android_kernel_motorola_sm6225/lib/zlib_deflate/deflate.c
Richard Purdie 4f3865fb57 [PATCH] zlib_inflate: Upgrade library code to a recent version
Upgrade the zlib_inflate implementation in the kernel from a patched
version 1.1.3/4 to a patched 1.2.3.

The code in the kernel is about seven years old and I noticed that the
external zlib library's inflate performance was significantly faster (~50%)
than the code in the kernel on ARM (and faster again on x86_32).

For comparison the newer deflate code is 20% slower on ARM and 50% slower
on x86_32 but gives an approx 1% compression ratio improvement.  I don't
consider this to be an improvement for kernel use so have no plans to
change the zlib_deflate code.

Various changes have been made to the zlib code in the kernel, the most
significant being the extra functions/flush option used by ppp_deflate.
This update reimplements the features PPP needs to ensure it continues to
work.

This code has been tested on ARM under both JFFS2 (with zlib compression
enabled) and ppp_deflate and on x86_32.  JFFS2 sees an approx.  10% real
world file read speed improvement.

This patch also removes ZLIB_VERSION as it no longer has a correct value.
We don't need version checks anyway as the kernel's module handling will
take care of that for us.  This removal is also more in keeping with the
zlib author's wishes (http://www.zlib.net/zlib_faq.html#faq24) and I've
added something to the zlib.h header to note its a modified version.

Signed-off-by: Richard Purdie <rpurdie@rpsys.net>
Acked-by: Joern Engel <joern@wh.fh-wedel.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-22 15:05:58 -07:00

1253 lines
43 KiB
C

/* +++ deflate.c */
/* deflate.c -- compress data using the deflation algorithm
* Copyright (C) 1995-1996 Jean-loup Gailly.
* For conditions of distribution and use, see copyright notice in zlib.h
*/
/*
* ALGORITHM
*
* The "deflation" process depends on being able to identify portions
* of the input text which are identical to earlier input (within a
* sliding window trailing behind the input currently being processed).
*
* The most straightforward technique turns out to be the fastest for
* most input files: try all possible matches and select the longest.
* The key feature of this algorithm is that insertions into the string
* dictionary are very simple and thus fast, and deletions are avoided
* completely. Insertions are performed at each input character, whereas
* string matches are performed only when the previous match ends. So it
* is preferable to spend more time in matches to allow very fast string
* insertions and avoid deletions. The matching algorithm for small
* strings is inspired from that of Rabin & Karp. A brute force approach
* is used to find longer strings when a small match has been found.
* A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
* (by Leonid Broukhis).
* A previous version of this file used a more sophisticated algorithm
* (by Fiala and Greene) which is guaranteed to run in linear amortized
* time, but has a larger average cost, uses more memory and is patented.
* However the F&G algorithm may be faster for some highly redundant
* files if the parameter max_chain_length (described below) is too large.
*
* ACKNOWLEDGEMENTS
*
* The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
* I found it in 'freeze' written by Leonid Broukhis.
* Thanks to many people for bug reports and testing.
*
* REFERENCES
*
* Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
* Available in ftp://ds.internic.net/rfc/rfc1951.txt
*
* A description of the Rabin and Karp algorithm is given in the book
* "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
*
* Fiala,E.R., and Greene,D.H.
* Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
*
*/
#include <linux/module.h>
#include <linux/zutil.h>
#include "defutil.h"
/* ===========================================================================
* Function prototypes.
*/
typedef enum {
need_more, /* block not completed, need more input or more output */
block_done, /* block flush performed */
finish_started, /* finish started, need only more output at next deflate */
finish_done /* finish done, accept no more input or output */
} block_state;
typedef block_state (*compress_func) (deflate_state *s, int flush);
/* Compression function. Returns the block state after the call. */
static void fill_window (deflate_state *s);
static block_state deflate_stored (deflate_state *s, int flush);
static block_state deflate_fast (deflate_state *s, int flush);
static block_state deflate_slow (deflate_state *s, int flush);
static void lm_init (deflate_state *s);
static void putShortMSB (deflate_state *s, uInt b);
static void flush_pending (z_streamp strm);
static int read_buf (z_streamp strm, Byte *buf, unsigned size);
static uInt longest_match (deflate_state *s, IPos cur_match);
#ifdef DEBUG_ZLIB
static void check_match (deflate_state *s, IPos start, IPos match,
int length);
#endif
/* ===========================================================================
* Local data
*/
#define NIL 0
/* Tail of hash chains */
#ifndef TOO_FAR
# define TOO_FAR 4096
#endif
/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
/* Minimum amount of lookahead, except at the end of the input file.
