java.lang.Object

ch.randelshofer.fastdoubleparser.FastDoubleMath

final class FastDoubleMath extends Object

This class provides the mathematical functions needed by JavaDoubleParser.

References:

Daniel Lemire, fast_float number parsing library: 4x faster than strtod. MIT License.: github.com
Daniel Lemire, Number Parsing at a Gigabyte per Second, Software: Practice and Experience 51 (8), 2021. arXiv.2101.11408v3 [cs.DS] 24 Feb 2021: arxiv.org
Clinger WD (1990). How to read floating point numbers accurately. ACM SIGPLAN Notices.: - no link -

Field Summary

Fields

Modifier and Type

Field

Description

static final int

DOUBLE_EXPONENT_BIAS

Bias used in the exponent of a double.

(package private) static final int

DOUBLE_MAX_EXPONENT_POWER_OF_TEN

Largest power of 10 value of the exponent.

private static final int

DOUBLE_MAX_EXPONENT_POWER_OF_TWO

(package private) static final int

DOUBLE_MIN_EXPONENT_POWER_OF_TEN

Smallest power of 10 value of the exponent.

private static final int

DOUBLE_MIN_EXPONENT_POWER_OF_TWO

private static final double[]

DOUBLE_POWERS_OF_TEN

Precomputed powers of ten from 10^0 to 10^22.

static final int

DOUBLE_SIGNIFICAND_WIDTH

The number of bits in the significand, including the implicit bit.

(package private) static final long[]

MANTISSA_128

A complement to mantissa_64 complete to a 128-bit mantissa.

(package private) static final long[]

MANTISSA_64

When mapping numbers from decimal to binary, we go from w * 10^q to m * 2^p, but we have 10^q = 5^q * 2^q, so effectively we are trying to match w * 2^q * 5^q to m * 2^p.

static final int

MAX_REQUIRED_DIGITS
Constructor Summary

Constructors

Modifier

Constructor

Description

private

FastDoubleMath()

Don't let anyone instantiate this class.
Method Summary

Modifier and Type

Method

Description

(package private) static double

fastScalb(double number, long scaleFactor)

This is a faster alternative to Math.scalb(double, int).

(package private) static double

tryDecFloatToDoubleTruncated(boolean isNegative, long significand, int exponent, boolean isSignificandTruncated, int exponentOfTruncatedSignificand)

Tries to compute significand * 10^exponent exactly using a fast algorithm; and if isNegative is true, negate the result; the significand can be truncated.

(package private) static double

tryDecToDoubleWithFastAlgorithm(boolean isNegative, long significand, int power)

Tries to compute significand * 10^power exactly using a fast algorithm; and if isNegative is true, negate the result.

(package private) static double

tryHexFloatToDoubleTruncated(boolean isNegative, long significand, long exponent, boolean isSignificandTruncated, long exponentOfTruncatedSignificand)

Tries to compute significand * 2^exponent exactly using a fast algorithm; and if isNegative is true, negate the result; the significand can be truncated.

Methods inherited from class Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details
- DOUBLE_EXPONENT_BIAS
  public static final int DOUBLE_EXPONENT_BIAS
  
  Bias used in the exponent of a double.
  
  See Also:
  
  Constant Field Values
- DOUBLE_SIGNIFICAND_WIDTH
  public static final int DOUBLE_SIGNIFICAND_WIDTH
  
  The number of bits in the significand, including the implicit bit.
  
  See Also:
  
  Constant Field Values
- MAX_REQUIRED_DIGITS
  public static final int MAX_REQUIRED_DIGITS
  
  See Also:
  
  Constant Field Values
- DOUBLE_MIN_EXPONENT_POWER_OF_TEN
  static final int DOUBLE_MIN_EXPONENT_POWER_OF_TEN
  
  Smallest power of 10 value of the exponent.
  The smallest non-zero double is 2^−1074.
  We take as input numbers of the form w x 10^q where w invalid input: '<' 2^64.
  We have that w * 10^-343 < 2^(63-343) * 5^-343 < 2^-1076.
  However, we have that (2^64-1) * 10^-342 = (2^64 - 1) * 2^-342 * 5^-342 > 2^−1074. Thus, it is possible for a number of the form w * 10^-342 where w is a 64-bit value to be a non-zero double.
  ********
  If we are solely interested in the *normal* numbers then the smallest value is 2^-1022. We can generate a value larger than 2^-1022 with expressions of the form w * 10^-326. Thus, we need to pick SMALLEST_POWER_OF_TEN >= -326.
  
  See Also:
  
  Constant Field Values
- DOUBLE_MAX_EXPONENT_POWER_OF_TEN
  static final int DOUBLE_MAX_EXPONENT_POWER_OF_TEN
  
  Largest power of 10 value of the exponent.
  Any number of form w * 10^309 where w >= 1 is going to be infinite in a double, so we never need to worry about powers of 10 greater than 308.
  
