mirror of
https://github.com/opencv/opencv.git
synced 2024-12-03 00:10:21 +08:00
0de26fd78e
Zlib-ng is zlib replacement with optimizations for "next generation" systems. Its optimization may benifits image library decode and encode speed such as libpng. In our tests, if using zlib-ng and libpng combination on a x86_64 machine with AVX2, the time of `imdecode` amd `imencode` will drop 20% approximately. This patch enables zlib-ng's optimization if `CV_DISABLE_OPTIMIZATION` is OFF. Since Zlib-ng can dispatch intrinsics on the fly, port work is much easier. Related discussion: https://github.com/opencv/opencv/issues/22573
294 lines
10 KiB
C
294 lines
10 KiB
C
/* deflate_medium.c -- The deflate_medium deflate strategy
|
|
*
|
|
* Copyright (C) 2013 Intel Corporation. All rights reserved.
|
|
* Authors:
|
|
* Arjan van de Ven <arjan@linux.intel.com>
|
|
*
|
|
* For conditions of distribution and use, see copyright notice in zlib.h
|
|
*/
|
|
#ifndef NO_MEDIUM_STRATEGY
|
|
#include "zbuild.h"
|
|
#include "deflate.h"
|
|
#include "deflate_p.h"
|
|
#include "functable.h"
|
|
|
|
struct match {
|
|
uint16_t match_start;
|
|
uint16_t match_length;
|
|
uint16_t strstart;
|
|
uint16_t orgstart;
|
|
};
|
|
|
|
static int emit_match(deflate_state *s, struct match match) {
|
|
int bflush = 0;
|
|
|
|
/* matches that are not long enough we need to emit as literals */
|
|
if (match.match_length < WANT_MIN_MATCH) {
|
|
while (match.match_length) {
|
|
bflush += zng_tr_tally_lit(s, s->window[match.strstart]);
|
|
s->lookahead--;
|
|
match.strstart++;
|
|
match.match_length--;
|
|
}
|
|
return bflush;
|
|
}
|
|
|
|
check_match(s, match.strstart, match.match_start, match.match_length);
|
|
|
|
bflush += zng_tr_tally_dist(s, match.strstart - match.match_start, match.match_length - STD_MIN_MATCH);
|
|
|
|
s->lookahead -= match.match_length;
|
|
return bflush;
|
|
}
|
|
|
|
static void insert_match(deflate_state *s, struct match match) {
|
|
if (UNLIKELY(s->lookahead <= (unsigned int)(match.match_length + WANT_MIN_MATCH)))
|
|
return;
|
|
|
|
/* matches that are not long enough we need to emit as literals */
|
|
if (LIKELY(match.match_length < WANT_MIN_MATCH)) {
|
|
match.strstart++;
|
|
match.match_length--;
|
|
if (UNLIKELY(match.match_length > 0)) {
|
|
if (match.strstart >= match.orgstart) {
|
|
if (match.strstart + match.match_length - 1 >= match.orgstart) {
|
|
functable.insert_string(s, match.strstart, match.match_length);
|
|
} else {
|
|
functable.insert_string(s, match.strstart, match.orgstart - match.strstart + 1);
|
|
}
|
|
match.strstart += match.match_length;
|
|
match.match_length = 0;
|
|
}
|
|
}
|
|
return;
|
|
}
|
|
|
|
/* Insert new strings in the hash table only if the match length
|
|
* is not too large. This saves time but degrades compression.
|
|
*/
|
|
if (match.match_length <= 16 * s->max_insert_length && s->lookahead >= WANT_MIN_MATCH) {
|
|
match.match_length--; /* string at strstart already in table */
|
|
match.strstart++;
|
|
|
|
if (LIKELY(match.strstart >= match.orgstart)) {
|
|
if (LIKELY(match.strstart + match.match_length - 1 >= match.orgstart)) {
|
|
functable.insert_string(s, match.strstart, match.match_length);
|
|
} else {
|
|
functable.insert_string(s, match.strstart, match.orgstart - match.strstart + 1);
|
|
}
|
|
} else if (match.orgstart < match.strstart + match.match_length) {
|
|
functable.insert_string(s, match.orgstart, match.strstart + match.match_length - match.orgstart);
|
|
}
|
|
match.strstart += match.match_length;
|
|
match.match_length = 0;
|
|
} else {
|
|
match.strstart += match.match_length;
|
|
match.match_length = 0;
|
|
|
|
if (match.strstart >= (STD_MIN_MATCH - 2))
|
|
functable.quick_insert_string(s, match.strstart + 2 - STD_MIN_MATCH);
|
|
|
|
/* If lookahead < WANT_MIN_MATCH, ins_h is garbage, but it does not
|
|
* matter since it will be recomputed at next deflate call.
