Encoding information for a single floating-point format. More...

#include <float.hh>

Public Types
enum	floatclass { normalized = 0 , infinity = 1 , zero = 2 , nan = 3 , denormalized = 4 }
	The various classes of floating-point encodings. More...

Public Member Functions
	FloatFormat (int4 sz)
	Construct default IEEE 754 standard settings.

int4	getSize (void) const
	Get the size of the encoding in bytes.

double	getHostFloat (uintb encoding, floatclass *type) const
	Convert an encoding into host's double.

uintb	getEncoding (double host) const
	Convert host's double into this encoding.

uintb	convertEncoding (uintb encoding, const FloatFormat *formin) const
	Convert between two different formats.

uintb	extractFractionalCode (uintb x) const
	Extract the fractional part of the encoding.

bool	extractSign (uintb x) const
	Extract the sign bit from the encoding.

int4	extractExponentCode (uintb x) const
	Extract the exponent from the encoding.

string	printDecimal (double host, bool forcesci) const
	Print given value as a decimal string.

uintb	opEqual (uintb a, uintb b) const
	Equality comparison (==)

uintb	opNotEqual (uintb a, uintb b) const
	Inequality comparison (!=)

uintb	opLess (uintb a, uintb b) const
	Less-than comparison (<)

uintb	opLessEqual (uintb a, uintb b) const
	Less-than-or-equal comparison (<=)

uintb	opNan (uintb a) const
	Test if Not-a-Number (NaN)

uintb	opAdd (uintb a, uintb b) const
	Addition (+)

uintb	opDiv (uintb a, uintb b) const
	Division (/)

uintb	opMult (uintb a, uintb b) const
	Multiplication (*)

uintb	opSub (uintb a, uintb b) const
	Subtraction (-)

uintb	opNeg (uintb a) const
	Unary negate.

uintb	opAbs (uintb a) const
	Absolute value (abs)

uintb	opSqrt (uintb a) const
	Square root (sqrt)

uintb	opTrunc (uintb a, int4 sizeout) const
	Convert floating-point to integer.

uintb	opCeil (uintb a) const
	Ceiling (ceil)

uintb	opFloor (uintb a) const
	Floor (floor)

uintb	opRound (uintb a) const
	Round.

uintb	opInt2Float (uintb a, int4 sizein) const
	Convert integer to floating-point.

uintb	opFloat2Float (uintb a, const FloatFormat &outformat) const
	Convert between floating-point precisions.

Private Member Functions
uintb	setFractionalCode (uintb x, uintb code) const
	Set the fractional part of an encoded value.

uintb	setSign (uintb x, bool sign) const
	Set the sign bit of an encoded value.

uintb	setExponentCode (uintb x, uintb code) const
	Set the exponent of an encoded value.

uintb	getZeroEncoding (bool sgn) const
	Get an encoded zero value.

uintb	getInfinityEncoding (bool sgn) const
	Get an encoded infinite value.

uintb	getNaNEncoding (bool sgn) const
	Get an encoded NaN value.

void	calcPrecision (void)
	Calculate the decimal precision of this format.

Static Private Member Functions
static double	createFloat (bool sign, uintb signif, int4 exp)
	Create a double given sign, fractional, and exponent.

static floatclass	extractExpSig (double x, bool sgn, uintb signif, int4 *exp)
	Extract the sign, fractional, and exponent from a given floating-point value.

static bool	roundToNearestEven (uintb &signif, int4 lowbitpos)
	Round a floating point value to the nearest even.

Private Attributes
int4	size
	Size of float in bytes (this format)

int4	signbit_pos
	Bit position of sign bit.

int4	frac_pos
	(lowest) bit position of fractional part

int4	frac_size
	Number of bits in fractional part.

int4	exp_pos
	(lowest) bit position of exponent

int4	exp_size
	Number of bits in exponent.

int4	bias
	What to add to real exponent to get encoding.

int4	maxexponent
	Maximum possible exponent.

int4	decimalMinPrecision
	Minimum decimal digits of precision guaranteed by the format.

int4	decimalMaxPrecision
	Maximum decimal digits of precision needed to uniquely represent value.

bool	jbitimplied
	Set to true if integer bit of 1 is assumed.

Detailed Description

Encoding information for a single floating-point format.

This class supports manipulation of a single floating-point encoding. An encoding can be converted to and from the host format and convenience methods allow p-code floating-point operations to be performed on natively encoded operands. This follows the IEEE754 standards.

Member Enumeration Documentation

◆ floatclass

enum ghidra::FloatFormat::floatclass

The various classes of floating-point encodings.