* See deflate.c for comments about the MIN_MATCH+1.
*/
/* Values for max_lazy_match, good_match and max_chain_length, depending on
* the desired pack level (0..9). The values given below have been tuned to
* exclude worst case performance for pathological files. Better values may be
* found for specific files.
*/
typedef struct config_s {
ush good_length; /* reduce lazy search above this match length */
ush max_lazy; /* do not perform lazy search above this match length */
ush nice_length; /* quit search above this match length */
ush max_chain;
compress_func func;
} config;
static const config configuration_table[10] = {
/* good lazy nice chain */
/* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */
/* 1 */ {4, 4, 8, 4, deflate_fast}, /* maximum speed, no lazy matches */
/* 2 */ {4, 5, 16, 8, deflate_fast},
/* 3 */ {4, 6, 32, 32, deflate_fast},
/* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */
/* 5 */ {8, 16, 32, 32, deflate_slow},
/* 6 */ {8, 16, 128, 128, deflate_slow},
/* 7 */ {8, 32, 128, 256, deflate_slow},
/* 8 */ {32, 128, 258, 1024, deflate_slow},
/* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* maximum compression */
/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
* For deflate_fast() (levels <= 3) good is ignored and lazy has a different
* meaning.
*/
#define EQUAL 0
/* result of memcmp for equal strings */
/* ===========================================================================
* Update a hash value with the given input byte
* IN assertion: all calls to 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.
*/
#define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
/* ===========================================================================
* 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 to INSERT_STRING are made with consecutive
* input characters and the first MIN_MATCH bytes of str are valid
* (except for the last MIN_MATCH-1 bytes of the input file).
*/
#define INSERT_STRING(s, str, match_head) \
(UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
s->prev[(str) & s->w_mask] = match_head = s->head[s->ins_h], \
s->head[s->ins_h] = (Pos)(str))
/* ===========================================================================
* Initialize the hash table (avoiding 64K overflow for 16 bit systems).
* prev[] will be initialized on the fly.
*/
#define CLEAR_HASH(s) \
s->head[s->hash_size-1] = NIL; \
memset((char *)s->head, 0, (unsigned)(s->hash_size-1)*sizeof(*s->head));
/* ========================================================================= */
int zlib_deflateInit2(
z_streamp strm,
int level,
int method,
int windowBits,
int memLevel,
int strategy
)
{
deflate_state *s;
int noheader = 0;
deflate_workspace *mem;
ush *overlay;
/* We overlay pending_buf and d_buf+l_buf. This works since the average
* output size for (length,distance) codes is <= 24 bits.
*/
if (strm == NULL) return Z_STREAM_ERROR;
strm->msg = NULL;
if (level == Z_DEFAULT_COMPRESSION) level = 6;
mem = (deflate_workspace *) strm->workspace;
if (windowBits < 0) { /* undocumented feature: suppress zlib header */
noheader = 1;
windowBits = -windowBits;
}
if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
windowBits < 9 || windowBits > 15 || level < 0 || level > 9 ||
strategy < 0 || strategy > Z_HUFFMAN_ONLY) {
return Z_STREAM_ERROR;
}
s = (deflate_state *) &(mem->deflate_memory);
strm->state = (struct internal_state *)s;
s->strm = strm;
s->noheader = noheader;
s->w_bits = windowBits;
s->w_size = 1 << s->w_bits;
s->w_mask = s->w_size - 1;
s->hash_bits = memLevel + 7;
s->hash_size = 1 << s->hash_bits;
s->hash_mask = s->hash_size - 1;
s->hash_shift = ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
s->window = (Byte *) mem->window_memory;
s->prev = (Pos *) mem->prev_memory;
s->head = (Pos *) mem->head_memory;
s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
overlay = (ush *) mem->overlay_memory;
s->pending_buf = (uch *) overlay;
s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
s->level = level;
s->strategy = strategy;
s->method = (Byte)method;
return zlib_deflateReset(strm);
}
/* ========================================================================= */
#if 0
int zlib_deflateSetDictionary(
z_streamp strm,
const Byte *dictionary,
uInt dictLength
)
{
deflate_state *s;
uInt length = dictLength;
uInt n;
IPos hash_head = 0;
if (strm == NULL || strm->state == NULL || dictionary == NULL)
return Z_STREAM_ERROR;
s = (deflate_state *) strm->state;
if (s->status != INIT_STATE) return Z_STREAM_ERROR;
strm->adler = zlib_adler32(strm->adler, dictionary, dictLength);
if (length < MIN_MATCH) return Z_OK;
if (length > MAX_DIST(s)) {
length = MAX_DIST(s);
#ifndef USE_DICT_HEAD
dictionary += dictLength - length; /* use the tail of the dictionary */
#endif
}
memcpy((char *)s->window, dictionary, length);
s->strstart = length;
s->block_start = (long)length;
/* Insert all strings in the hash table (except for the last two bytes).
