> 5; // divide by 32 + byte bitPos = (byte)(bitNum & 0x1F); // get the lowest 5 bits + + uint mask = (uint)1 << bitPos; + this.data[bytePos] |= mask; + + if (bytePos >= this.dataLength) + this.dataLength = (int)bytePos + 1; + } + + + //*********************************************************************** + // Sets the value of the specified bit to 0 + // The Least Significant Bit position is 0. + //*********************************************************************** + + public void unsetBit(uint bitNum) + { + uint bytePos = bitNum >> 5; + + if (bytePos < this.dataLength) + { + byte bitPos = (byte)(bitNum & 0x1F); + + uint mask = (uint)1 << bitPos; + uint mask2 = 0xFFFFFFFF ^ mask; + + this.data[bytePos] &= mask2; + + if (this.dataLength > 1 && this.data[this.dataLength - 1] == 0) + this.dataLength--; + } + } + + + //*********************************************************************** + // Returns a value that is equivalent to the integer square root + // of the BigInteger. + // + // The integer square root of "this" is defined as the largest integer n + // such that (n * n) <= this + // + //*********************************************************************** + + public BigInteger sqrt() + { + uint numBits = (uint)this.bitCount(); + + if ((numBits & 0x1) != 0) // odd number of bits + numBits = (numBits >> 1) + 1; + else + numBits = (numBits >> 1); + + uint bytePos = numBits >> 5; + byte bitPos = (byte)(numBits & 0x1F); + + uint mask; + + BigInteger result = new BigInteger(); + if (bitPos == 0) + mask = 0x80000000; + else + { + mask = (uint)1 << bitPos; + bytePos++; + } + result.dataLength = (int)bytePos; + + for (int i = (int)bytePos - 1; i >= 0; i--) + { + while (mask != 0) + { + // guess + result.data[i] ^= mask; + + // undo the guess if its square is larger than this + if ((result * result) > this) + result.data[i] ^= mask; + + mask >>= 1; + } + mask = 0x80000000; + } + return result; + } + + + //*********************************************************************** + // Returns the k_th number in the Lucas Sequence reduced modulo n. + // + // Uses index doubling to speed up the process. For example, to calculate V(k), + // we maintain two numbers in the sequence V(n) and V(n+1). + // + // To obtain V(2n), we use the identity + // V(2n) = (V(n) * V(n)) - (2 * Q^n) + // To obtain V(2n+1), we first write it as + // V(2n+1) = V((n+1) + n) + // and use the identity + // V(m+n) = V(m) * V(n) - Q * V(m-n) + // Hence, + // V((n+1) + n) = V(n+1) * V(n) - Q^n * V((n+1) - n) + // = V(n+1) * V(n) - Q^n * V(1) + // = V(n+1) * V(n) - Q^n * P + // + // We use k in its binary expansion and perform index doubling for each + // bit position. For each bit position that is set, we perform an + // index doubling followed by an index addition. This means that for V(n), + // we need to update it to V(2n+1). For V(n+1), we need to update it to + // V((2n+1)+1) = V(2*(n+1)) + // + // This function returns + // [0] = U(k) + // [1] = V(k) + // [2] = Q^n + // + // Where U(0) = 0 % n, U(1) = 1 % n + // V(0) = 2 % n, V(1) = P % n + //*********************************************************************** + + public static BigInteger[] LucasSequence(BigInteger P, BigInteger Q, + BigInteger k, BigInteger n) + { + if (k.dataLength == 1 && k.data[0] == 0) + { + BigInteger[] result = new BigInteger[3]; + + result[0] = 0; result[1] = 2 % n; result[2] = 1 % n; + return result; + } + + // calculate constant = b^(2k) / m + // for Barrett Reduction + BigInteger constant = new BigInteger(); + + int nLen = n.dataLength << 1; + constant.data[nLen] = 0x00000001; + constant.dataLength = nLen + 1; + + constant = constant / n; + + // calculate values of s and t + int s = 0; + + for (int index = 0; index < k.dataLength; index++) + { + uint mask = 0x01; + + for (int i = 0; i < 32; i++) + { + if ((k.data[index] & mask) != 0) + { + index = k.dataLength; // to break the outer loop + break; + } + mask <<= 1; + s++; + } + } + + BigInteger t = k >> s; + + //Console.WriteLine("s = " + s + " t = " + t); + return LucasSequenceHelper(P, Q, t, n, constant, s); + } + + + //*********************************************************************** + // Performs the calculation of the kth term in the Lucas Sequence. + // For details of the algorithm, see reference [9]. + // + // k must be odd. i.e LSB == 1 + //*********************************************************************** + + private static BigInteger[] LucasSequenceHelper(BigInteger P, BigInteger Q, + BigInteger k, BigInteger n, + BigInteger constant, int s) + { + BigInteger[] result = new BigInteger[3]; + + if ((k.data[0] & 0x00000001) == 0) + throw (new ArgumentException("Argument k must be odd.")); + + int numbits = k.bitCount(); + uint mask = (uint)0x1 << ((numbits & 0x1F) - 1); + + // v = v0, v1 = v1, u1 = u1, Q_k = Q^0 + + BigInteger v = 2 % n, Q_k = 1 % n, + v1 = P % n, u1 = Q_k; + bool flag = true; + + for (int i = k.dataLength - 1; i >= 0; i--) // iterate on the binary expansion of k + { + //Console.WriteLine("round"); + while (mask != 0) + { + if (i == 0 && mask == 0x00000001) // last bit + break; + + if ((k.data[i] & mask) != 0) // bit is set + { + // index doubling with addition + + u1 = (u1 * v1) % n; + + v = ((v * v1) - (P * Q_k)) % n; + v1 = n.BarrettReduction(v1 * v1, n, constant); + v1 = (v1 - ((Q_k * Q) << 1)) % n; + + if (flag) + flag = false; + else + Q_k = n.BarrettReduction(Q_k * Q_k, n, constant); + + Q_k = (Q_k * Q) % n; + } + else + { + // index doubling + u1 = ((u1 * v) - Q_k) % n; + + v1 = ((v * v1) - (P * Q_k)) % n; + v = n.BarrettReduction(v * v, n, constant); + v = (v - (Q_k << 1)) % n; + + if (flag) + { + Q_k = Q % n; + flag = false; + } + else + Q_k = n.BarrettReduction(Q_k * Q_k, n, constant); + } + + mask >>= 1; + } + mask = 0x80000000; + } + + // at this point u1 = u(n+1) and v = v(n) + // since the last bit always 1, we need to transform u1 to u(2n+1) and v to v(2n+1) + + u1 = ((u1 * v) - Q_k) % n; + v = ((v * v1) - (P * Q_k)) % n; + if (flag) + flag = false; + else + Q_k = n.BarrettReduction(Q_k * Q_k, n, constant); + + Q_k = (Q_k * Q) % n; + + + for (int i = 0; i < s; i++) + { + // index doubling + u1 = (u1 * v) % n; + v = ((v * v) - (Q_k << 1)) % n; + + if (flag) + { + Q_k = Q % n; + flag = false; + } + else + Q_k = n.BarrettReduction(Q_k * Q_k, n, constant); + } + + result[0] = u1; + result[1] = v; + result[2] = Q_k; + + return result; + } + + + //*********************************************************************** + // Tests the correct implementation of the /, %, * and + operators + //*********************************************************************** + + public static void MulDivTest(int rounds) + { + Random rand = new Random(); + byte[] val = new byte[64]; + byte[] val2 = new byte[64]; + + for (int count = 0; count < rounds; count++) + { + // generate 2 numbers of random length + int t1 = 0; + while (t1 == 0) + t1 = (int)(rand.NextDouble() * 65); + + int t2 = 0; + while (t2 == 0) + t2 = (int)(rand.NextDouble() * 65); + + bool done = false; + while (!done) + { + for (int i = 0; i < 64; i++) + { + if (i < t1) + val[i] = (byte)(rand.NextDouble() * 256); + else + val[i] = 0; + + if (val[i] != 0) + done = true; + } + } + + done = false; + while (!done) + { + for (int i = 0; i < 64; i++) + { + if (i < t2) + val2[i] = (byte)(rand.NextDouble() * 256); + else + val2[i] = 0; + + if (val2[i] != 0) + done = true; + } + } + + while (val[0] == 0) + val[0] = (byte)(rand.NextDouble() * 256); + while (val2[0] == 0) + val2[0] = (byte)(rand.NextDouble() * 256); + + Console.WriteLine(count); + BigInteger bn1 = new BigInteger(val, t1); + BigInteger bn2 = new BigInteger(val2, t2); + + + // Determine the quotient and remainder by dividing + // the first number by the second. + + BigInteger bn3 = bn1 / bn2; + BigInteger bn4 = bn1 % bn2; + + // Recalculate the number + BigInteger bn5 = (bn3 * bn2) + bn4; + + // Make sure they're the same + if (bn5 != bn1) + { + Console.WriteLine("Error at " + count); + Console.WriteLine(bn1 + "\n"); + Console.WriteLine(bn2 + "\n"); + Console.WriteLine(bn3 + "\n"); + Console.WriteLine(bn4 + "\n"); + Console.WriteLine(bn5 + "\n"); + return; + } + } + } + + + //*********************************************************************** + // Tests the correct implementation of the modulo exponential function + // using RSA encryption and decryption (using pre-computed encryption and + // decryption keys). + //*********************************************************************** + + public static void RSATest(int rounds) + { + Random rand = new Random(1); + byte[] val = new byte[64]; + + // private and public key + BigInteger bi_e = new BigInteger("a932b948feed4fb2b692609bd22164fc9edb59fae7880cc1eaff7b3c9626b7e5b241c27a974833b2622ebe09beb451917663d47232488f23a117fc97720f1e7", 16); + BigInteger bi_d = new BigInteger("4adf2f7a89da93248509347d2ae506d683dd3a16357e859a980c4f77a4e2f7a01fae289f13a851df6e9db5adaa60bfd2b162bbbe31f7c8f828261a6839311929d2cef4f864dde65e556ce43c89bbbf9f1ac5511315847ce9cc8dc92470a747b8792d6a83b0092d2e5ebaf852c85cacf34278efa99160f2f8aa7ee7214de07b7", 16); + BigInteger bi_n = new BigInteger("e8e77781f36a7b3188d711c2190b560f205a52391b3479cdb99fa010745cbeba5f2adc08e1de6bf38398a0487c4a73610d94ec36f17f3f46ad75e17bc1adfec99839589f45f95ccc94cb2a5c500b477eb3323d8cfab0c8458c96f0147a45d27e45a4d11d54d77684f65d48f15fafcc1ba208e71e921b9bd9017c16a5231af7f", 16); + + Console.WriteLine("e =\n" + bi_e.ToString(10)); + Console.WriteLine("\nd =\n" + bi_d.ToString(10)); + Console.WriteLine("\nn =\n" + bi_n.ToString(10) + "\n"); + + for (int count = 0; count < rounds; count++) + { + // generate data of random length + int t1 = 0; + while (t1 == 0) + t1 = (int)(rand.NextDouble() * 65); + + bool done = false; + while (!done) + { + for (int i = 0; i < 64; i++) + { + if (i < t1) + val[i] = (byte)(rand.NextDouble() * 256); + else + val[i] = 0; + + if (val[i] != 0) + done = true; + } + } + + while (val[0] == 0) + val[0] = (byte)(rand.NextDouble() * 256); + + Console.Write("Round = " + count); + + // encrypt and decrypt data + BigInteger bi_data = new BigInteger(val, t1); + BigInteger bi_encrypted = bi_data.modPow(bi_e, bi_n); + BigInteger bi_decrypted = bi_encrypted.modPow(bi_d, bi_n); + + // compare + if (bi_decrypted != bi_data) + { + Console.WriteLine("\nError at round " + count); + Console.WriteLine(bi_data + "\n"); + return; + } + Console.WriteLine(" <PASSED>."); + } + + } + + + //*********************************************************************** + // Tests the correct implementation of the modulo exponential and + // inverse modulo functions using RSA encryption and decryption. The two + // pseudoprimes p and q are fixed, but the two RSA keys are generated + // for each round of testing. + //*********************************************************************** + + public static void RSATest2(int rounds) + { + Random rand = new Random(); + byte[] val = new byte[64]; + + byte[] pseudoPrime1 = { + (byte)0x85, (byte)0x84, (byte)0x64, (byte)0xFD, (byte)0x70, (byte)0x6A, + (byte)0x9F, (byte)0xF0, (byte)0x94, (byte)0x0C, (byte)0x3E, (byte)0x2C, + (byte)0x74, (byte)0x34, (byte)0x05, (byte)0xC9, (byte)0x55, (byte)0xB3, + (byte)0x85, (byte)0x32, (byte)0x98, (byte)0x71, (byte)0xF9, (byte)0x41, + (byte)0x21, (byte)0x5F, (byte)0x02, (byte)0x9E, (byte)0xEA, (byte)0x56, + (byte)0x8D, (byte)0x8C, (byte)0x44, (byte)0xCC, (byte)0xEE, (byte)0xEE, + (byte)0x3D, (byte)0x2C, (byte)0x9D, (byte)0x2C, (byte)0x12, (byte)0x41, + (byte)0x1E, (byte)0xF1, (byte)0xC5, (byte)0x32, (byte)0xC3, (byte)0xAA, + (byte)0x31, (byte)0x4A, (byte)0x52, (byte)0xD8, (byte)0xE8, (byte)0xAF, + (byte)0x42, (byte)0xF4, (byte)0x72, (byte)0xA1, (byte)0x2A, (byte)0x0D, + (byte)0x97, (byte)0xB1, (byte)0x31, (byte)0xB3, + }; + + byte[] pseudoPrime2 = { + (byte)0x99, (byte)0x98, (byte)0xCA, (byte)0xB8, (byte)0x5E, (byte)0xD7, + (byte)0xE5, (byte)0xDC, (byte)0x28, (byte)0x5C, (byte)0x6F, (byte)0x0E, + (byte)0x15, (byte)0x09, (byte)0x59, (byte)0x6E, (byte)0x84, (byte)0xF3, + (byte)0x81, (byte)0xCD, (byte)0xDE, (byte)0x42, (byte)0xDC, (byte)0x93, + (byte)0xC2, (byte)0x7A, (byte)0x62, (byte)0xAC, (byte)0x6C, (byte)0xAF, + (byte)0xDE, (byte)0x74, (byte)0xE3, (byte)0xCB, (byte)0x60, (byte)0x20, + (byte)0x38, (byte)0x9C, (byte)0x21, (byte)0xC3, (byte)0xDC, (byte)0xC8, + (byte)0xA2, (byte)0x4D, (byte)0xC6, (byte)0x2A, (byte)0x35, (byte)0x7F, + (byte)0xF3, (byte)0xA9, (byte)0xE8, (byte)0x1D, (byte)0x7B, (byte)0x2C, + (byte)0x78, (byte)0xFA, (byte)0xB8, (byte)0x02, (byte)0x55, (byte)0x80, + (byte)0x9B, (byte)0xC2, (byte)0xA5, (byte)0xCB, + }; + + + BigInteger bi_p = new BigInteger(pseudoPrime1); + BigInteger bi_q = new BigInteger(pseudoPrime2); + BigInteger bi_pq = (bi_p - 1) * (bi_q - 1); + BigInteger bi_n = bi_p * bi_q; + + for (int count = 0; count < rounds; count++) + { + // generate private and public key + BigInteger bi_e = bi_pq.genCoPrime(512, rand); + BigInteger bi_d = bi_e.modInverse(bi_pq); + + Console.WriteLine("\ne =\n" + bi_e.ToString(10)); + Console.WriteLine("\nd =\n" + bi_d.ToString(10)); + Console.WriteLine("\nn =\n" + bi_n.ToString(10) + "\n"); + + // generate data of random length + int t1 = 0; + while (t1 == 0) + t1 = (int)(rand.NextDouble() * 65); + + bool done = false; + while (!done) + { + for (int i = 0; i < 64; i++) + { + if (i < t1) + val[i] = (byte)(rand.NextDouble() * 256); + else + val[i] = 0; + + if (val[i] != 0) + done = true; + } + } + + while (val[0] == 0) + val[0] = (byte)(rand.NextDouble() * 256); + + Console.Write("Round = " + count); + + // encrypt and decrypt data + BigInteger bi_data = new BigInteger(val, t1); + BigInteger bi_encrypted = bi_data.modPow(bi_e, bi_n); + BigInteger bi_decrypted = bi_encrypted.modPow(bi_d, bi_n); + + // compare + if (bi_decrypted != bi_data) + { + Console.WriteLine("\nError at round " + count); + Console.WriteLine(bi_data + "\n"); + return; + } + Console.WriteLine(" <PASSED>."); + } + + } + + + //*********************************************************************** + // Tests the correct implementation of sqrt() method. + //*********************************************************************** + + public static void SqrtTest(int rounds) + { + Random rand = new Random(); + for (int count = 0; count < rounds; count++) + { + // generate data of random length + int t1 = 0; + while (t1 == 0) + t1 = (int)(rand.NextDouble() * 1024); + + Console.Write("Round = " + count); + + BigInteger a = new BigInteger(); + a.genRandomBits(t1, rand); + + BigInteger b = a.sqrt(); + BigInteger c = (b + 1) * (b + 1); + + // check that b is the largest integer such that b*b <= a + if (c <= a) + { + Console.WriteLine("\nError at round " + count); + Console.WriteLine(a + "\n"); + return; + } + Console.WriteLine(" <PASSED>."); + } + } + } +} -- Gitblit v1.8.0