opencv/3rdparty/zlib-ng/insert_string_tpl.h
Letu Ren 0de26fd78e Add zlib-ng as an alternative zlib implementation
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
2024-01-14 14:58:47 +08:00

109 lines
3.5 KiB
C

#ifndef INSERT_STRING_H_
#define INSERT_STRING_H_
/* insert_string.h -- Private insert_string functions shared with more than
* one insert string implementation
*
* Copyright (C) 1995-2013 Jean-loup Gailly and Mark Adler
*
* Copyright (C) 2013 Intel Corporation. All rights reserved.
* Authors:
* Wajdi Feghali <wajdi.k.feghali@intel.com>
* Jim Guilford <james.guilford@intel.com>
* Vinodh Gopal <vinodh.gopal@intel.com>
* Erdinc Ozturk <erdinc.ozturk@intel.com>
* Jim Kukunas <james.t.kukunas@linux.intel.com>
*
* Portions are Copyright (C) 2016 12Sided Technology, LLC.
* Author:
* Phil Vachon <pvachon@12sidedtech.com>
*
* For conditions of distribution and use, see copyright notice in zlib.h
*
*/
#ifndef HASH_CALC_OFFSET
# define HASH_CALC_OFFSET 0
#endif
#ifndef HASH_CALC_MASK
# define HASH_CALC_MASK HASH_MASK
#endif
#ifndef HASH_CALC_READ
# if BYTE_ORDER == LITTLE_ENDIAN
# define HASH_CALC_READ \
memcpy(&val, strstart, sizeof(val));
# else
# define HASH_CALC_READ \
val = ((uint32_t)(strstart[0])); \
val |= ((uint32_t)(strstart[1]) << 8); \
val |= ((uint32_t)(strstart[2]) << 16); \
val |= ((uint32_t)(strstart[3]) << 24);
# endif
#endif
/* ===========================================================================
* Update a hash value with the given input byte
* IN assertion: all calls to UPDATE_HASH are made with consecutive
* input characters, so that a running hash key can be computed from the
* previous key instead of complete recalculation each time.
*/
Z_INTERNAL uint32_t UPDATE_HASH(deflate_state *const s, uint32_t h, uint32_t val) {
(void)s;
HASH_CALC(s, h, val);
return h & HASH_CALC_MASK;
}
/* ===========================================================================
* Quick insert string str in the dictionary and set match_head to the previous head
* of the hash chain (the most recent string with same hash key). Return
* the previous length of the hash chain.
*/
Z_INTERNAL Pos QUICK_INSERT_STRING(deflate_state *const s, uint32_t str) {
Pos head;
uint8_t *strstart = s->window + str + HASH_CALC_OFFSET;
uint32_t val, hm;
HASH_CALC_VAR_INIT;
HASH_CALC_READ;
HASH_CALC(s, HASH_CALC_VAR, val);
HASH_CALC_VAR &= HASH_CALC_MASK;
hm = HASH_CALC_VAR;
head = s->head[hm];
if (LIKELY(head != str)) {
s->prev[str & s->w_mask] = head;
s->head[hm] = (Pos)str;
}
return head;
}
/* ===========================================================================
* Insert string str in the dictionary and set match_head to the previous head
* of the hash chain (the most recent string with same hash key). Return
* the previous length of the hash chain.
* IN assertion: all calls to INSERT_STRING are made with consecutive
* input characters and the first STD_MIN_MATCH bytes of str are valid
* (except for the last STD_MIN_MATCH-1 bytes of the input file).
*/
Z_INTERNAL void INSERT_STRING(deflate_state *const s, uint32_t str, uint32_t count) {
uint8_t *strstart = s->window + str + HASH_CALC_OFFSET;
uint8_t *strend = strstart + count;
for (Pos idx = (Pos)str; strstart < strend; idx++, strstart++) {
uint32_t val, hm;
HASH_CALC_VAR_INIT;
HASH_CALC_READ;
HASH_CALC(s, HASH_CALC_VAR, val);
HASH_CALC_VAR &= HASH_CALC_MASK;
hm = HASH_CALC_VAR;
Pos head = s->head[hm];
if (LIKELY(head != idx)) {
s->prev[idx & s->w_mask] = head;
s->head[hm] = idx;
}
}
}
#endif