本文整理汇总了C#中Rational.Abs方法的典型用法代码示例。如果您正苦于以下问题：C# Rational.Abs方法的具体用法？C# Rational.Abs怎么用？C# Rational.Abs使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类Rational的用法示例。
在下文中一共展示了Rational.Abs方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的C#代码示例。
示例1: PowRound
public static BigDecimal PowRound(BigDecimal x, Rational q)
{
/** Special cases: x^1=x and x^0 = 1
*/
if (q.CompareTo(BigInteger.One) == 0)
return x;
if (q.Sign == 0)
return BigDecimal.One;
if (q.IsInteger) {
/* We are sure that the denominator is positive here, because normalize() has been
* called during constrution etc.
*/
return PowRound(x, q.Numerator);
}
/* Refuse to operate on the general negative basis. The integer q have already been handled above.
*/
if (x.CompareTo(BigDecimal.Zero) < 0)
throw new ArithmeticException("Cannot power negative " + x);
if (q.IsIntegerFraction) {
/* Newton method with first estimate in double precision.
* The disadvantage of this first line here is that the result must fit in the
* standard range of double precision numbers exponents.
*/
double estim = System.Math.Pow(x.ToDouble(), q.ToDouble());
var res = new BigDecimal(estim);
/* The error in x^q is q*x^(q-1)*Delta(x).
* The relative error is q*Delta(x)/x, q times the relative error of x.
*/
var reserr = new BigDecimal(0.5*q.Abs().ToDouble()
*x.Ulp().Divide(x.Abs(), MathContext.Decimal64).ToDouble());
/* The main point in branching the cases above is that this conversion
* will succeed for numerator and denominator of q.
*/
int qa = q.Numerator.ToInt32();
int qb = q.Denominator.ToInt32();
/* Newton iterations. */
BigDecimal xpowa = PowRound(x, qa);
for (;;) {
/* numerator and denominator of the Newton term. The major
* disadvantage of this implementation is that the updates of the powers
* of the new estimate are done in full precision calling BigDecimal.pow(),
* which becomes slow if the denominator of q is large.
*/
BigDecimal nu = res.Pow(qb).Subtract(xpowa);
BigDecimal de = MultiplyRound(res.Pow(qb - 1), q.Denominator);
/* estimated correction */
BigDecimal eps = nu.Divide(de, MathContext.Decimal64);
BigDecimal err = res.Multiply(reserr, MathContext.Decimal64);
int precDiv = 2 + ErrorToPrecision(eps, err);
if (precDiv <= 0) {
/* The case when the precision is already reached and any precision
* will do. */
eps = nu.Divide(de, MathContext.Decimal32);
} else {
eps = nu.Divide(de, new MathContext(precDiv));
}
res = SubtractRound(res, eps);
/* reached final precision if the relative error fell below reserr,
* |eps/res| < reserr
*/
if (eps.Divide(res, MathContext.Decimal64).Abs().CompareTo(reserr) < 0) {
/* delete the bits of extra precision kept in this
* working copy.
*/
return res.Round(new MathContext(ErrorToPrecision(reserr.ToDouble())));
}
}
}
/* The error in x^q is q*x^(q-1)*Delta(x) + Delta(q)*x^q*log(x).
* The relative error is q/x*Delta(x) + Delta(q)*log(x). Convert q to a floating point
* number such that its relative error becomes negligible: Delta(q)/q << Delta(x)/x/log(x) .
*/
int precq = 3 + ErrorToPrecision((x.Ulp().Divide(x, MathContext.Decimal64)).ToDouble()
/System.Math.Log(x.ToDouble()));
/* Perform the actual calculation as exponentiation of two floating point numbers.
*/
return Pow(x, q.ToBigDecimal(new MathContext(precq)));
}