summaryrefslogtreecommitdiff
path: root/Utilities/cmliblzma/liblzma/lzma/lzma_encoder_optimum_fast.c
blob: f983126433e23b31df88d413da165bead1d6082c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
///////////////////////////////////////////////////////////////////////////////
//
/// \file       lzma_encoder_optimum_fast.c
//
//  Author:     Igor Pavlov
//
//  This file has been put into the public domain.
//  You can do whatever you want with this file.
//
///////////////////////////////////////////////////////////////////////////////

#include "lzma_encoder_private.h"


#define change_pair(small_dist, big_dist) \
	(((big_dist) >> 7) > (small_dist))


extern void
lzma_lzma_optimum_fast(lzma_coder *LZMA_RESTRICT coder, lzma_mf *LZMA_RESTRICT mf,
		uint32_t *LZMA_RESTRICT back_res, uint32_t *LZMA_RESTRICT len_res)
{
	const uint8_t *buf;
	uint32_t buf_avail;
	uint32_t i;
	uint32_t rep_len = 0;
	uint32_t rep_index = 0;
	uint32_t back_main = 0;
	uint32_t limit;

	const uint32_t nice_len = mf->nice_len;

	uint32_t len_main;
	uint32_t matches_count;
	if (mf->read_ahead == 0) {
		len_main = mf_find(mf, &matches_count, coder->matches);
	} else {
		assert(mf->read_ahead == 1);
		len_main = coder->longest_match_length;
		matches_count = coder->matches_count;
	}

	buf = mf_ptr(mf) - 1;
	buf_avail = my_min(mf_avail(mf) + 1, MATCH_LEN_MAX);

	if (buf_avail < 2) {
		// There's not enough input left to encode a match.
		*back_res = UINT32_MAX;
		*len_res = 1;
		return;
	}

	// Look for repeated matches; scan the previous four match distances
	for (i = 0; i < REP_DISTANCES; ++i) {
		uint32_t len;

		// Pointer to the beginning of the match candidate
		const uint8_t *const buf_back = buf - coder->reps[i] - 1;

		// If the first two bytes (2 == MATCH_LEN_MIN) do not match,
		// this rep is not useful.
		if (not_equal_16(buf, buf_back))
			continue;

		// The first two bytes matched.
		// Calculate the length of the match.
		for (len = 2; len < buf_avail
				&& buf[len] == buf_back[len]; ++len) ;

		// If we have found a repeated match that is at least
		// nice_len long, return it immediately.
		if (len >= nice_len) {
			*back_res = i;
			*len_res = len;
			mf_skip(mf, len - 1);
			return;
		}

		if (len > rep_len) {
			rep_index = i;
			rep_len = len;
		}
	}

	// We didn't find a long enough repeated match. Encode it as a normal
	// match if the match length is at least nice_len.
	if (len_main >= nice_len) {
		*back_res = coder->matches[matches_count - 1].dist
				+ REP_DISTANCES;
		*len_res = len_main;
		mf_skip(mf, len_main - 1);
		return;
	}

	if (len_main >= 2) {
		back_main = coder->matches[matches_count - 1].dist;

		while (matches_count > 1 && len_main ==
				coder->matches[matches_count - 2].len + 1) {
			if (!change_pair(coder->matches[
						matches_count - 2].dist,
					back_main))
				break;

			--matches_count;
			len_main = coder->matches[matches_count - 1].len;
			back_main = coder->matches[matches_count - 1].dist;
		}

		if (len_main == 2 && back_main >= 0x80)
			len_main = 1;
	}

	if (rep_len >= 2) {
		if (rep_len + 1 >= len_main
				|| (rep_len + 2 >= len_main
					&& back_main > (UINT32_C(1) << 9))
				|| (rep_len + 3 >= len_main
					&& back_main > (UINT32_C(1) << 15))) {
			*back_res = rep_index;
			*len_res = rep_len;
			mf_skip(mf, rep_len - 1);
			return;
		}
	}

	if (len_main < 2 || buf_avail <= 2) {
		*back_res = UINT32_MAX;
		*len_res = 1;
		return;
	}

	// Get the matches for the next byte. If we find a better match,
	// the current byte is encoded as a literal.
	coder->longest_match_length = mf_find(mf,
			&coder->matches_count, coder->matches);

	if (coder->longest_match_length >= 2) {
		const uint32_t new_dist = coder->matches[
				coder->matches_count - 1].dist;

		if ((coder->longest_match_length >= len_main
					&& new_dist < back_main)
				|| (coder->longest_match_length == len_main + 1
					&& !change_pair(back_main, new_dist))
				|| (coder->longest_match_length > len_main + 1)
				|| (coder->longest_match_length + 1 >= len_main
					&& len_main >= 3
					&& change_pair(new_dist, back_main))) {
			*back_res = UINT32_MAX;
			*len_res = 1;
			return;
		}
	}

	// In contrast to LZMA SDK, dictionary could not have been moved
	// between mf_find() calls, thus it is safe to just increment
	// the old buf pointer instead of recalculating it with mf_ptr().
	++buf;

	limit = len_main - 1;

	for (i = 0; i < REP_DISTANCES; ++i) {
		uint32_t len;

		const uint8_t *const buf_back = buf - coder->reps[i] - 1;

		if (not_equal_16(buf, buf_back))
			continue;

		for (len = 2; len < limit
				&& buf[len] == buf_back[len]; ++len) ;

		if (len >= limit) {
			*back_res = UINT32_MAX;
			*len_res = 1;
			return;
		}
	}

	*back_res = back_main + REP_DISTANCES;
	*len_res = len_main;
	mf_skip(mf, len_main - 2);
	return;
}