2013-06-30 20:55:49 +04:00
|
|
|
///////////////////////////////////////////////////////////////////////////////
|
|
|
|
//
|
|
|
|
/// \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 *restrict coder, lzma_mf *restrict mf,
|
|
|
|
uint32_t *restrict back_res, uint32_t *restrict len_res)
|
|
|
|
{
|
2014-07-14 00:21:58 +04:00
|
|
|
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;
|
|
|
|
|
2013-06-30 20:55:49 +04:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2014-07-14 00:21:58 +04:00
|
|
|
buf = mf_ptr(mf) - 1;
|
|
|
|
buf_avail = my_min(mf_avail(mf) + 1, MATCH_LEN_MAX);
|
2013-06-30 20:55:49 +04:00
|
|
|
|
|
|
|
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
|
2014-07-14 00:21:58 +04:00
|
|
|
for (i = 0; i < REP_DISTANCES; ++i) {
|
|
|
|
uint32_t len;
|
2013-06-30 20:55:49 +04:00
|
|
|
|
|
|
|
// 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;
|
|
|
|
|
2014-07-14 00:21:58 +04:00
|
|
|
limit = len_main - 1;
|
|
|
|
|
|
|
|
for (i = 0; i < REP_DISTANCES; ++i) {
|
|
|
|
uint32_t len;
|
2013-06-30 20:55:49 +04:00
|
|
|
|
|
|
|
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;
|
|
|
|
}
|