Enumerator
normalized	A normal floating-point number.
infinity	An encoding representing an infinite value.
zero	An encoding of the value zero.
nan	An invalid encoding, Not-a-Number.
denormalized	A denormalized encoding (for very small values)

Constructor & Destructor Documentation

◆ FloatFormat()

ghidra::FloatFormat::FloatFormat ( int4 sz )

Construct default IEEE 754 standard settings.

Set format for a given encoding size according to IEEE 754 standards

Parameters

sz	is the size of the encoding in bytes

References bias, calcPrecision(), exp_pos, exp_size, frac_pos, frac_size, jbitimplied, maxexponent, signbit_pos, and size.

Member Function Documentation

◆ convertEncoding()

uintb ghidra::FloatFormat::convertEncoding	(	uintb	encoding,
		const FloatFormat *	formin
	)		const

Convert between two different formats.

Parameters

encoding	is the value in the other FloatFormat
formin	is the other FloatFormat

Returns: the equivalent value in this FloatFormat

References bias, extractExponentCode(), extractFractionalCode(), extractSign(), frac_size, getInfinityEncoding(), getNaNEncoding(), getZeroEncoding(), jbitimplied, maxexponent, roundToNearestEven(), setExponentCode(), setFractionalCode(), and setSign().

Referenced by opFloat2Float(), and ghidra::SubfloatFlow::setReplacement().

◆ createFloat()

double ghidra::FloatFormat::createFloat	(	bool	sign,
		uintb	signif,
		int4	exp
	)

staticprivate

Create a double given sign, fractional, and exponent.

Parameters

sign	is set to true if the value should be negative
signif	is the fractional part
exp	is the exponent

Returns: the constructed floating-point value

Referenced by getHostFloat().

◆ extractExponentCode()

int4 ghidra::FloatFormat::extractExponentCode ( uintb x ) const

Extract the exponent from the encoding.

Parameters

x	is an encoded floating-point value

Returns: the (signed) exponent

References exp_pos, and exp_size.

Referenced by convertEncoding(), getHostFloat(), and ghidra::ScoreUnionFields::scoreConstantFit().

◆ extractExpSig()

FloatFormat::floatclass ghidra::FloatFormat::extractExpSig	(	double	x,
		bool *	sgn,
		uintb *	signif,
		int4 *	exp
	)

staticprivate

Extract the sign, fractional, and exponent from a given floating-point value.

Parameters

x	is the given value
sgn	passes back the sign
signif	passes back the fractional part
exp	passes back the exponent

Returns: the floating-point class of the value

References infinity, nan, normalized, and zero.

Referenced by getEncoding().

◆ extractFractionalCode()

uintb ghidra::FloatFormat::extractFractionalCode ( uintb x ) const

Extract the fractional part of the encoding.

Parameters

x	is an encoded floating-point value

Returns: the fraction part of the value aligned to the top of the word

References frac_pos, and frac_size.

Referenced by convertEncoding(), and getHostFloat().

◆ extractSign()

bool ghidra::FloatFormat::extractSign ( uintb x ) const

Extract the sign bit from the encoding.

Parameters

x	is an encoded floating-point value

Returns: the sign bit

References signbit_pos.

Referenced by convertEncoding(), getHostFloat(), and ghidra::PrintC::push_float().

◆ getEncoding()

uintb ghidra::FloatFormat::getEncoding ( double host ) const

Convert host's double into this encoding.

Parameters

host	is the double value to convert

Returns: the equivalent encoded value

References bias, extractExpSig(), frac_size, getInfinityEncoding(), getNaNEncoding(), getZeroEncoding(), infinity, jbitimplied, maxexponent, nan, roundToNearestEven(), setExponentCode(), setFractionalCode(), setSign(), and zero.

Referenced by opAbs(), opAdd(), opCeil(), opDiv(), opFloor(), opInt2Float(), opMult(), opNeg(), opRound(), opSqrt(), and opSub().

◆ getHostFloat()

double ghidra::FloatFormat::getHostFloat	(	uintb	encoding,
		floatclass *	type
	)		const

Convert an encoding into host's double.

Parameters

encoding	is the encoding value
type	points to the floating-point class, which is passed back

Returns: the equivalent double value

References bias, createFloat(), denormalized, extractExponentCode(), extractFractionalCode(), extractSign(), infinity, jbitimplied, maxexponent, nan, normalized, and zero.

Referenced by opAbs(), opAdd(), opCeil(), opDiv(), opEqual(), opFloor(), opLess(), opLessEqual(), opMult(), opNan(), opNeg(), opNotEqual(), opRound(), opSqrt(), opSub(), opTrunc(), and ghidra::PrintC::push_float().

◆ getInfinityEncoding()

uintb ghidra::FloatFormat::getInfinityEncoding ( bool sgn ) const

private

Get an encoded infinite value.

Parameters

sgn	is set to true for negative infinity, false for positive

Returns: the encoded infinity

References maxexponent, setExponentCode(), setFractionalCode(), and setSign().