* s->lookahead stays null, so s->ins_h will be recomputed at the next
* call of fill_window.
*/
s->ins_h = s->window[0];
UPDATE_HASH(s, s->ins_h, s->window[1]);
for (n = 0; n <= length - MIN_MATCH; n++) {
INSERT_STRING(s, n, hash_head);
}
if (hash_head) hash_head = 0; /* to make compiler happy */
return Z_OK;
}
#endif /* 0 */
/* ========================================================================= */
int zlib_deflateReset(
z_streamp strm
)
{
deflate_state *s;
if (strm == NULL || strm->state == NULL)
return Z_STREAM_ERROR;
strm->total_in = strm->total_out = 0;
strm->msg = NULL;
strm->data_type = Z_UNKNOWN;
s = (deflate_state *)strm->state;
s->pending = 0;
s->pending_out = s->pending_buf;
if (s->noheader < 0) {
s->noheader = 0; /* was set to -1 by deflate(..., Z_FINISH); */
}
s->status = s->noheader ? BUSY_STATE : INIT_STATE;
strm->adler = 1;
s->last_flush = Z_NO_FLUSH;
zlib_tr_init(s);
lm_init(s);
return Z_OK;
}
/* ========================================================================= */
#if 0
int zlib_deflateParams(
z_streamp strm,
int level,
int strategy
)
{
deflate_state *s;
compress_func func;
int err = Z_OK;
if (strm == NULL || strm->state == NULL) return Z_STREAM_ERROR;
s = (deflate_state *) strm->state;
if (level == Z_DEFAULT_COMPRESSION) {
level = 6;
}
if (level < 0 || level > 9 || strategy < 0 || strategy > Z_HUFFMAN_ONLY) {
return Z_STREAM_ERROR;
}
func = configuration_table[s->level].func;
if (func != configuration_table[level].func && strm->total_in != 0) {
/* Flush the last buffer: */
err = zlib_deflate(strm, Z_PARTIAL_FLUSH);
}
if (s->level != level) {
s->level = level;
s->max_lazy_match = configuration_table[level].max_lazy;
s->good_match = configuration_table[level].good_length;
s->nice_match = configuration_table[level].nice_length;
s->max_chain_length = configuration_table[level].max_chain;
}
s->strategy = strategy;
return err;
}
#endif /* 0 */
/* =========================================================================
* Put a short in the pending buffer. The 16-bit value is put in MSB order.
* IN assertion: the stream state is correct and there is enough room in
* pending_buf.
*/
static void putShortMSB(
deflate_state *s,
uInt b
)
{
put_byte(s, (Byte)(b >> 8));
put_byte(s, (Byte)(b & 0xff));
}
/* =========================================================================
* Flush as much pending output as possible. All deflate() output goes
* through this function so some applications may wish to modify it
* to avoid allocating a large strm->next_out buffer and copying into it.
* (See also read_buf()).
*/
static void flush_pending(
z_streamp strm
)
{
deflate_state *s = (deflate_state *) strm->state;
unsigned len = s->pending;
if (len > strm->avail_out) len = strm->avail_out;
if (len == 0) return;
if (strm->next_out != NULL) {
memcpy(strm->next_out, s->pending_out, len);
strm->next_out += len;
}
s->pending_out += len;
strm->total_out += len;
strm->avail_out -= len;
s->pending -= len;
if (s->pending == 0) {
s->pending_out = s->pending_buf;
}
}
/* ========================================================================= */
int zlib_deflate(
z_streamp strm,
int flush
)
{
int old_flush; /* value of flush param for previous deflate call */
deflate_state *s;
if (strm == NULL || strm->state == NULL ||
flush > Z_FINISH || flush < 0) {
return Z_STREAM_ERROR;
}
s = (deflate_state *) strm->state;
if ((strm->next_in == NULL && strm->avail_in != 0) ||
(s->status == FINISH_STATE && flush != Z_FINISH)) {
return Z_STREAM_ERROR;
}
if (strm->avail_out == 0) return Z_BUF_ERROR;
s->strm = strm; /* just in case */
old_flush = s->last_flush;
s->last_flush = flush;
/* Write the zlib header */
if (s->status == INIT_STATE) {
uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
uInt level_flags = (s->level-1) >> 1;
if (level_flags > 3) level_flags = 3;
header |= (level_flags << 6);
if (s->strstart != 0) header |= PRESET_DICT;
header += 31 - (header % 31);
s->status = BUSY_STATE;
putShortMSB(s, header);
/* Save the adler32 of the preset dictionary: */
if (s->strstart != 0) {
putShortMSB(s, (uInt)(strm->adler >> 16));
putShortMSB(s, (uInt)(strm->adler & 0xffff));
}
strm->adler = 1L;
}
/* Flush as much pending output as possible */
if (s->pending != 0) {
flush_pending(strm);
if (strm->avail_out == 0) {
/* Since avail_out is 0, deflate will be called again with
* more output space, but possibly with both pending and
* avail_in equal to zero. There won't be anything to do,
* but this is not an error situation so make sure we
* return OK instead of BUF_ERROR at next call of deflate:
*/
s->last_flush = -1;
return Z_OK;
}
/* Make sure there is something to do and avoid duplicate consecutive
* flushes. For repeated and useless calls with Z_FINISH, we keep
* returning Z_STREAM_END instead of Z_BUFF_ERROR.
