2 * @file ed448goldilocks/scalar.c
6 * Copyright (c) 2015-2016 Cryptography Research, Inc. \n
7 * Released under the MIT License. See LICENSE.txt for license information.
9 * @brief Decaf high-level functions.
11 * @warning This file was automatically generated in Python.
12 * Please do not edit it.
15 #include "constant_time.h"
19 #define API_NS(_id) decaf_448_##_id
20 #define SCALAR_BITS DECAF_448_SCALAR_BITS
21 #define SCALAR_SER_BYTES DECAF_448_SCALAR_BYTES
22 #define SCALAR_LIMBS DECAF_448_SCALAR_LIMBS
23 #define scalar_t API_NS(scalar_t)
25 static const decaf_word_t MONTGOMERY_FACTOR = (decaf_word_t)0x3bd440fae918bc5ull;
26 static const scalar_t sc_p = {{{
27 SC_LIMB(0x2378c292ab5844f3), SC_LIMB(0x216cc2728dc58f55), SC_LIMB(0xc44edb49aed63690), SC_LIMB(0xffffffff7cca23e9), SC_LIMB(0xffffffffffffffff), SC_LIMB(0xffffffffffffffff), SC_LIMB(0x3fffffffffffffff)
29 SC_LIMB(0xe3539257049b9b60), SC_LIMB(0x7af32c4bc1b195d9), SC_LIMB(0x0d66de2388ea1859), SC_LIMB(0xae17cf725ee4d838), SC_LIMB(0x1a9cc14ba3c47c44), SC_LIMB(0x2052bcb7e4d070af), SC_LIMB(0x3402a939f823b729)
31 /* End of template stuff */
33 #define WBITS DECAF_WORD_BITS /* NB this may be different from ARCH_WORD_BITS */
35 const scalar_t API_NS(scalar_one) = {{{1}}}, API_NS(scalar_zero) = {{{0}}};
37 /** {extra,accum} - sub +? p
38 * Must have extra <= 1
40 static DECAF_NOINLINE void sc_subx(
42 const decaf_word_t accum[SCALAR_LIMBS],
47 decaf_dsword_t chain = 0;
49 for (i=0; i<SCALAR_LIMBS; i++) {
50 chain = (chain + accum[i]) - sub->limb[i];
54 decaf_word_t borrow = chain+extra; /* = 0 or -1 */
57 for (i=0; i<SCALAR_LIMBS; i++) {
58 chain = (chain + out->limb[i]) + (p->limb[i] & borrow);
64 static DECAF_NOINLINE void sc_montmul (
70 decaf_word_t accum[SCALAR_LIMBS+1] = {0};
71 decaf_word_t hi_carry = 0;
73 for (i=0; i<SCALAR_LIMBS; i++) {
74 decaf_word_t mand = a->limb[i];
75 const decaf_word_t *mier = b->limb;
77 decaf_dword_t chain = 0;
78 for (j=0; j<SCALAR_LIMBS; j++) {
79 chain += ((decaf_dword_t)mand)*mier[j] + accum[j];
85 mand = accum[0] * MONTGOMERY_FACTOR;
88 for (j=0; j<SCALAR_LIMBS; j++) {
89 chain += (decaf_dword_t)mand*mier[j] + accum[j];
90 if (j) accum[j-1] = chain;
96 hi_carry = chain >> WBITS;
99 sc_subx(out, accum, sc_p, sc_p, hi_carry);
102 void API_NS(scalar_mul) (
108 sc_montmul(out,out,sc_r2);
111 /* PERF: could implement this */
112 static DECAF_INLINE void sc_montsqr (scalar_t out, const scalar_t a) {
116 decaf_error_t API_NS(scalar_invert) (
120 /* Fermat's little theorem, sliding window.
121 * Sliding window is fine here because the modulus isn't secret.
