michael@0: /* This Source Code Form is subject to the terms of the Mozilla Public
michael@0:  * License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0:  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
michael@0: 
michael@0: #include "ecp_fp.h"
michael@0: #include <stdlib.h>
michael@0: 
michael@0: #define ECFP_BSIZE 192
michael@0: #define ECFP_NUMDOUBLES 8
michael@0: 
michael@0: #include "ecp_fpinc.c"
michael@0: 
michael@0: /* Performs a single step of reduction, just on the uppermost float
michael@0:  * (assumes already tidied), and then retidies. Note, this does not
michael@0:  * guarantee that the result will be less than p. */
michael@0: void
michael@0: ecfp192_singleReduce(double *d, const EC_group_fp * group)
michael@0: {
michael@0: 	double q;
michael@0: 
michael@0: 	ECFP_ASSERT(group->doubleBitSize == 24);
michael@0: 	ECFP_ASSERT(group->primeBitSize == 192);
michael@0: 	ECFP_ASSERT(group->numDoubles == 8);
michael@0: 
michael@0: 	q = d[ECFP_NUMDOUBLES - 1] - ecfp_beta_192;
michael@0: 	q += group->bitSize_alpha;
michael@0: 	q -= group->bitSize_alpha;
michael@0: 
michael@0: 	d[ECFP_NUMDOUBLES - 1] -= q;
michael@0: 	d[0] += q * ecfp_twom192;
michael@0: 	d[2] += q * ecfp_twom128;
michael@0: 	ecfp_positiveTidy(d, group);
michael@0: }
michael@0: 
michael@0: /* 
michael@0:  * Performs imperfect reduction.  This might leave some negative terms,
michael@0:  * and one more reduction might be required for the result to be between 0 
michael@0:  * and p-1. x should be be an array of at least 16, and r at least 8 x and 
michael@0:  * r can be the same, but then the upper parts of r are not zeroed */
michael@0: void
michael@0: ecfp_reduce_192(double *r, double *x, const EC_group_fp * group)
michael@0: {
michael@0: 	double x8, x9, x10, q;
michael@0: 
michael@0: 	ECFP_ASSERT(group->doubleBitSize == 24);
michael@0: 	ECFP_ASSERT(group->primeBitSize == 192);
michael@0: 	ECFP_ASSERT(group->numDoubles == 8);
michael@0: 
michael@0: 	/* Tidy just the upper portion, the lower part can wait */
michael@0: 	ecfp_tidyUpper(x, group);
michael@0: 
michael@0: 	x8 = x[8] + x[14] * ecfp_twom128;	/* adds bits 16-40 */
michael@0: 	x9 = x[9] + x[15] * ecfp_twom128;	/* adds bits 16-40 */
michael@0: 
michael@0: 	/* Tidy up, or we won't have enough bits later to add it in */
michael@0: 
michael@0: 	q = x8 + group->alpha[9];
michael@0: 	q -= group->alpha[9];
michael@0: 	x8 -= q;
michael@0: 	x9 += q;
michael@0: 
michael@0: 	q = x9 + group->alpha[10];
michael@0: 	q -= group->alpha[10];
michael@0: 	x9 -= q;
michael@0: 	x10 = x[10] + q;
michael@0: 
michael@0: 	r[7] = x[7] + x[15] * ecfp_twom192 + x[13] * ecfp_twom128;	/* adds
michael@0: 																 * bits
michael@0: 																 * 0-40 */
michael@0: 	r[6] = x[6] + x[14] * ecfp_twom192 + x[12] * ecfp_twom128;
michael@0: 	r[5] = x[5] + x[13] * ecfp_twom192 + x[11] * ecfp_twom128;
michael@0: 	r[4] = x[4] + x[12] * ecfp_twom192 + x10 * ecfp_twom128;
