|
1 /* This Source Code Form is subject to the terms of the Mozilla Public |
|
2 * License, v. 2.0. If a copy of the MPL was not distributed with this |
|
3 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
|
4 |
|
5 #include "ecp_fp.h" |
|
6 #include <stdlib.h> |
|
7 |
|
8 #define ECFP_BSIZE 192 |
|
9 #define ECFP_NUMDOUBLES 8 |
|
10 |
|
11 #include "ecp_fpinc.c" |
|
12 |
|
13 /* Performs a single step of reduction, just on the uppermost float |
|
14 * (assumes already tidied), and then retidies. Note, this does not |
|
15 * guarantee that the result will be less than p. */ |
|
16 void |
|
17 ecfp192_singleReduce(double *d, const EC_group_fp * group) |
|
18 { |
|
19 double q; |
|
20 |
|
21 ECFP_ASSERT(group->doubleBitSize == 24); |
|
22 ECFP_ASSERT(group->primeBitSize == 192); |
|
23 ECFP_ASSERT(group->numDoubles == 8); |
|
24 |
|
25 q = d[ECFP_NUMDOUBLES - 1] - ecfp_beta_192; |
|
26 q += group->bitSize_alpha; |
|
27 q -= group->bitSize_alpha; |
|
28 |
|
29 d[ECFP_NUMDOUBLES - 1] -= q; |
|
30 d[0] += q * ecfp_twom192; |
|
31 d[2] += q * ecfp_twom128; |
|
32 ecfp_positiveTidy(d, group); |
|
33 } |
|
34 |
|
35 /* |
|
36 * Performs imperfect reduction. This might leave some negative terms, |
|
37 * and one more reduction might be required for the result to be between 0 |
|
38 * and p-1. x should be be an array of at least 16, and r at least 8 x and |
|
39 * r can be the same, but then the upper parts of r are not zeroed */ |
|
40 void |
|
41 ecfp_reduce_192(double *r, double *x, const EC_group_fp * group) |
|
42 { |
|
43 double x8, x9, x10, q; |
|
44 |
|
45 ECFP_ASSERT(group->doubleBitSize == 24); |
|
46 ECFP_ASSERT(group->primeBitSize == 192); |
|
47 ECFP_ASSERT(group->numDoubles == 8); |
|
48 |
|
49 /* Tidy just the upper portion, the lower part can wait */ |
|
50 ecfp_tidyUpper(x, group); |
|
51 |
|
52 x8 = x[8] + x[14] * ecfp_twom128; /* adds bits 16-40 */ |
|
53 x9 = x[9] + x[15] * ecfp_twom128; /* adds bits 16-40 */ |
|
54 |
|
55 /* Tidy up, or we won't have enough bits later to add it in */ |
|
56 |
|
57 q = x8 + group->alpha[9]; |
|
58 q -= group->alpha[9]; |
|
59 x8 -= q; |
|
60 x9 += q; |
|
61 |
|
62 q = x9 + group->alpha[10]; |
|
63 q -= group->alpha[10]; |
|
64 x9 -= q; |
|
65 x10 = x[10] + q; |
|
66 |
|
67 r[7] = x[7] + x[15] * ecfp_twom192 + x[13] * ecfp_twom128; /* adds |
|
68 * bits |
|
69 * 0-40 */ |
|
70 r[6] = x[6] + x[14] * ecfp_twom192 + x[12] * ecfp_twom128; |
|
71 r[5] = x[5] + x[13] * ecfp_twom192 + x[11] * ecfp_twom128; |
|
72 r[4] = x[4] + x[12] * ecfp_twom192 + x10 * ecfp_twom128; |
|
73 r[3] = x[3] + x[11] * ecfp_twom192 + x9 * ecfp_twom128; /* adds bits |