Referenced by convertEncoding(), and getEncoding().

◆ getNaNEncoding()

uintb ghidra::FloatFormat::getNaNEncoding ( bool sgn ) const

private

Get an encoded NaN value.

Parameters

sgn	is set to true for negative NaN, false for positive

Returns: the encoded NaN

References maxexponent, setExponentCode(), setFractionalCode(), and setSign().

Referenced by convertEncoding(), and getEncoding().

◆ getZeroEncoding()

uintb ghidra::FloatFormat::getZeroEncoding ( bool sgn ) const

private

Get an encoded zero value.

Parameters

sgn	is set to true for negative zero, false for positive

Returns: the encoded zero

References setExponentCode(), setFractionalCode(), and setSign().

Referenced by convertEncoding(), and getEncoding().

◆ opAbs()

uintb ghidra::FloatFormat::opAbs ( uintb a ) const

Absolute value (abs)

Parameters

a	is an encoded floating-point value

Returns: abs(a)

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatAbs::evaluateUnary().

◆ opAdd()

uintb ghidra::FloatFormat::opAdd	(	uintb	a,
		uintb	b
	)		const

Addition (+)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: a + b

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatAdd::evaluateBinary().

◆ opCeil()

uintb ghidra::FloatFormat::opCeil ( uintb a ) const

Ceiling (ceil)

Parameters

a	is an encoded floating-point value

Returns: ceil(a)

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatCeil::evaluateUnary().

◆ opDiv()

uintb ghidra::FloatFormat::opDiv	(	uintb	a,
		uintb	b
	)		const

Division (/)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: a / b

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatDiv::evaluateBinary().

◆ opEqual()

uintb ghidra::FloatFormat::opEqual	(	uintb	a,
		uintb	b
	)		const

Equality comparison (==)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: true if (a == b)

References getHostFloat().

Referenced by ghidra::OpBehaviorFloatEqual::evaluateBinary().

◆ opFloat2Float()

uintb ghidra::FloatFormat::opFloat2Float	(	uintb	a,
		const FloatFormat &	outformat
	)		const

Convert between floating-point precisions.

Parameters

a	is an encoded floating-point value
outformat	is the desired output FloatFormat

Returns: a converted to the output FloatFormat

References convertEncoding().

Referenced by ghidra::OpBehaviorFloatFloat2Float::evaluateUnary().

◆ opFloor()

uintb ghidra::FloatFormat::opFloor ( uintb a ) const

Floor (floor)

Parameters

a	is an encoded floating-point value

Returns: floor(a)

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatFloor::evaluateUnary().

◆ opInt2Float()

uintb ghidra::FloatFormat::opInt2Float	(	uintb	a,
		int4	sizein
	)		const

Convert integer to floating-point.

Parameters

a	is a signed integer value
sizein	is the number of bytes in the integer encoding

Returns: a converted to an encoded floating-point value

References getEncoding().

Referenced by ghidra::OpBehaviorFloatInt2Float::evaluateUnary().

◆ opLess()

uintb ghidra::FloatFormat::opLess	(	uintb	a,
		uintb	b
	)		const

Less-than comparison (<)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: true if (a < b)

References getHostFloat().

Referenced by ghidra::OpBehaviorFloatLess::evaluateBinary().

◆ opLessEqual()

uintb ghidra::FloatFormat::opLessEqual	(	uintb	a,
		uintb	b
	)		const

Less-than-or-equal comparison (<=)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: true if (a <= b)

References getHostFloat().

Referenced by ghidra::OpBehaviorFloatLessEqual::evaluateBinary().

◆ opMult()

uintb ghidra::FloatFormat::opMult	(	uintb	a,
		uintb	b
	)		const

Multiplication (*)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: a * b

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatMult::evaluateBinary().

◆ opNan()

uintb ghidra::FloatFormat::opNan ( uintb a ) const

Test if Not-a-Number (NaN)

Parameters

a	is an encoded floating-point value

Returns: true if a is Not-a-Number

References getHostFloat(), and nan.

Referenced by ghidra::OpBehaviorFloatNan::evaluateUnary().

◆ opNeg()

uintb ghidra::FloatFormat::opNeg ( uintb a ) const

Unary negate.

Parameters

a	is an encoded floating-point value

Returns: -a

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatNeg::evaluateUnary().

◆ opNotEqual()

uintb ghidra::FloatFormat::opNotEqual	(	uintb	a,
		uintb	b
	)		const

Inequality comparison (!=)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: true if (a != b)

References getHostFloat().

Referenced by ghidra::OpBehaviorFloatNotEqual::evaluateBinary().

◆ opRound()

uintb ghidra::FloatFormat::opRound ( uintb a ) const

Round.

