xref: /freebsd/contrib/bearssl/src/int/i15_muladd.c (revision 580744621f33383027108364dcadad718df46ffe)
1 /*
2  * Copyright (c) 2017 Thomas Pornin <pornin@bolet.org>
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining
5  * a copy of this software and associated documentation files (the
6  * "Software"), to deal in the Software without restriction, including
7  * without limitation the rights to use, copy, modify, merge, publish,
8  * distribute, sublicense, and/or sell copies of the Software, and to
9  * permit persons to whom the Software is furnished to do so, subject to
10  * the following conditions:
11  *
12  * The above copyright notice and this permission notice shall be
13  * included in all copies or substantial portions of the Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
19  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
20  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
21  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
24 
25 #include "inner.h"
26 
27 /*
28  * Constant-time division. The divisor must not be larger than 16 bits,
29  * and the quotient must fit on 17 bits.
30  */
31 static uint32_t
32 divrem16(uint32_t x, uint32_t d, uint32_t *r)
33 {
34 	int i;
35 	uint32_t q;
36 
37 	q = 0;
38 	d <<= 16;
39 	for (i = 16; i >= 0; i --) {
40 		uint32_t ctl;
41 
42 		ctl = LE(d, x);
43 		q |= ctl << i;
44 		x -= (-ctl) & d;
45 		d >>= 1;
46 	}
47 	if (r != NULL) {
48 		*r = x;
49 	}
50 	return q;
51 }
52 
53 /* see inner.h */
54 void
55 br_i15_muladd_small(uint16_t *x, uint16_t z, const uint16_t *m)
56 {
57 	/*
58 	 * Constant-time: we accept to leak the exact bit length of the
59 	 * modulus m.
60 	 */
61 	unsigned m_bitlen, mblr;
62 	size_t u, mlen;
63 	uint32_t hi, a0, a, b, q;
64 	uint32_t cc, tb, over, under;
65 
66 	/*
67 	 * Simple case: the modulus fits on one word.
68 	 */
69 	m_bitlen = m[0];
70 	if (m_bitlen == 0) {
71 		return;
72 	}
73 	if (m_bitlen <= 15) {
74 		uint32_t rem;
75 
76 		divrem16(((uint32_t)x[1] << 15) | z, m[1], &rem);
77 		x[1] = rem;
78 		return;
79 	}
80 	mlen = (m_bitlen + 15) >> 4;
81 	mblr = m_bitlen & 15;
82 
83 	/*
84 	 * Principle: we estimate the quotient (x*2^15+z)/m by
85 	 * doing a 30/15 division with the high words.
86 	 *
87 	 * Let:
88 	 *   w = 2^15
89 	 *   a = (w*a0 + a1) * w^N + a2
90 	 *   b = b0 * w^N + b2
91 	 * such that:
92 	 *   0 <= a0 < w
93 	 *   0 <= a1 < w
94 	 *   0 <= a2 < w^N
95 	 *   w/2 <= b0 < w
96 	 *   0 <= b2 < w^N
97 	 *   a < w*b
98 	 * I.e. the two top words of a are a0:a1, the top word of b is
99 	 * b0, we ensured that b0 is "full" (high bit set), and a is
100 	 * such that the quotient q = a/b fits on one word (0 <= q < w).
101 	 *
102 	 * If a = b*q + r (with 0 <= r < q), then we can estimate q by
103 	 * using a division on the top words:
104 	 *   a0*w + a1 = b0*u + v (with 0 <= v < b0)
105 	 * Then the following holds:
106 	 *   0 <= u <= w
107 	 *   u-2 <= q <= u
108 	 */
109 	hi = x[mlen];
110 	if (mblr == 0) {
111 		a0 = x[mlen];
112 		memmove(x + 2, x + 1, (mlen - 1) * sizeof *x);
113 		x[1] = z;
114 		a = (a0 << 15) + x[mlen];
115 		b = m[mlen];
116 	} else {
117 		a0 = (x[mlen] << (15 - mblr)) | (x[mlen - 1] >> mblr);
118 		memmove(x + 2, x + 1, (mlen - 1) * sizeof *x);
119 		x[1] = z;
120 		a = (a0 << 15) | (((x[mlen] << (15 - mblr))
121 			| (x[mlen - 1] >> mblr)) & 0x7FFF);
122 		b = (m[mlen] << (15 - mblr)) | (m[mlen - 1] >> mblr);
123 	}
124 	q = divrem16(a, b, NULL);
125 
126 	/*
127 	 * We computed an estimate for q, but the real one may be q,
128 	 * q-1 or q-2; moreover, the division may have returned a value
129 	 * 8000 or even 8001 if the two high words were identical, and
130 	 * we want to avoid values beyond 7FFF. We thus adjust q so
131 	 * that the "true" multiplier will be q+1, q or q-1, and q is
132 	 * in the 0000..7FFF range.
133 	 */
134 	q = MUX(EQ(b, a0), 0x7FFF, q - 1 + ((q - 1) >> 31));
135 
136 	/*
137 	 * We subtract q*m from x (x has an extra high word of value 'hi').
138 	 * Since q may be off by 1 (in either direction), we may have to
139 	 * add or subtract m afterwards.
140 	 *
141 	 * The 'tb' flag will be true (1) at the end of the loop if the
142 	 * result is greater than or equal to the modulus (not counting
143 	 * 'hi' or the carry).
144 	 */
145 	cc = 0;
146 	tb = 1;
147 	for (u = 1; u <= mlen; u ++) {
148 		uint32_t mw, zl, xw, nxw;
149 
150 		mw = m[u];
151 		zl = MUL15(mw, q) + cc;
152 		cc = zl >> 15;
153 		zl &= 0x7FFF;
154 		xw = x[u];
155 		nxw = xw - zl;
156 		cc += nxw >> 31;
157 		nxw &= 0x7FFF;
158 		x[u] = nxw;
159 		tb = MUX(EQ(nxw, mw), tb, GT(nxw, mw));
160 	}
161 
162 	/*
163 	 * If we underestimated q, then either cc < hi (one extra bit
164 	 * beyond the top array word), or cc == hi and tb is true (no
165 	 * extra bit, but the result is not lower than the modulus).
166 	 *
167 	 * If we overestimated q, then cc > hi.
168 	 */
169 	over = GT(cc, hi);
170 	under = ~over & (tb | LT(cc, hi));
171 	br_i15_add(x, m, over);
172 	br_i15_sub(x, m, under);
173 }
174