*/
} else if (strm->avail_in == 0 && flush <= old_flush &&
flush != Z_FINISH) {
return Z_BUF_ERROR;
}
/* User must not provide more input after the first FINISH: */
if (s->status == FINISH_STATE && strm->avail_in != 0) {
return Z_BUF_ERROR;
}
/* Start a new block or continue the current one.
*/
if (strm->avail_in != 0 || s->lookahead != 0 ||
(flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
block_state bstate;
bstate = (*(configuration_table[s->level].func))(s, flush);
if (bstate == finish_started || bstate == finish_done) {
s->status = FINISH_STATE;
}
if (bstate == need_more || bstate == finish_started) {
if (strm->avail_out == 0) {
s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
}
return Z_OK;
/* If flush != Z_NO_FLUSH && avail_out == 0, the next call
* of deflate should use the same flush parameter to make sure
* that the flush is complete. So we don't have to output an
* empty block here, this will be done at next call. This also
* ensures that for a very small output buffer, we emit at most
* one empty block.
*/
}
if (bstate == block_done) {
if (flush == Z_PARTIAL_FLUSH) {
zlib_tr_align(s);
} else if (flush == Z_PACKET_FLUSH) {
/* Output just the 3-bit `stored' block type value,
but not a zero length. */
zlib_tr_stored_type_only(s);
} else { /* FULL_FLUSH or SYNC_FLUSH */
zlib_tr_stored_block(s, (char*)0, 0L, 0);
/* For a full flush, this empty block will be recognized
* as a special marker by inflate_sync().
*/
if (flush == Z_FULL_FLUSH) {
CLEAR_HASH(s); /* forget history */
}
}
flush_pending(strm);
if (strm->avail_out == 0) {
s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
return Z_OK;
}
}
}
Assert(strm->avail_out > 0, "bug2");
if (flush != Z_FINISH) return Z_OK;
if (s->noheader) return Z_STREAM_END;
/* Write the zlib trailer (adler32) */
putShortMSB(s, (uInt)(strm->adler >> 16));
putShortMSB(s, (uInt)(strm->adler & 0xffff));
flush_pending(strm);
/* If avail_out is zero, the application will call deflate again
* to flush the rest.
*/
s->noheader = -1; /* write the trailer only once! */
return s->pending != 0 ? Z_OK : Z_STREAM_END;
}
/* ========================================================================= */
int zlib_deflateEnd(
z_streamp strm
)
{
int status;
deflate_state *s;
if (strm == NULL || strm->state == NULL) return Z_STREAM_ERROR;
s = (deflate_state *) strm->state;
status = s->status;
if (status != INIT_STATE && status != BUSY_STATE &&
status != FINISH_STATE) {
return Z_STREAM_ERROR;
}
strm->state = NULL;
return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
}
/* =========================================================================
* Copy the source state to the destination state.
