MastersThesis/PQ_TIIGER_TLS/sal/miracl-winx86-11-04-24/includes/fp16_BLS48556.h

309 lines
8.4 KiB
C
Raw Permalink Normal View History

2024-04-15 09:53:30 +00:00
/*
* Copyright (c) 2012-2020 MIRACL UK Ltd.
*
* This file is part of MIRACL Core
* (see https://github.com/miracl/core).
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef FP16_BLS48556_H
#define FP16_BLS48556_H
#include "fp8_BLS48556.h"
#include "config_curve_BLS48556.h"
using namespace core;
namespace BLS48556 {
/**
@brief FP16 Structure - towered over two FP4
*/
typedef struct
{
FP8 a; /**< real part of FP16 */
FP8 b; /**< imaginary part of FP16 */
} FP16;
/* FP16 prototypes */
/** @brief Tests for FP16 equal to zero
*
@param x FP16 number to be tested
@return 1 if zero, else returns 0
*/
extern int FP16_iszilch(FP16 *x);
/** @brief Tests for FP16 equal to unity
*
@param x FP16 number to be tested
@return 1 if unity, else returns 0
*/
extern int FP16_isunity(FP16 *x);
/** @brief Tests for equality of two FP16s
*
@param x FP16 instance to be compared
@param y FP16 instance to be compared
@return 1 if x=y, else returns 0
*/
extern int FP16_equals(FP16 *x, FP16 *y);
/** @brief Serialize in FP16
*
@param b buffer for output
@param x FP16 number to be serialized
*/
extern void FP16_toBytes(char *b,FP16 *x);
/** @brief Serialize out FP16
*
@param x FP16 number to be serialized
@param b buffer for input
*/
extern void FP16_fromBytes(FP16 *x,char *b);
/** @brief Tests for FP16 having only a real part and no imaginary part
*
@param x FP16 number to be tested
@return 1 if real, else returns 0
*/
extern int FP16_isreal(FP16 *x);
/** @brief Initialise FP16 from two FP8s
*
@param x FP16 instance to be initialised
@param a FP8 to form real part of FP16
@param b FP8 to form imaginary part of FP16
*/
extern void FP16_from_FP8s(FP16 *x, FP8 *a, FP8 *b);
/** @brief Initialise FP16 from single FP8
*
Imaginary part is set to zero
@param x FP16 instance to be initialised
@param a FP8 to form real part of FP16
*/
extern void FP16_from_FP8(FP16 *x, FP8 *a);
/** @brief Initialise FP16 from single FP8
*
real part is set to zero
@param x FP16 instance to be initialised
@param a FP8 to form imaginary part of FP16
*/
extern void FP16_from_FP8H(FP16 *x, FP8 *a);
/** @brief Copy FP16 to another FP16
*
@param x FP16 instance, on exit = y
@param y FP16 instance to be copied
*/
extern void FP16_copy(FP16 *x, FP16 *y);
/** @brief Set FP16 to zero
*
@param x FP16 instance to be set to zero
*/
extern void FP16_zero(FP16 *x);
/** @brief Set FP16 to unity
*
@param x FP16 instance to be set to one
*/
extern void FP16_one(FP16 *x);
/** @brief Negation of FP16
*
@param x FP16 instance, on exit = -y
@param y FP16 instance
*/
extern void FP16_neg(FP16 *x, FP16 *y);
/** @brief Conjugation of FP16
*
If y=(a,b) on exit x=(a,-b)
@param x FP16 instance, on exit = conj(y)
@param y FP16 instance
*/
extern void FP16_conj(FP16 *x, FP16 *y);
/** @brief Negative conjugation of FP16
*
If y=(a,b) on exit x=(-a,b)
@param x FP16 instance, on exit = -conj(y)
@param y FP16 instance
*/
extern void FP16_nconj(FP16 *x, FP16 *y);
/** @brief addition of two FP16s
*
@param x FP16 instance, on exit = y+z
@param y FP16 instance
@param z FP16 instance
*/
extern void FP16_add(FP16 *x, FP16 *y, FP16 *z);
/** @brief subtraction of two FP16s
*
@param x FP16 instance, on exit = y-z
@param y FP16 instance
@param z FP16 instance
*/
extern void FP16_sub(FP16 *x, FP16 *y, FP16 *z);
/** @brief Multiplication of an FP16 by an FP8
*
@param x FP16 instance, on exit = y*a
@param y FP16 instance
@param a FP4 multiplier
*/
extern void FP16_pmul(FP16 *x, FP16 *y, FP8 *a);
/** @brief Multiplication of an FP16 by an FP2
*
@param x FP16 instance, on exit = y*a
@param y FP16 instance
@param a FP2 multiplier
*/
extern void FP16_qmul(FP16 *x, FP16 *y, FP2 *a);
/** @brief Multiplication of an FP16 by an FP
*