|
|
*/
|
|
}
|
|
}
|
|
|
|
static void fizzle_matches(deflate_state *s, struct match *current, struct match *next) {
|
|
Pos limit;
|
|
unsigned char *match, *orig;
|
|
int changed = 0;
|
|
struct match c, n;
|
|
/* step zero: sanity checks */
|
|
|
|
if (current->match_length <= 1)
|
|
return;
|
|
|
|
if (UNLIKELY(current->match_length > 1 + next->match_start))
|
|
return;
|
|
|
|
if (UNLIKELY(current->match_length > 1 + next->strstart))
|
|
return;
|
|
|
|
match = s->window - current->match_length + 1 + next->match_start;
|
|
orig = s->window - current->match_length + 1 + next->strstart;
|
|
|
|
/* quick exit check.. if this fails then don't bother with anything else */
|
|
if (LIKELY(*match != *orig))
|
|
return;
|
|
|
|
c = *current;
|
|
n = *next;
|
|
|
|
/* step one: try to move the "next" match to the left as much as possible */
|
|
limit = next->strstart > MAX_DIST(s) ? next->strstart - (Pos)MAX_DIST(s) : 0;
|
|
|
|
match = s->window + n.match_start - 1;
|
|
orig = s->window + n.strstart - 1;
|
|
|
|
while (*match == *orig) {
|
|
if (UNLIKELY(c.match_length < 1))
|
|
break;
|
|
if (UNLIKELY(n.strstart <= limit))
|
|
break;
|
|
if (UNLIKELY(n.match_length >= 256))
|
|
break;
|
|
if (UNLIKELY(n.match_start <= 1))
|
|
break;
|
|
|
|
n.strstart--;
|
|
n.match_start--;
|
|
n.match_length++;
|
|
c.match_length--;
|
|
match--;
|
|
orig--;
|
|
changed++;
|
|
}
|
|
|
|
if (!changed)
|
|
return;
|
|
|
|
if (c.match_length <= 1 && n.match_length != 2) {
|
|
n.orgstart++;
|
|
*current = c;
|
|
*next = n;
|
|
} else {
|
|
return;
|
|
}
|
|
}
|
|
|
|
Z_INTERNAL block_state deflate_medium(deflate_state *s, int flush) {
|
|
/* Align the first struct to start on a new cacheline, this allows us to fit both structs in one cacheline */
|
|
ALIGNED_(16) struct match current_match;
|
|
struct match next_match;
|
|
|
|
/* For levels below 5, don't check the next position for a better match */
|
|
int early_exit = s->level < 5;
|
|
|
|
memset(¤t_match, 0, sizeof(struct match));
|
|
memset(&next_match, 0, sizeof(struct match));
|
|
|
|
for (;;) {
|
|
Pos hash_head = 0; /* head of the hash chain */
|
|
int bflush = 0; /* set if current block must be flushed */
|
|
int64_t dist;
|
|
|
|
/* Make sure that we always have enough lookahead, except
|
|
* at the end of the input file. We need STD_MAX_MATCH bytes
|
|
* for the next match, plus WANT_MIN_MATCH bytes to insert the
|
|
* string following the next current_match.
|
|
*/
|
|
if (s->lookahead < MIN_LOOKAHEAD) {
|
|
PREFIX(fill_window)(s);
|
|
if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
|
|
return need_more;
|
|
}
|
|
if (UNLIKELY(s->lookahead == 0))
|
|
break; /* flush the current block */
|
|
next_match.match_length = 0;
|
|
}
|
|
|
|
/* Insert the string window[strstart .. strstart+2] in the
|
|
* dictionary, and set hash_head to the head of the hash chain:
|
|
*/
|
|
|
|
/* If we already have a future match from a previous round, just use that */
|
|
if (!early_exit && next_match.match_length > 0) {
|
|
current_match = next_match;
|
|
next_match.match_length = 0;
|
|
} else {
|
|
hash_head = 0;
|
|
if (s->lookahead >= WANT_MIN_MATCH) {
|
|
hash_head = functable.quick_insert_string(s, s->strstart);
|
|
}
|
|
|
|
current_match.strstart = (uint16_t)s->strstart;
|
|
current_match.orgstart = current_match.strstart;
|
|
|
|
/* Find the longest match, discarding those <= prev_length.