*/
#if 0
int zlib_deflateCopy (
z_streamp dest,
z_streamp source
)
{
#ifdef MAXSEG_64K
return Z_STREAM_ERROR;
#else
deflate_state *ds;
deflate_state *ss;
ush *overlay;
deflate_workspace *mem;
if (source == NULL || dest == NULL || source->state == NULL) {
return Z_STREAM_ERROR;
}
ss = (deflate_state *) source->state;
*dest = *source;
mem = (deflate_workspace *) dest->workspace;
ds = &(mem->deflate_memory);
dest->state = (struct internal_state *) ds;
*ds = *ss;
ds->strm = dest;
ds->window = (Byte *) mem->window_memory;
ds->prev = (Pos *) mem->prev_memory;
ds->head = (Pos *) mem->head_memory;
overlay = (ush *) mem->overlay_memory;
ds->pending_buf = (uch *) overlay;
memcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
memcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
memcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
memcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
ds->l_desc.dyn_tree = ds->dyn_ltree;
ds->d_desc.dyn_tree = ds->dyn_dtree;
ds->bl_desc.dyn_tree = ds->bl_tree;
return Z_OK;
#endif
}
#endif /* 0 */
/* ===========================================================================
* Read a new buffer from the current input stream, update the adler32
* and total number of bytes read. All deflate() input goes through
* this function so some applications may wish to modify it to avoid
* allocating a large strm->next_in buffer and copying from it.
* (See also flush_pending()).
*/
static int read_buf(
z_streamp strm,
Byte *buf,
unsigned size
)
{
unsigned len = strm->avail_in;
if (len > size) len = size;
if (len == 0) return 0;
strm->avail_in -= len;
if (!((deflate_state *)(strm->state))->noheader) {
strm->adler = zlib_adler32(strm->adler, strm->next_in, len);
}
memcpy(buf, strm->next_in, len);
strm->next_in += len;
strm->total_in += len;
return (int)len;
}
/* ===========================================================================
* Initialize the "longest match" routines for a new zlib stream
*/
static void lm_init(
deflate_state *s
)
{
s->window_size = (ulg)2L*s->w_size;
CLEAR_HASH(s);
/* Set the default configuration parameters:
*/
s->max_lazy_match = configuration_table[s->level].max_lazy;
s->good_match = configuration_table[s->level].good_length;
s->nice_match = configuration_table[s->level].nice_length;
s->max_chain_length = configuration_table[s->level].max_chain;
s->strstart = 0;
s->block_start = 0L;
s->lookahead = 0;
s->match_length = s->prev_length = MIN_MATCH-1;
s->match_available = 0;
s->ins_h = 0;
}
/* ===========================================================================
* Set match_start to the longest match starting at the given string and
* return its length. Matches shorter or equal to prev_length are discarded,
* in which case the result is equal to prev_length and match_start is
* garbage.
* IN assertions: cur_match is the head of the hash chain for the current
* string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
* OUT assertion: the match length is not greater than s->lookahead.
*/
/* For 80x86 and 680x0, an optimized version will be provided in match.asm or
* match.S. The code will be functionally equivalent.
*/
static uInt longest_match(
deflate_state *s,
IPos cur_match /* current match */
)
{
unsigned chain_length = s->max_chain_length;/* max hash chain length */
register Byte *scan = s->window + s->strstart; /* current string */
register Byte *match; /* matched string */
register int len; /* length of current match */
int best_len = s->prev_length; /* best match length so far */
int nice_match = s->nice_match; /* stop if match long enough */
IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
s->strstart - (IPos)MAX_DIST(s) : NIL;
/* Stop when cur_match becomes <= limit. To simplify the code,
* we prevent matches with the string of window index 0.
*/
Pos *prev = s->prev;
uInt wmask = s->w_mask;
#ifdef UNALIGNED_OK
/* Compare two bytes at a time. Note: this is not always beneficial.
* Try with and without -DUNALIGNED_OK to check.
*/
register Byte *strend = s->window + s->strstart + MAX_MATCH - 1;
register ush scan_start = *(ush*)scan;
register ush scan_end = *(ush*)(scan+best_len-1);
#else
register Byte *strend = s->window + s->strstart + MAX_MATCH;
register Byte scan_end1 = scan[best_len-1];
register Byte scan_end = scan[best_len];
#endif
/* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
* It is easy to get rid of this optimization if necessary.
*/
Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
/* Do not waste too much time if we already have a good match: */
if (s->prev_length >= s->good_match) {
chain_length >>= 2;
}
/* Do not look for matches beyond the end of the input. This is necessary
* to make deflate deterministic.
*/
if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
do {
Assert(cur_match < s->strstart, "no future");
match = s->window + cur_match;
/* Skip to next match if the match length cannot increase
* or if the match length is less than 2:
*/
#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
/* This code assumes sizeof(unsigned short) == 2. Do not use
* UNALIGNED_OK if your compiler uses a different size.
*/
if (*(ush*)(match+best_len-1) != scan_end ||
*(ush*)match != scan_start) continue;
/* It is not necessary to compare scan[2] and match[2] since they are
* always equal when the other bytes match, given that the hash keys
* are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
* strstart+3, +5, ... up to strstart+257. We check for insufficient
* lookahead only every 4th comparison; the 128th check will be made
* at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
* necessary to put more guard bytes at the end of the window, or
* to check more often for insufficient lookahead.