Parameters

a	is an encoded floating-point value

Returns: round(a)

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatRound::evaluateUnary().

◆ opSqrt()

uintb ghidra::FloatFormat::opSqrt ( uintb a ) const

Square root (sqrt)

Parameters

a	is an encoded floating-point value

Returns: sqrt(a)

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatSqrt::evaluateUnary().

◆ opSub()

uintb ghidra::FloatFormat::opSub	(	uintb	a,
		uintb	b
	)		const

Subtraction (-)

Parameters

a	is the first floating-point value
b	is the second floating-point value

Returns: a - b

References getEncoding(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatSub::evaluateBinary().

◆ opTrunc()

uintb ghidra::FloatFormat::opTrunc	(	uintb	a,
		int4	sizeout
	)		const

Convert floating-point to integer.

Parameters

a	is an encoded floating-point value
sizeout	is the desired encoding size of the output

Returns: an integer encoding of a

References ghidra::calc_mask(), and getHostFloat().

Referenced by ghidra::OpBehaviorFloatTrunc::evaluateUnary().

◆ printDecimal()

string ghidra::FloatFormat::printDecimal	(	double	host,
		bool	forcesci
	)		const

Print given value as a decimal string.

The string should be printed with the minimum number of digits to uniquely specify the underlying binary value. This currently only works for the 32-bit and 64-bit IEEE 754 formats. If the forcesci parameter is true, the string will always be printed using scientific notation.

Parameters

host	is the given value already converted to the host's double format.
forcesci	is true if the value should be printed in scientific notation.

Returns: the decimal representation as a string

References decimalMaxPrecision, decimalMinPrecision, and size.

Referenced by ghidra::PrintC::push_float().

◆ roundToNearestEven()

bool ghidra::FloatFormat::roundToNearestEven	(	uintb &	signif,
		int4	lowbitpos
	)

staticprivate

Round a floating point value to the nearest even.

Parameters

signif	the significant bits of a floating point value
lowbitpos	the position in signif of the floating point

Returns: true if we rounded up

Referenced by convertEncoding(), and getEncoding().

◆ setExponentCode()

uintb ghidra::FloatFormat::setExponentCode	(	uintb	x,
		uintb	code
	)		const

private

Set the exponent of an encoded value.

Parameters

x	is an encoded value (with exponent set to zero)
code	is the exponent to set

Returns: the encoded value with the new exponent

References exp_pos.

Referenced by convertEncoding(), getEncoding(), getInfinityEncoding(), getNaNEncoding(), and getZeroEncoding().

◆ setFractionalCode()

uintb ghidra::FloatFormat::setFractionalCode	(	uintb	x,
		uintb	code
	)		const

private

Set the fractional part of an encoded value.

Parameters

x	is an encoded value (with fraction part set to zero)
code	is the new fractional value to set

Returns: the encoded value with the fractional filled in

References frac_pos, and frac_size.

Referenced by convertEncoding(), getEncoding(), getInfinityEncoding(), getNaNEncoding(), and getZeroEncoding().

◆ setSign()

uintb ghidra::FloatFormat::setSign	(	uintb	x,
		bool	sign
	)		const

private

Set the sign bit of an encoded value.

Parameters

x	is an encoded value (with sign set to zero)
sign	is the sign bit to set

Returns: the encoded value with the sign bit set

References signbit_pos.

Referenced by convertEncoding(), getEncoding(), getInfinityEncoding(), getNaNEncoding(), and getZeroEncoding().

The documentation for this class was generated from the following files:

/builds/appsec/reveng-binrw/ghidraplusplus/src/core/float.hh
/builds/appsec/reveng-binrw/ghidraplusplus/src/core/float.cc

Public Types

Public Member Functions

Private Member Functions

Static Private Member Functions

Private Attributes

Detailed Description

Member Enumeration Documentation

◆ floatclass

Constructor & Destructor Documentation

◆ FloatFormat()

Member Function Documentation

◆ convertEncoding()

◆ createFloat()

◆ extractExponentCode()

◆ extractExpSig()

◆ extractFractionalCode()

◆ extractSign()

◆ getEncoding()

◆ getHostFloat()

◆ getInfinityEncoding()

◆ getNaNEncoding()

◆ getZeroEncoding()

◆ opAbs()

◆ opAdd()

◆ opCeil()

◆ opDiv()

◆ opEqual()

◆ opFloat2Float()

◆ opFloor()

◆ opInt2Float()

◆ opLess()

◆ opLessEqual()

◆ opMult()

◆ opNan()

◆ opNeg()

◆ opNotEqual()

◆ opRound()

◆ opSqrt()

◆ opSub()

◆ opTrunc()

◆ printDecimal()

◆ roundToNearestEven()

◆ setExponentCode()

◆ setFractionalCode()

◆ setSign()