com.github.kiprobinson.bigfraction.util

Class FloatUtil

java.lang.Object

com.github.kiprobinson.bigfraction.util.FloatUtil

public final class FloatUtil
extends Object

Additional utilities for working with floay values. Useful to prevent hard-coding IEEE 754 constants in code. Consider this a complement to the methods provided in Float.

See Also:

Wikipedia overview of IEEE 754 float-precision floating point specifications

Field Summary

Fields

Modifier and Type	Field and Description
static int	MAX_EXPONENT The largest permitted exponent value, when using an exponent with offset.
static int	MAX_EXPONENT_BITS The largest permitted exponent value, when using raw exponent bits.
static int	MAX_MANTISSA The largest permitted mantissa value.
static int	MIN_EXPONENT The smallest permitted exponent value, when using an exponent with offset.

Method Summary

Methods

Modifier and Type	Method and Description
static int[]	getAllParts(float f) Returns an array containing the parts of the float.
static int[]	getAllParts(float f, boolean exponentAsBits) Returns an array containing the parts of the float.
static int	getExponent(float f) Returns the exponent, after adding the exponent offset to the exponent bits.
static int	getExponentBits(float f) Returns the raw exponent bits, without adjusting for the offset.
static float	getFloat(int sign, int exponent, int mantissa) Creates a new float primitive using the provided component bits.
static float	getFloat(int sign, int exponent, int mantissa, boolean exponentAsBits) Creates a new float primitive using the provided component bits.
static int	getMantissa(float f) Returns the mantissa bits (bits 51-0).
static int	getSign(float f) Returns the sign bit (bit 63).
static boolean	isFinite(float f) Returns true if f is finite--not infinite and not NaN.
static boolean	isSubnormal(float f) Returns whether or not this float is a subnormal value.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

MAX_MANTISSA

public static final int MAX_MANTISSA

The largest permitted mantissa value.

See Also:

Constant Field Values

MAX_EXPONENT_BITS

public static final int MAX_EXPONENT_BITS

The largest permitted exponent value, when using raw exponent bits.

See Also:

Constant Field Values

MIN_EXPONENT

public static final int MIN_EXPONENT

The smallest permitted exponent value, when using an exponent with offset.

See Also:

Constant Field Values

MAX_EXPONENT

public static final int MAX_EXPONENT

The largest permitted exponent value, when using an exponent with offset.

See Also:

Constant Field Values

Method Detail

isFinite

public static boolean isFinite(float f)

Returns true if f is finite--not infinite and not NaN. (Equivalent to Float.isFinite() available from Java 8.)

Parameters:

f - a float value

Returns:

whether this value is finite.

getSign

public static int getSign(float f)

Returns the sign bit (bit 63). 0=positive, 1=negative. This is returned even for NaN values.

Parameters:

f - a float value

Returns:

the sign bit

getExponent

public static int getExponent(float f)

Returns the exponent, after adding the exponent offset to the exponent bits. In other words, the value in bits 62-52, plus 0x3ff.

Examples:

• getExponent(0.5) == -1 because 1.0 == 1.0 * 2^-1
• getExponent(1.0) == 0 because 1.0 == 1.0 * 2^0
• getExponent(2.0) == 1 because 2.0 == 1.0 * 2^1
• Subnormal values: Always returns -1023.
• Zero: Always returns -1023.
• Infinity: Always returns 1024.
• NaN: Always returns 1024.

Parameters:

f - any float value

Returns:

adjusted exponent

getExponentBits

public static int getExponentBits(float f)

Returns the raw exponent bits, without adjusting for the offset. In other words, the value in bits 62-52.

Examples:

• getExponentBits(0.5) == 0x3fe
• getExponentBits(1.0) == 0x3ff
• getExponentBits(2.0) == 0x400
• Subnormal values: Always returns 0x000.
• Zero: Always returns 0x000.
• Infinity: Always returns 0x7ff.
• NaN: Always returns 0x7ff.

Parameters:

f - a float value

Returns:

raw exponent bits

getMantissa

public static int getMantissa(float f)

Returns the mantissa bits (bits 51-0). Returned even for NaN values.

Parameters:

f - a float value

Returns:

mantissa bits

isSubnormal

public static boolean isSubnormal(float f)

Returns whether or not this float is a subnormal value.

Parameters:

f - a float value

Returns:

whether or not this float is a subnormal value

getAllParts

public static int[] getAllParts(float f)

Returns an array containing the parts of the float. Avoids the overhead of four separate function calls.

Parameters:

f - a float value

Returns:

array with four elements:

• return[0]: Equivalent to getSign(f)
• return[1]: Equivalent to getExponent(f)
• return[2]: Equivalent to getMantissa(f)
• return[3]: Equivalent to isSubnormal(f) - Uses zero for false, non-zero for true.

getAllParts

public static int[] getAllParts(float f,
                boolean exponentAsBits)

Returns an array containing the parts of the float. Avoids the overhead of four separate function calls.

Parameters:

f - a float value

exponentAsBits - whether to return exponent as raw bits rather than adjusted value

Returns:

array with four elements:

• return[0]: Equivalent to getSign(f)
• return[1]: Equivalent to (exponentAsBits ? getExponentBits(f) : getExponent(f))
• return[2]: Equivalent to getMantissa(f)
• return[3]: Equivalent to isSubnormal(f) - Uses zero for false, non-zero for true.

getFloat

public static float getFloat(int sign,
             int exponent,
             int mantissa)

Creates a new float primitive using the provided component bits. Assumes the exponent parameter is signed.

Parameters:

sign - sign bit

exponent - adjusted exponent

mantissa - mantissa bits

Returns:

The float value represented by the provided binary parts.

Throws:

IllegalArgumentException - if any of the parts contain invalid bits.

getFloat

public static float getFloat(int sign,
             int exponent,
             int mantissa,
             boolean exponentAsBits)

Creates a new float primitive using the provided component bits.

Parameters:

sign - sign bit

exponent - exponent (either raw bits or adjusted value)

mantissa - mantissa bits

exponentAsBits - If true, assumes that exponent parameter represents the actual exponent bits. If false, IEEE exponent offset value will be added to the offset to get the bits.

Returns:

The float value represented by the provided binary parts.

Throws:

IllegalArgumentException - if any of the parts contain invalid bits.