*/
Assert(scan[2] == match[2], "scan[2]?");
scan++, match++;
do {
} while (*(ush*)(scan+=2) == *(ush*)(match+=2) &&
*(ush*)(scan+=2) == *(ush*)(match+=2) &&
*(ush*)(scan+=2) == *(ush*)(match+=2) &&
*(ush*)(scan+=2) == *(ush*)(match+=2) &&
scan < strend);
/* The funny "do {}" generates better code on most compilers */
/* Here, scan <= window+strstart+257 */
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
if (*scan == *match) scan++;
len = (MAX_MATCH - 1) - (int)(strend-scan);
scan = strend - (MAX_MATCH-1);
#else /* UNALIGNED_OK */
if (match[best_len] != scan_end ||
match[best_len-1] != scan_end1 ||
*match != *scan ||
*++match != scan[1]) continue;
/* The check at best_len-1 can be removed because it will be made
* again later. (This heuristic is not always a win.)
* It is not necessary to compare scan[2] and match[2] since they
* are always equal when the other bytes match, given that
* the hash keys are equal and that HASH_BITS >= 8.
*/
scan += 2, match++;
Assert(*scan == *match, "match[2]?");
/* We check for insufficient lookahead only every 8th comparison;
* the 256th check will be made at strstart+258.
*/
do {
} while (*++scan == *++match && *++scan == *++match &&
*++scan == *++match && *++scan == *++match &&
*++scan == *++match && *++scan == *++match &&
*++scan == *++match && *++scan == *++match &&
scan < strend);
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
len = MAX_MATCH - (int)(strend - scan);
scan = strend - MAX_MATCH;
#endif /* UNALIGNED_OK */
if (len > best_len) {
s->match_start = cur_match;
best_len = len;
if (len >= nice_match) break;
#ifdef UNALIGNED_OK
scan_end = *(ush*)(scan+best_len-1);
#else
scan_end1 = scan[best_len-1];
scan_end = scan[best_len];
#endif
}
} while ((cur_match = prev[cur_match & wmask]) > limit
&& --chain_length != 0);
if ((uInt)best_len <= s->lookahead) return best_len;
return s->lookahead;
}
#ifdef DEBUG_ZLIB
/* ===========================================================================
* Check that the match at match_start is indeed a match.
*/
static void check_match(
deflate_state *s,
IPos start,
IPos match,
int length
)
{
/* check that the match is indeed a match */
if (memcmp((char *)s->window + match,
(char *)s->window + start, length) != EQUAL) {
fprintf(stderr, " start %u, match %u, length %d\n",
start, match, length);
do {
fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
} while (--length != 0);
z_error("invalid match");
}
if (z_verbose > 1) {
fprintf(stderr,"\\[%d,%d]", start-match, length);
do { putc(s->window[start++], stderr); } while (--length != 0);
}
}
#else
# define check_match(s, start, match, length)
#endif
/* ===========================================================================
* Fill the window when the lookahead becomes insufficient.
* Updates strstart and lookahead.
*
* IN assertion: lookahead < MIN_LOOKAHEAD
* OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
* At least one byte has been read, or avail_in == 0; reads are
* performed for at least two bytes (required for the zip translate_eol
* option -- not supported here).
*/
static void fill_window(
deflate_state *s
)
{
register unsigned n, m;
register Pos *p;
unsigned more; /* Amount of free space at the end of the window. */
uInt wsize = s->w_size;
do {
more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
/* Deal with !@#$% 64K limit: */
if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
more = wsize;
} else if (more == (unsigned)(-1)) {
/* Very unlikely, but possible on 16 bit machine if strstart == 0
* and lookahead == 1 (input done one byte at time)
*/
more--;
/* If the window is almost full and there is insufficient lookahead,
* move the upper half to the lower one to make room in the upper half.
*/
} else if (s->strstart >= wsize+MAX_DIST(s)) {
memcpy((char *)s->window, (char *)s->window+wsize,
(unsigned)wsize);
s->match_start -= wsize;
s->strstart -= wsize; /* we now have strstart >= MAX_DIST */
s->block_start -= (long) wsize;
/* Slide the hash table (could be avoided with 32 bit values
at the expense of memory usage). We slide even when level == 0
to keep the hash table consistent if we switch back to level > 0
later. (Using level 0 permanently is not an optimal usage of
zlib, so we don't care about this pathological case.)