123 const int SCALAR_WINDOW_BITS = 3;
124 scalar_t precmp[1<<SCALAR_WINDOW_BITS];
125 const int LAST = (1<<SCALAR_WINDOW_BITS)-1;
127 /* Precompute precmp = [a^1,a^3,...] */
128 sc_montmul(precmp[0],a,sc_r2);
129 if (LAST > 0) sc_montmul(precmp[LAST],precmp[0],precmp[0]);
132 for (i=1; i<=LAST; i++) {
133 sc_montmul(precmp[i],precmp[i-1],precmp[LAST]);
137 unsigned residue = 0, trailing = 0, started = 0;
138 for (i=SCALAR_BITS-1; i>=-SCALAR_WINDOW_BITS; i--) {
140 if (started) sc_montsqr(out,out);
142 decaf_word_t w = (i>=0) ? sc_p->limb[i/WBITS] : 0;
143 if (i >= 0 && i<WBITS) {
148 residue = (residue<<1) | ((w>>(i%WBITS))&1);
149 if (residue>>SCALAR_WINDOW_BITS != 0) {
150 assert(trailing == 0);
155 if (trailing > 0 && (trailing & ((1<<SCALAR_WINDOW_BITS)-1)) == 0) {
157 sc_montmul(out,out,precmp[trailing>>(SCALAR_WINDOW_BITS+1)]);
159 API_NS(scalar_copy)(out,precmp[trailing>>(SCALAR_WINDOW_BITS+1)]);
171 sc_montmul(out,out,API_NS(scalar_one));
172 decaf_bzero(precmp, sizeof(precmp));
173 return decaf_succeed_if(~API_NS(scalar_eq)(out,API_NS(scalar_zero)));
176 void API_NS(scalar_sub) (
181 sc_subx(out, a->limb, b, sc_p, 0);
184 void API_NS(scalar_add) (
189 decaf_dword_t chain = 0;
191 for (i=0; i<SCALAR_LIMBS; i++) {
192 chain = (chain + a->limb[i]) + b->limb[i];
193 out->limb[i] = chain;
196 sc_subx(out, out->limb, sc_p, sc_p, chain);
200 API_NS(scalar_set_unsigned) (
204 memset(out,0,sizeof(scalar_t));
206 for (; i<sizeof(uint64_t)/sizeof(decaf_word_t); i++) {
208 #if DECAF_WORD_BITS < 64
209 w >>= 8*sizeof(decaf_word_t);
219 decaf_word_t diff = 0;
221 for (i=0; i<SCALAR_LIMBS; i++) {
222 diff |= a->limb[i] ^ b->limb[i];
224 return mask_to_bool(word_is_zero(diff));
227 static DECAF_INLINE void scalar_decode_short (
229 const unsigned char *ser,
232 unsigned int i,j,k=0;
233 for (i=0; i<SCALAR_LIMBS; i++) {
234 decaf_word_t out = 0;
235 for (j=0; j<sizeof(decaf_word_t) && k<nbytes; j++,k++) {
236 out |= ((decaf_word_t)ser[k])<<(8*j);
242 decaf_error_t API_NS(scalar_decode)(
244 const unsigned char ser[SCALAR_SER_BYTES]
247 scalar_decode_short(s, ser, SCALAR_SER_BYTES);
248 decaf_dsword_t accum = 0;
249 for (i=0; i<SCALAR_LIMBS; i++) {
250 accum = (accum + s->limb[i] - sc_p->limb[i]) >> WBITS;
252 /* Here accum == 0 or -1 */
254 API_NS(scalar_mul)(s,s,API_NS(scalar_one)); /* ham-handed reduce */
256 return decaf_succeed_if(~word_is_zero(accum));
259 void API_NS(scalar_destroy) (
262 decaf_bzero(scalar, sizeof(scalar_t));
265 void API_NS(scalar_decode_long)(
267 const unsigned char *ser,
271 API_NS(scalar_copy)(s, API_NS(scalar_zero));
278 i = ser_len - (ser_len%SCALAR_SER_BYTES);
279 if (i==ser_len) i -= SCALAR_SER_BYTES;
281 scalar_decode_short(t1, &ser[i], ser_len-i);
283 if (ser_len == sizeof(scalar_t)) {
285 /* ham-handed reduce */
286 API_NS(scalar_mul)(s,t1,API_NS(scalar_one));
287 API_NS(scalar_destroy)(t1);
292 i -= SCALAR_SER_BYTES;
293 sc_montmul(t1,t1,sc_r2);
294 ignore_result( API_NS(scalar_decode)(t2, ser+i) );
295 API_NS(scalar_add)(t1, t1, t2);
298 API_NS(scalar_copy)(s, t1);
299 API_NS(scalar_destroy)(t1);
300 API_NS(scalar_destroy)(t2);
303 void API_NS(scalar_encode)(
304 unsigned char ser[SCALAR_SER_BYTES],
307 unsigned int i,j,k=0;
308 for (i=0; i<SCALAR_LIMBS; i++) {
309 for (j=0; j<sizeof(decaf_word_t); j++,k++) {
310 ser[k] = s->limb[i] >> (8*j);
315 void API_NS(scalar_cond_sel) (
321 constant_time_select(out,a,b,sizeof(scalar_t),bool_to_mask(pick_b),sizeof(out->limb[0]));
324 void API_NS(scalar_halve) (
328 decaf_word_t mask = -(a->limb[0] & 1);
329 decaf_dword_t chain = 0;
331 for (i=0; i<SCALAR_LIMBS; i++) {
332 chain = (chain + a->limb[i]) + (sc_p->limb[i] & mask);
333 out->limb[i] = chain;
334 chain >>= DECAF_WORD_BITS;
336 for (i=0; i<SCALAR_LIMBS-1; i++) {
337 out->limb[i] = out->limb[i]>>1 | out->limb[i+1]<<(WBITS-1);
339 out->limb[i] = out->limb[i]>>1 | chain<<(WBITS-1);