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/* see inner.h */
28uint32_t
29br_i15_decode_mod(uint16_t *x, const void *src, size_t len, const uint16_t *m)
30{
31	/*
32	 * Two-pass algorithm: in the first pass, we determine whether the
33	 * value fits; in the second pass, we do the actual write.
34	 *
35	 * During the first pass, 'r' contains the comparison result so
36	 * far:
37	 *  0x00000000   value is equal to the modulus
38	 *  0x00000001   value is greater than the modulus
39	 *  0xFFFFFFFF   value is lower than the modulus
40	 *
41	 * Since we iterate starting with the least significant bytes (at
42	 * the end of src[]), each new comparison overrides the previous
43	 * except when the comparison yields 0 (equal).
44	 *
45	 * During the second pass, 'r' is either 0xFFFFFFFF (value fits)
46	 * or 0x00000000 (value does not fit).
47	 *
48	 * We must iterate over all bytes of the source, _and_ possibly
49	 * some extra virtual bytes (with value 0) so as to cover the
50	 * complete modulus as well. We also add 4 such extra bytes beyond
51	 * the modulus length because it then guarantees that no accumulated
52	 * partial word remains to be processed.
53	 */
54	const unsigned char *buf;
55	size_t mlen, tlen;
56	int pass;
57	uint32_t r;
58
59	buf = src;
60	mlen = (m[0] + 15) >> 4;
61	tlen = (mlen << 1);
62	if (tlen < len) {
63		tlen = len;
64	}
65	tlen += 4;
66	r = 0;
67	for (pass = 0; pass < 2; pass ++) {
68		size_t u, v;
69		uint32_t acc;
70		int acc_len;
71
72		v = 1;
73		acc = 0;
74		acc_len = 0;
75		for (u = 0; u < tlen; u ++) {
76			uint32_t b;
77
78			if (u < len) {
79				b = buf[len - 1 - u];
80			} else {
81				b = 0;
82			}
83			acc |= (b << acc_len);
84			acc_len += 8;
85			if (acc_len >= 15) {
86				uint32_t xw;
87
88				xw = acc & (uint32_t)0x7FFF;
89				acc_len -= 15;
90				acc = b >> (8 - acc_len);
91				if (v <= mlen) {
92					if (pass) {
93						x[v] = r & xw;
94					} else {
95						uint32_t cc;
96
97						cc = (uint32_t)CMP(xw, m[v]);
98						r = MUX(EQ(cc, 0), r, cc);
99					}
100				} else {
101					if (!pass) {
102						r = MUX(EQ(xw, 0), r, 1);
103					}
104				}
105				v ++;
106			}
107		}
108
109		/*
110		 * When we reach this point at the end of the first pass:
111		 * r is either 0, 1 or -1; we want to set r to 0 if it
112		 * is equal to 0 or 1, and leave it to -1 otherwise.
113		 *
114		 * When we reach this point at the end of the second pass:
115		 * r is either 0 or -1; we want to leave that value
116		 * untouched. This is a subcase of the previous.
117		 */
118		r >>= 1;
119		r |= (r << 1);
120	}
121
122	x[0] = m[0];
123	return r & (uint32_t)1;
124}
125