hpdyad.h File Reference

High-precision dyadic number operations. More...

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	hpdyad_t
	Fixed-size high-precision dyadic number (hpdyad) with separate sign and binary exponent. More...

Macros
#define	HPDYAD_MAX_LIMBS   32
	Maximum number of 32-bit limbs in fixed-size representation (1024 bits total).

Functions
unsigned	ctz64 (unsigned long long x)
	Count trailing zero bits in a 64-bit unsigned integer.
int	msb32 (unsigned int x)
	Find index of most significant bit in a 32-bit unsigned integer.
void	hpdyad_set_ull (hpdyad_t *a, unsigned long long x, signed char sign)
	Initialize hpdyad from unsigned long long with sign.
void	hpdyad_normalize (hpdyad_t *a)
	Normalize hpdyad: trim leading zeros and left-shift mantissa so MSB of top limb is 1.
void	hpdyad_mul (hpdyad_t *out, const hpdyad_t *a, const hpdyad_t *b)
	Multiply two high precision dyadic numbers (hpdyad's): out = a * b.
void	hpdyad_shr_bits_sticky (hpdyad_t *dst, const hpdyad_t *src, unsigned int bits)
	Right-shift mantissa with sticky bit (value-preserving)
void	hpdyad_shl_bits (hpdyad_t *dst, const hpdyad_t *src, int bits)
	Left-shift mantissa by specified number of bits (value-preserving)
void	hpdyad_add (hpdyad_t *out, const hpdyad_t *a, const hpdyad_t *b)
	Add two high precision dyadic numbers (hpdyad's): out = a + b.
double	hpdyad_to_double (const hpdyad_t *a)
	Convert hpdyad to double with round-to-nearest-even.

Detailed Description

High-precision dyadic number operations.

Function Documentation

ctz64()

unsigned ctz64

(

unsigned long long

x

)

Count trailing zero bits in a 64-bit unsigned integer.

Parameters

[in]

x

input value

Returns

number of trailing zeros (0..63); returns 64 if x == 0

hpdyad_add()

void hpdyad_add	(	hpdyad_t *	out,
		const hpdyad_t *	a,
		const hpdyad_t *	b )

Add two high precision dyadic numbers (hpdyad's): out = a + b.

Parameters

[out]	out	result (supports aliasing)
[in]	a	first operand
[in]	b	second operand

hpdyad_mul()

void hpdyad_mul	(	hpdyad_t *	out,
		const hpdyad_t *	a,
		const hpdyad_t *	b )

Multiply two high precision dyadic numbers (hpdyad's): out = a * b.

Parameters

[out]	out	result (may alias a or b)
[in]	a	first operand
[in]	b	second operand

hpdyad_normalize()

void hpdyad_normalize

(

hpdyad_t *

a

)

Normalize hpdyad: trim leading zeros and left-shift mantissa so MSB of top limb is 1.

Parameters

[in,out]

a

pointer to hpdyad to normalize

Returns

void

hpdyad_set_ull()

void hpdyad_set_ull	(	hpdyad_t *	a,
		unsigned long long	x,
		signed char	sign )

Initialize hpdyad from unsigned long long with sign.

Parameters

[out]	a	destination hpdyad
[in]	x	unsigned integer value
[in]	sign	+1 or -1

hpdyad_shl_bits()

void hpdyad_shl_bits	(	hpdyad_t *	dst,
		const hpdyad_t *	src,
		int	bits )

Left-shift mantissa by specified number of bits (value-preserving)

Parameters

[out]	dst	destination hpdyad
[in]	src	source hpdyad
[in]	bits	number of bits to shift left

Returns

void

hpdyad_shr_bits_sticky()

void hpdyad_shr_bits_sticky	(	hpdyad_t *	dst,
		const hpdyad_t *	src,
		unsigned int	bits )

Right-shift mantissa with sticky bit (value-preserving)

Shifts mantissa right by bits, increments exp2 by bits to preserve value. Any 1-bit shifted out ORs into bit 0 (sticky). Does NOT normalize.

Parameters

[out]	dst	destination hpdyad
[in]	src	source hpdyad
[in]	bits	number of bits to shift right

hpdyad_to_double()

double hpdyad_to_double

(

const hpdyad_t *

a

)

Convert hpdyad to double with round-to-nearest-even.

Parameters

[in]

a

pointer to normalized hpdyad

Returns

closest double representation; ±DBL_MAX on overflow

msb32()

int msb32

(

unsigned int

x

)

Find index of most significant bit in a 32-bit unsigned integer.

Parameters

[in]

x

input value

Returns

bit index 0..31; returns 0 for x == 0 (defensive fallback)