|
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 224 |
|
9 #define ECFP_NUMDOUBLES 10 |
|
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 ecfp224_singleReduce(double *r, const EC_group_fp * group) |
|
18 { |
|
19 double q; |
|
20 |
|
21 ECFP_ASSERT(group->doubleBitSize == 24); |
|
22 ECFP_ASSERT(group->primeBitSize == 224); |
|
23 ECFP_ASSERT(group->numDoubles == 10); |
|
24 |
|
25 q = r[ECFP_NUMDOUBLES - 1] - ecfp_beta_224; |
|
26 q += group->bitSize_alpha; |
|
27 q -= group->bitSize_alpha; |
|
28 |
|
29 r[ECFP_NUMDOUBLES - 1] -= q; |
|
30 r[0] -= q * ecfp_twom224; |
|
31 r[4] += q * ecfp_twom128; |
|
32 |
|
33 ecfp_positiveTidy(r, group); |
|
34 } |
|
35 |
|
36 /* |
|
37 * Performs imperfect reduction. This might leave some negative terms, |
|
38 * and one more reduction might be required for the result to be between 0 |
|
39 * and p-1. x should be be an array of at least 20, and r at least 10 x |
|
40 * and r can be the same, but then the upper parts of r are not zeroed */ |
|
41 void |
|
42 ecfp224_reduce(double *r, double *x, const EC_group_fp * group) |
|
43 { |
|
44 |
|
45 double x10, x11, x12, x13, x14, q; |
|
46 |
|
47 ECFP_ASSERT(group->doubleBitSize == 24); |
|
48 ECFP_ASSERT(group->primeBitSize == 224); |
|
49 ECFP_ASSERT(group->numDoubles == 10); |
|
50 |
|
51 /* Tidy just the upper bits of x. Don't need to tidy the lower ones |
|
52 * yet. */ |
|
53 ecfp_tidyUpper(x, group); |
|
54 |
|
55 x10 = x[10] + x[16] * ecfp_twom128; |
|
56 x11 = x[11] + x[17] * ecfp_twom128; |
|
57 x12 = x[12] + x[18] * ecfp_twom128; |
|
58 x13 = x[13] + x[19] * ecfp_twom128; |
|
59 |
|
60 /* Tidy up, or we won't have enough bits later to add it in */ |
|
61 q = x10 + group->alpha[11]; |
|
62 q -= group->alpha[11]; |
|
63 x10 -= q; |
|
64 x11 = x11 + q; |
|
65 |
|
66 q = x11 + group->alpha[12]; |
|
67 q -= group->alpha[12]; |
|
68 x11 -= q; |
|
69 x12 = x12 + q; |
|
70 |
|
71 q = x12 + group->alpha[13]; |
|
72 q -= group->alpha[13]; |
|
73 x12 -= q; |
|
74 x13 = x13 + q; |
|
75 |
|
76 q = x13 + group->alpha[14]; |
|
77 q -= group->alpha[14]; |
|
78 x13 -= q; |
|
79 x14 = x[14] + q; |
|
80 |
|
81 r[9] = x[9] + x[15] * ecfp_twom128 - x[19] * ecfp_twom224; |
|
82 r[8] = x[8] + x14 * ecfp_twom128 - x[18] * ecfp_twom224; |
|
83 r[7] = x[7] + x13 * ecfp_twom128 - x[17] * ecfp_twom224; |
|
84 r[6] = x[6] + x12 * ecfp_twom128 - x[16] * ecfp_twom224; |
|
85 r[5] = x[5] + x11 * ecfp_twom128 - x[15] * ecfp_twom224; |
|
86 r[4] = x[4] + x10 * ecfp_twom128 - x14 * ecfp_twom224; |
|