*/
n = s->hash_size;
p = &s->head[n];
do {
m = *--p;
*p = (Pos)(m >= wsize ? m-wsize : NIL);
} while (--n);
n = wsize;
p = &s->prev[n];
do {
m = *--p;
*p = (Pos)(m >= wsize ? m-wsize : NIL);
/* If n is not on any hash chain, prev[n] is garbage but
* its value will never be used.
*/
} while (--n);
more += wsize;
}
if (s->strm->avail_in == 0) return;
/* If there was no sliding:
* strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
* more == window_size - lookahead - strstart
* => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
* => more >= window_size - 2*WSIZE + 2
* In the BIG_MEM or MMAP case (not yet supported),
* window_size == input_size + MIN_LOOKAHEAD &&
* strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
* Otherwise, window_size == 2*WSIZE so more >= 2.
* If there was sliding, more >= WSIZE. So in all cases, more >= 2.
*/
Assert(more >= 2, "more < 2");
n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
s->lookahead += n;
/* Initialize the hash value now that we have some input: */
if (s->lookahead >= MIN_MATCH) {
s->ins_h = s->window[s->strstart];
UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
#if MIN_MATCH != 3
Call UPDATE_HASH() MIN_MATCH-3 more times
#endif
}
/* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
* but this is not important since only literal bytes will be emitted.
*/
} while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
}
/* ===========================================================================
* Flush the current block, with given end-of-file flag.
* IN assertion: strstart is set to the end of the current match.
*/
#define FLUSH_BLOCK_ONLY(s, eof) { \
zlib_tr_flush_block(s, (s->block_start >= 0L ? \
(char *)&s->window[(unsigned)s->block_start] : \
NULL), \
(ulg)((long)s->strstart - s->block_start), \
(eof)); \
s->block_start = s->strstart; \
flush_pending(s->strm); \
Tracev((stderr,"[FLUSH]")); \
}
/* Same but force premature exit if necessary. */
#define FLUSH_BLOCK(s, eof) { \
FLUSH_BLOCK_ONLY(s, eof); \
if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
}
/* ===========================================================================
* Copy without compression as much as possible from the input stream, return
* the current block state.
* This function does not insert new strings in the dictionary since
* uncompressible data is probably not useful. This function is used
* only for the level=0 compression option.
* NOTE: this function should be optimized to avoid extra copying from
* window to pending_buf.
*/
static block_state deflate_stored(
deflate_state *s,
int flush
)
{
/* Stored blocks are limited to 0xffff bytes, pending_buf is limited
* to pending_buf_size, and each stored block has a 5 byte header:
*/
ulg max_block_size = 0xffff;
ulg max_start;
if (max_block_size > s->pending_buf_size - 5) {
max_block_size = s->pending_buf_size - 5;
}
/* Copy as much as possible from input to output: */
for (;;) {
/* Fill the window as much as possible: */
if (s->lookahead <= 1) {
Assert(s->strstart < s->w_size+MAX_DIST(s) ||
s->block_start >= (long)s->w_size, "slide too late");
fill_window(s);
if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
if (s->lookahead == 0) break; /* flush the current block */
}
Assert(s->block_start >= 0L, "block gone");
s->strstart += s->lookahead;
s->lookahead = 0;
/* Emit a stored block if pending_buf will be full: */
max_start = s->block_start + max_block_size;
if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
/* strstart == 0 is possible when wraparound on 16-bit machine */
s->lookahead = (uInt)(s->strstart - max_start);
s->strstart = (uInt)max_start;
FLUSH_BLOCK(s, 0);
}
/* Flush if we may have to slide, otherwise block_start may become
* negative and the data will be gone:
*/
if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
FLUSH_BLOCK(s, 0);
}
}
FLUSH_BLOCK(s, flush == Z_FINISH);
return flush == Z_FINISH ? finish_done : block_done;
}
/* ===========================================================================
* Compress as much as possible from the input stream, return the current
* block state.
* This function does not perform lazy evaluation of matches and inserts
* new strings in the dictionary only for unmatched strings or for short
* matches. It is used only for the fast compression options.
*/
static block_state deflate_fast(
deflate_state *s,
int flush
)
{
IPos hash_head = NIL; /* head of the hash chain */
int bflush; /* set if current block must be flushed */
for (;;) {
/* Make sure that we always have enough lookahead, except
* at the end of the input file. We need MAX_MATCH bytes
* for the next match, plus MIN_MATCH bytes to insert the
* string following the next match.
*/
if (s->lookahead < MIN_LOOKAHEAD) {
fill_window(s);
if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
return need_more;
}
if (s->lookahead == 0) break; /* flush the current block */
}
/* Insert the string window[strstart .. strstart+2] in the
* dictionary, and set hash_head to the head of the hash chain:
*/
if (s->lookahead >= MIN_MATCH) {
INSERT_STRING(s, s->strstart, hash_head);
}
/* Find the longest match, discarding those <= prev_length.