michael@0: 	r[3] = x[3] + x[11] * ecfp_twom192 + x9 * ecfp_twom128;	/* adds bits
michael@0: 															 * 0-40 */
michael@0: 	r[2] = x[2] + x10 * ecfp_twom192 + x8 * ecfp_twom128;
michael@0: 	r[1] = x[1] + x9 * ecfp_twom192;	/* adds bits 16-40 */
michael@0: 	r[0] = x[0] + x8 * ecfp_twom192;
michael@0: 
michael@0: 	/* 
michael@0: 	 * Tidy up just r[group->numDoubles-2] so that the number of
michael@0: 	 * reductions is accurate plus or minus one.  (Rather than tidy all to 
michael@0: 	 * make it totally accurate) */
michael@0: 	q = r[ECFP_NUMDOUBLES - 2] + group->alpha[ECFP_NUMDOUBLES - 1];
michael@0: 	q -= group->alpha[ECFP_NUMDOUBLES - 1];
michael@0: 	r[ECFP_NUMDOUBLES - 2] -= q;
michael@0: 	r[ECFP_NUMDOUBLES - 1] += q;
michael@0: 
michael@0: 	/* Tidy up the excess bits on r[group->numDoubles-1] using reduction */
michael@0: 	/* Use ecfp_beta so we get a positive res */
michael@0: 	q = r[ECFP_NUMDOUBLES - 1] - ecfp_beta_192;
michael@0: 	q += group->bitSize_alpha;
michael@0: 	q -= group->bitSize_alpha;
michael@0: 
michael@0: 	r[ECFP_NUMDOUBLES - 1] -= q;
michael@0: 	r[0] += q * ecfp_twom192;
michael@0: 	r[2] += q * ecfp_twom128;
michael@0: 
michael@0: 	/* Tidy the result */
michael@0: 	ecfp_tidyShort(r, group);
michael@0: }
michael@0: 
michael@0: /* Sets group to use optimized calculations in this file */
michael@0: mp_err
michael@0: ec_group_set_nistp192_fp(ECGroup *group)
michael@0: {
michael@0: 	EC_group_fp *fpg;
michael@0: 
michael@0: 	/* Allocate memory for floating point group data */
michael@0: 	fpg = (EC_group_fp *) malloc(sizeof(EC_group_fp));
michael@0: 	if (fpg == NULL) {
michael@0: 		return MP_MEM;
michael@0: 	}
michael@0: 
michael@0: 	fpg->numDoubles = ECFP_NUMDOUBLES;
michael@0: 	fpg->primeBitSize = ECFP_BSIZE;
michael@0: 	fpg->orderBitSize = 192;
michael@0: 	fpg->doubleBitSize = 24;
michael@0: 	fpg->numInts = (ECFP_BSIZE + ECL_BITS - 1) / ECL_BITS;
michael@0: 	fpg->aIsM3 = 1;
michael@0: 	fpg->ecfp_singleReduce = &ecfp192_singleReduce;
michael@0: 	fpg->ecfp_reduce = &ecfp_reduce_192;
michael@0: 	fpg->ecfp_tidy = &ecfp_tidy;
michael@0: 
michael@0: 	fpg->pt_add_jac_aff = &ecfp192_pt_add_jac_aff;
michael@0: 	fpg->pt_add_jac = &ecfp192_pt_add_jac;
michael@0: 	fpg->pt_add_jm_chud = &ecfp192_pt_add_jm_chud;
michael@0: 	fpg->pt_add_chud = &ecfp192_pt_add_chud;
michael@0: 	fpg->pt_dbl_jac = &ecfp192_pt_dbl_jac;
michael@0: 	fpg->pt_dbl_jm = &ecfp192_pt_dbl_jm;
michael@0: 	fpg->pt_dbl_aff2chud = &ecfp192_pt_dbl_aff2chud;
michael@0: 	fpg->precompute_chud = &ecfp192_precompute_chud;
michael@0: 	fpg->precompute_jac = &ecfp192_precompute_jac;
michael@0: 
michael@0: 	group->point_mul = &ec_GFp_point_mul_wNAF_fp;
michael@0: 	group->points_mul = &ec_pts_mul_basic;
michael@0: 	group->extra1 = fpg;
michael@0: 	group->extra_free = &ec_GFp_extra_free_fp;
michael@0: 
michael@0: 	ec_set_fp_precision(fpg);
michael@0: 	fpg->bitSize_alpha = ECFP_TWO192 * fpg->alpha[0];
michael@0: 
michael@0: 	return MP_OKAY;
michael@0: }