87 r[3] = x[3] - x13 * ecfp_twom224; |
|
88 r[2] = x[2] - x12 * ecfp_twom224; |
|
89 r[1] = x[1] - x11 * ecfp_twom224; |
|
90 r[0] = x[0] - x10 * ecfp_twom224; |
|
91 |
|
92 /* |
|
93 * Tidy up just r[ECFP_NUMDOUBLES-2] so that the number of reductions |
|
94 * is accurate plus or minus one. (Rather than tidy all to make it |
|
95 * totally accurate) */ |
|
96 q = r[ECFP_NUMDOUBLES - 2] + group->alpha[ECFP_NUMDOUBLES - 1]; |
|
97 q -= group->alpha[ECFP_NUMDOUBLES - 1]; |
|
98 r[ECFP_NUMDOUBLES - 2] -= q; |
|
99 r[ECFP_NUMDOUBLES - 1] += q; |
|
100 |
|
101 /* Tidy up the excess bits on r[ECFP_NUMDOUBLES-1] using reduction */ |
|
102 /* Use ecfp_beta so we get a positive res */ |
|
103 q = r[ECFP_NUMDOUBLES - 1] - ecfp_beta_224; |
|
104 q += group->bitSize_alpha; |
|
105 q -= group->bitSize_alpha; |
|
106 |
|
107 r[ECFP_NUMDOUBLES - 1] -= q; |
|
108 r[0] -= q * ecfp_twom224; |
|
109 r[4] += q * ecfp_twom128; |
|
110 |
|
111 ecfp_tidyShort(r, group); |
|
112 } |
|
113 |
|
114 /* Sets group to use optimized calculations in this file */ |
|
115 mp_err |
|
116 ec_group_set_nistp224_fp(ECGroup *group) |
|
117 { |
|
118 |
|
119 EC_group_fp *fpg; |
|
120 |
|
121 /* Allocate memory for floating point group data */ |
|
122 fpg = (EC_group_fp *) malloc(sizeof(EC_group_fp)); |
|
123 if (fpg == NULL) { |
|
124 return MP_MEM; |
|
125 } |
|
126 |
|
127 fpg->numDoubles = ECFP_NUMDOUBLES; |
|
128 fpg->primeBitSize = ECFP_BSIZE; |
|
129 fpg->orderBitSize = 224; |
|
130 fpg->doubleBitSize = 24; |
|
131 fpg->numInts = (ECFP_BSIZE + ECL_BITS - 1) / ECL_BITS; |
|
132 fpg->aIsM3 = 1; |
|
133 fpg->ecfp_singleReduce = &ecfp224_singleReduce; |
|
134 fpg->ecfp_reduce = &ecfp224_reduce; |
|
135 fpg->ecfp_tidy = &ecfp_tidy; |
|
136 |
|
137 fpg->pt_add_jac_aff = &ecfp224_pt_add_jac_aff; |
|
138 fpg->pt_add_jac = &ecfp224_pt_add_jac; |
|
139 fpg->pt_add_jm_chud = &ecfp224_pt_add_jm_chud; |
|
140 fpg->pt_add_chud = &ecfp224_pt_add_chud; |
|
141 fpg->pt_dbl_jac = &ecfp224_pt_dbl_jac; |
|
142 fpg->pt_dbl_jm = &ecfp224_pt_dbl_jm; |
|
143 fpg->pt_dbl_aff2chud = &ecfp224_pt_dbl_aff2chud; |
|
144 fpg->precompute_chud = &ecfp224_precompute_chud; |
|
145 fpg->precompute_jac = &ecfp224_precompute_jac; |
|
146 |
|
147 group->point_mul = &ec_GFp_point_mul_wNAF_fp; |
|
148 group->points_mul = &ec_pts_mul_basic; |
|
149 group->extra1 = fpg; |
|
150 group->extra_free = &ec_GFp_extra_free_fp; |
|
151 |
|
152 ec_set_fp_precision(fpg); |
|
153 fpg->bitSize_alpha = ECFP_TWO224 * fpg->alpha[0]; |
|
154 |
|
155 return MP_OKAY; |
|
156 } |