|
74 * 0-40 */ |
|
75 r[2] = x[2] + x10 * ecfp_twom192 + x8 * ecfp_twom128; |
|
76 r[1] = x[1] + x9 * ecfp_twom192; /* adds bits 16-40 */ |
|
77 r[0] = x[0] + x8 * ecfp_twom192; |
|
78 |
|
79 /* |
|
80 * Tidy up just r[group->numDoubles-2] so that the number of |
|
81 * reductions is accurate plus or minus one. (Rather than tidy all to |
|
82 * make it totally accurate) */ |
|
83 q = r[ECFP_NUMDOUBLES - 2] + group->alpha[ECFP_NUMDOUBLES - 1]; |
|
84 q -= group->alpha[ECFP_NUMDOUBLES - 1]; |
|
85 r[ECFP_NUMDOUBLES - 2] -= q; |
|
86 r[ECFP_NUMDOUBLES - 1] += q; |
|
87 |
|
88 /* Tidy up the excess bits on r[group->numDoubles-1] using reduction */ |
|
89 /* Use ecfp_beta so we get a positive res */ |
|
90 q = r[ECFP_NUMDOUBLES - 1] - ecfp_beta_192; |
|
91 q += group->bitSize_alpha; |
|
92 q -= group->bitSize_alpha; |
|
93 |
|
94 r[ECFP_NUMDOUBLES - 1] -= q; |
|
95 r[0] += q * ecfp_twom192; |
|
96 r[2] += q * ecfp_twom128; |
|
97 |
|
98 /* Tidy the result */ |
|
99 ecfp_tidyShort(r, group); |
|
100 } |
|
101 |
|
102 /* Sets group to use optimized calculations in this file */ |
|
103 mp_err |
|
104 ec_group_set_nistp192_fp(ECGroup *group) |
|
105 { |
|
106 EC_group_fp *fpg; |
|
107 |
|
108 /* Allocate memory for floating point group data */ |
|
109 fpg = (EC_group_fp *) malloc(sizeof(EC_group_fp)); |
|
110 if (fpg == NULL) { |
|
111 return MP_MEM; |
|
112 } |
|
113 |
|
114 fpg->numDoubles = ECFP_NUMDOUBLES; |
|
115 fpg->primeBitSize = ECFP_BSIZE; |
|
116 fpg->orderBitSize = 192; |
|
117 fpg->doubleBitSize = 24; |
|
118 fpg->numInts = (ECFP_BSIZE + ECL_BITS - 1) / ECL_BITS; |
|
119 fpg->aIsM3 = 1; |
|
120 fpg->ecfp_singleReduce = &ecfp192_singleReduce; |
|
121 fpg->ecfp_reduce = &ecfp_reduce_192; |
|
122 fpg->ecfp_tidy = &ecfp_tidy; |
|
123 |
|
124 fpg->pt_add_jac_aff = &ecfp192_pt_add_jac_aff; |
|
125 fpg->pt_add_jac = &ecfp192_pt_add_jac; |
|
126 fpg->pt_add_jm_chud = &ecfp192_pt_add_jm_chud; |
|
127 fpg->pt_add_chud = &ecfp192_pt_add_chud; |
|
128 fpg->pt_dbl_jac = &ecfp192_pt_dbl_jac; |
|
129 fpg->pt_dbl_jm = &ecfp192_pt_dbl_jm; |
|
130 fpg->pt_dbl_aff2chud = &ecfp192_pt_dbl_aff2chud; |
|
131 fpg->precompute_chud = &ecfp192_precompute_chud; |
|
132 fpg->precompute_jac = &ecfp192_precompute_jac; |
|
133 |
|
134 group->point_mul = &ec_GFp_point_mul_wNAF_fp; |
|
135 group->points_mul = &ec_pts_mul_basic; |
|
136 group->extra1 = fpg; |
|
137 group->extra_free = &ec_GFp_extra_free_fp; |
|
138 |
|
139 ec_set_fp_precision(fpg); |
|
140 fpg->bitSize_alpha = ECFP_TWO192 * fpg->alpha[0]; |
|
141 |
|
142 return MP_OKAY; |
|
143 } |