@param x FP16 instance, on exit = y*a
@param y FP16 instance
@param a FP multiplier
*/
extern void FP16_tmul(FP16 *x, FP16 *y, FP *a);
/** @brief Multiplication of an FP16 by a small integer
*
@param x FP16 instance, on exit = y*i
@param y FP16 instance
@param i an integer
*/
extern void FP16_imul(FP16 *x, FP16 *y, int i);
/** @brief Squaring an FP16
*
@param x FP16 instance, on exit = y^2
@param y FP16 instance
*/
extern void FP16_sqr(FP16 *x, FP16 *y);
/** @brief Multiplication of two FP16s
*
@param x FP16 instance, on exit = y*z
@param y FP16 instance
@param z FP16 instance
*/
extern void FP16_mul(FP16 *x, FP16 *y, FP16 *z);
/** @brief Inverting an FP16
*
@param x FP16 instance, on exit = 1/y
@param y FP16 instance
*/
extern void FP16_inv(FP16 *x, FP16 *y);
/** @brief Formats and outputs an FP16 to the console
*
@param x FP16 instance to be printed
*/
extern void FP16_output(FP16 *x);
/** @brief Formats and outputs an FP16 to the console in raw form (for debugging)
*
@param x FP16 instance to be printed
*/
extern void FP16_rawoutput(FP16 *x);
/** @brief multiplies an FP16 instance by irreducible polynomial sqrt(1+sqrt(-1))
*
@param x FP16 instance, on exit = sqrt(1+sqrt(-1)*x
*/
extern void FP16_times_i(FP16 *x);
/** @brief multiplies an FP16 instance by irreducible polynomial (1+sqrt(-1))
*
@param x FP16 instance, on exit = sqrt(1+sqrt(-1))^2*x
*/
extern void FP16_times_i2(FP16 *x);
/** @brief multiplies an FP16 instance by irreducible polynomial (1+sqrt(-1))
*
@param x FP16 instance, on exit = sqrt(1+sqrt(-1))^4*x
*/
extern void FP16_times_i4(FP16 *x);
/** @brief Normalises the components of an FP16
*
@param x FP16 instance to be normalised
*/
extern void FP16_norm(FP16 *x);
/** @brief Reduces all components of possibly unreduced FP16 mod Modulus
*
@param x FP16 instance, on exit reduced mod Modulus
*/
extern void FP16_reduce(FP16 *x);
/** @brief Raises an FP16 to the power of a BIG
*
@param x FP16 instance, on exit = y^b
@param y FP16 instance
@param b BIG number
*/
extern void FP16_pow(FP16 *x, FP16 *y, B560_29::BIG b);
/** @brief Raises an FP16 to the power of the internal modulus p, using the Frobenius
*
@param x FP16 instance, on exit = x^p
@param f FP2 precalculated Frobenius constant
*/
extern void FP16_frob(FP16 *x, FP2 *f);
/** @brief Calculates the XTR addition function r=w*x-conj(x)*y+z
*
@param r FP16 instance, on exit = w*x-conj(x)*y+z
@param w FP16 instance
@param x FP16 instance
@param y FP16 instance
@param z FP16 instance
*/
extern void FP16_xtr_A(FP16 *r, FP16 *w, FP16 *x, FP16 *y, FP16 *z);
/** @brief Calculates the XTR doubling function r=x^2-2*conj(x)
*
@param r FP16 instance, on exit = x^2-2*conj(x)
@param x FP16 instance
*/
extern void FP16_xtr_D(FP16 *r, FP16 *x);
/** @brief Calculates FP16 trace of an FP12 raised to the power of a BIG number
*
XTR single exponentiation
@param r FP16 instance, on exit = trace(w^b)
@param x FP16 instance, trace of an FP12 w
@param b BIG number
*/
extern void FP16_xtr_pow(FP16 *r, FP16 *x, B560_29::BIG b);
/** @brief Calculates FP16 trace of c^a.d^b, where c and d are derived from FP16 traces of FP12s
*
XTR double exponentiation
Assumes c=tr(x^m), d=tr(x^n), e=tr(x^(m-n)), f=tr(x^(m-2n))
@param r FP16 instance, on exit = trace(c^a.d^b)
@param c FP16 instance, trace of an FP12
@param d FP16 instance, trace of an FP12
@param e FP16 instance, trace of an FP12
@param f FP16 instance, trace of an FP12
@param a BIG number
@param b BIG number
*/
extern void FP16_xtr_pow2(FP16 *r, FP16 *c, FP16 *d, FP16 *e, FP16 *f, B560_29::BIG a, B560_29::BIG b);
/** @brief Conditional copy of FP16 number
*
Conditionally copies second parameter to the first (without branching)
@param x FP16 instance, set to y if s!=0
@param y another FP16 instance
@param s copy only takes place if not equal to 0
*/
extern void FP16_cmove(FP16 *x, FP16 *y, int s);
}
#endif