  See Also:
  
  Constant Field Values
- MANTISSA_64
  static final long[] MANTISSA_64
  
  When mapping numbers from decimal to binary, we go from w * 10^q to m * 2^p, but we have 10^q = 5^q * 2^q, so effectively we are trying to match w * 2^q * 5^q to m * 2^p.
  Thus, the powers of two are not a concern since they can be represented exactly using the binary notation, only the powers of five affect the binary significand.
  The mantissas of powers of ten from -308 to 308, extended out to sixty-four bits. The array contains the powers of ten approximated as a 64-bit mantissa. It goes from 10^-325 to 10^308 (inclusively). The mantissa is truncated, and never rounded up. Uses about 5 KB.
  
  long getMantissaHigh(int q) { MANTISSA_64[q - SMALLEST_POWER_OF_TEN]; }
- MANTISSA_128
  static final long[] MANTISSA_128
  
  A complement to mantissa_64 complete to a 128-bit mantissa.
  Uses about 5KB but is rarely accessed.
  UInt128 getMantissa128(int q) { return new UInt128( MANTISSA_64[q - SMALLEST_POWER_OF_TEN], MANTISSA_128[q - SMALLEST_POWER_OF_TEN]; ); }
- DOUBLE_MIN_EXPONENT_POWER_OF_TWO
  private static final int DOUBLE_MIN_EXPONENT_POWER_OF_TWO
  
  See Also:
  
  Constant Field Values
- DOUBLE_MAX_EXPONENT_POWER_OF_TWO
  private static final int DOUBLE_MAX_EXPONENT_POWER_OF_TWO
  
  See Also:
  
  Constant Field Values
- DOUBLE_POWERS_OF_TEN
  
  private static final double[] DOUBLE_POWERS_OF_TEN
  
  Precomputed powers of ten from 10^0 to 10^22. These can be represented exactly using the double type.
Constructor Details
- FastDoubleMath
  
  private FastDoubleMath()
  
  Don't let anyone instantiate this class.
Method Details
- tryDecFloatToDoubleTruncated
  
  static double tryDecFloatToDoubleTruncated(boolean isNegative, long significand, int exponent, boolean isSignificandTruncated, int exponentOfTruncatedSignificand)
  
  Tries to compute significand * 10^exponent exactly using a fast algorithm; and if isNegative is true, negate the result; the significand can be truncated.
  
  Parameters:
  
  isNegative - true if the sign is negative
  
  significand - the significand
  
  exponent - the exponent number (the power)
  
  isSignificandTruncated - true if significand has been truncated
  
  exponentOfTruncatedSignificand - the exponent number of the truncated significand
  
  Returns:
  
  the double value, or Double.NaN if the fast path failed.
- tryDecToDoubleWithFastAlgorithm
  
  static double tryDecToDoubleWithFastAlgorithm(boolean isNegative, long significand, int power)
  
  Tries to compute significand * 10^power exactly using a fast algorithm; and if isNegative is true, negate the result.
  This function will only work in some cases, when it does not work it returns NaN. This should work *most of the time* (like 99% of the time). We assume that power is in the [-325, 308] interval: the caller is responsible for this check.
  References:
  
  Noble Mushtak, Daniel Lemire. (2023) Fast Number Parsing Without Fallback.
  
  arxiv.org
  
  Parameters:
  
  isNegative - whether the number is negative
  
  significand - uint64 the significand
  
  power - the exponent number (the power)
  
  Returns:
  
  the computed double on success, Double.NaN on failure
- tryHexFloatToDoubleTruncated
  
  static double tryHexFloatToDoubleTruncated(boolean isNegative, long significand, long exponent, boolean isSignificandTruncated, long exponentOfTruncatedSignificand)
  
  Tries to compute significand * 2^exponent exactly using a fast algorithm; and if isNegative is true, negate the result; the significand can be truncated.
  
  Parameters:
  
  isNegative - true if the sign is negative
  
  significand - the significand (unsigned long, uint64)
  
  exponent - the exponent number (the power)
  
  isSignificandTruncated - true if significand has been truncated
  
  exponentOfTruncatedSignificand - the exponent number of the truncated significand
  
  Returns:
  
  the double value, or Double.NaN if the fast path failed.
- fastScalb
  
  static double fastScalb(double number, long scaleFactor)
  
  This is a faster alternative to Math.scalb(double, int).
  This method only works if scaleFactor is within the range of Double.MIN_EXPONENT through Double.MAX_EXPONENT (inclusive), so that we do not underflow or overflow.
  
  Parameters:
  
  number - a double number
  
  scaleFactor - the scale factor
  
  Returns:
  
  number × 2^scaleFactor

Class FastDoubleMath

Field Summary

Constructor Summary

Method Summary

Methods inherited from class Object

Field Details

DOUBLE_EXPONENT_BIAS

DOUBLE_SIGNIFICAND_WIDTH

MAX_REQUIRED_DIGITS

DOUBLE_MIN_EXPONENT_POWER_OF_TEN

DOUBLE_MAX_EXPONENT_POWER_OF_TEN

MANTISSA_64

MANTISSA_128

DOUBLE_MIN_EXPONENT_POWER_OF_TWO

DOUBLE_MAX_EXPONENT_POWER_OF_TWO

DOUBLE_POWERS_OF_TEN

Constructor Details

FastDoubleMath

Method Details

tryDecFloatToDoubleTruncated

tryDecToDoubleWithFastAlgorithm

tryHexFloatToDoubleTruncated

fastScalb