|
|
* At this point we have always match_length < WANT_MIN_MATCH
|
|
*/
|
|
|
|
dist = (int64_t)s->strstart - hash_head;
|
|
if (dist <= MAX_DIST(s) && dist > 0 && hash_head != 0) {
|
|
/* To simplify the code, we prevent matches with the string
|
|
* of window index 0 (in particular we have to avoid a match
|
|
* of the string with itself at the start of the input file).
|
|
*/
|
|
current_match.match_length = (uint16_t)functable.longest_match(s, hash_head);
|
|
current_match.match_start = (uint16_t)s->match_start;
|
|
if (UNLIKELY(current_match.match_length < WANT_MIN_MATCH))
|
|
current_match.match_length = 1;
|
|
if (UNLIKELY(current_match.match_start >= current_match.strstart)) {
|
|
/* this can happen due to some restarts */
|
|
current_match.match_length = 1;
|
|
}
|
|
} else {
|
|
/* Set up the match to be a 1 byte literal */
|
|
current_match.match_start = 0;
|
|
current_match.match_length = 1;
|
|
}
|
|
}
|
|
|
|
insert_match(s, current_match);
|
|
|
|
/* now, look ahead one */
|
|
if (LIKELY(!early_exit && s->lookahead > MIN_LOOKAHEAD && (uint32_t)(current_match.strstart + current_match.match_length) < (s->window_size - MIN_LOOKAHEAD))) {
|
|
s->strstart = current_match.strstart + current_match.match_length;
|
|
hash_head = functable.quick_insert_string(s, s->strstart);
|
|
|
|
next_match.strstart = (uint16_t)s->strstart;
|
|
next_match.orgstart = next_match.strstart;
|
|
|
|
/* Find the longest match, discarding those <= prev_length.
|
|
* At this point we have always match_length < WANT_MIN_MATCH
|
|
*/
|
|
|
|
dist = (int64_t)s->strstart - hash_head;
|
|
if (dist <= MAX_DIST(s) && dist > 0 && hash_head != 0) {
|
|
/* To simplify the code, we prevent matches with the string
|
|
* of window index 0 (in particular we have to avoid a match
|
|
* of the string with itself at the start of the input file).
|
|
*/
|
|
next_match.match_length = (uint16_t)functable.longest_match(s, hash_head);
|
|
next_match.match_start = (uint16_t)s->match_start;
|
|
if (UNLIKELY(next_match.match_start >= next_match.strstart)) {
|
|
/* this can happen due to some restarts */
|
|
next_match.match_length = 1;
|
|
}
|
|
if (next_match.match_length < WANT_MIN_MATCH)
|
|
next_match.match_length = 1;
|
|
else
|
|
fizzle_matches(s, ¤t_match, &next_match);
|
|
} else {
|
|
/* Set up the match to be a 1 byte literal */
|
|
next_match.match_start = 0;
|
|
next_match.match_length = 1;
|
|
}
|
|
|
|
s->strstart = current_match.strstart;
|
|
} else {
|
|
next_match.match_length = 0;
|
|
}
|
|
|
|
/* now emit the current match */
|
|
bflush = emit_match(s, current_match);
|
|
|
|
/* move the "cursor" forward */
|
|
s->strstart += current_match.match_length;
|
|
|
|
if (UNLIKELY(bflush))
|
|
FLUSH_BLOCK(s, 0);
|
|
}
|
|
s->insert = s->strstart < (STD_MIN_MATCH - 1) ? s->strstart : (STD_MIN_MATCH - 1);
|
|
if (flush == Z_FINISH) {
|
|
FLUSH_BLOCK(s, 1);
|
|
return finish_done;
|
|
}
|
|
if (UNLIKELY(s->sym_next))
|
|
FLUSH_BLOCK(s, 0);
|
|
|
|
return block_done;
|
|
}
|
|
#endif
|