* At this point we have always match_length < MIN_MATCH
*/
if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
/* 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).
*/
if (s->strategy != Z_HUFFMAN_ONLY) {
s->match_length = longest_match (s, hash_head);
}
/* longest_match() sets match_start */
}
if (s->match_length >= MIN_MATCH) {
check_match(s, s->strstart, s->match_start, s->match_length);
bflush = zlib_tr_tally(s, s->strstart - s->match_start,
s->match_length - MIN_MATCH);
s->lookahead -= s->match_length;
/* Insert new strings in the hash table only if the match length
* is not too large. This saves time but degrades compression.
*/
if (s->match_length <= s->max_insert_length &&
s->lookahead >= MIN_MATCH) {
s->match_length--; /* string at strstart already in hash table */
do {
s->strstart++;
INSERT_STRING(s, s->strstart, hash_head);
/* strstart never exceeds WSIZE-MAX_MATCH, so there are
* always MIN_MATCH bytes ahead.
*/
} while (--s->match_length != 0);
s->strstart++;
} else {
s->strstart += s->match_length;
s->match_length = 0;
s->ins_h = s->window[s->strstart];
UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
#if MIN_MATCH != 3
Call UPDATE_HASH() MIN_MATCH-3 more times
#endif
/* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
* matter since it will be recomputed at next deflate call.
*/
}
} else {
/* No match, output a literal byte */
Tracevv((stderr,"%c", s->window[s->strstart]));
bflush = zlib_tr_tally (s, 0, s->window[s->strstart]);
s->lookahead--;
s->strstart++;
}
if (bflush) FLUSH_BLOCK(s, 0);
}
FLUSH_BLOCK(s, flush == Z_FINISH);
return flush == Z_FINISH ? finish_done : block_done;
}
/* ===========================================================================
* Same as above, but achieves better compression. We use a lazy
* evaluation for matches: a match is finally adopted only if there is
* no better match at the next window position.
*/
static block_state deflate_slow(
deflate_state *s,
int flush
)
{
IPos hash_head = NIL; /* head of hash chain */
int bflush; /* set if current block must be flushed */
/* Process the input block. */
for (;;) {
/* Make sure that we always have enough lookahead, except
* at the end of the input file. We need MAX_MATCH bytes
* for the next match, plus MIN_MATCH bytes to insert the
* string following the next match.
*/
if (s->lookahead < MIN_LOOKAHEAD) {
fill_window(s);
if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
return need_more;
}
if (s->lookahead == 0) break; /* flush the current block */
}
/* Insert the string window[strstart .. strstart+2] in the
* dictionary, and set hash_head to the head of the hash chain:
*/
if (s->lookahead >= MIN_MATCH) {
INSERT_STRING(s, s->strstart, hash_head);
}
/* Find the longest match, discarding those <= prev_length.
*/
s->prev_length = s->match_length, s->prev_match = s->match_start;
s->match_length = MIN_MATCH-1;
if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
s->strstart - hash_head <= MAX_DIST(s)) {
/* 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).
*/
if (s->strategy != Z_HUFFMAN_ONLY) {
s->match_length = longest_match (s, hash_head);
}
/* longest_match() sets match_start */
if (s->match_length <= 5 && (s->strategy == Z_FILTERED ||
(s->match_length == MIN_MATCH &&
s->strstart - s->match_start > TOO_FAR))) {
/* If prev_match is also MIN_MATCH, match_start is garbage
* but we will ignore the current match anyway.
*/
s->match_length = MIN_MATCH-1;
}
}
/* If there was a match at the previous step and the current
* match is not better, output the previous match:
*/
if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
/* Do not insert strings in hash table beyond this. */
check_match(s, s->strstart-1, s->prev_match, s->prev_length);
bflush = zlib_tr_tally(s, s->strstart -1 - s->prev_match,
s->prev_length - MIN_MATCH);
/* Insert in hash table all strings up to the end of the match.
* strstart-1 and strstart are already inserted. If there is not
* enough lookahead, the last two strings are not inserted in
* the hash table.
*/
s->lookahead -= s->prev_length-1;
s->prev_length -= 2;
do {
if (++s->strstart <= max_insert) {
INSERT_STRING(s, s->strstart, hash_head);
}
} while (--s->prev_length != 0);
s->match_available = 0;
s->match_length = MIN_MATCH-1;
s->strstart++;
if (bflush) FLUSH_BLOCK(s, 0);
} else if (s->match_available) {
/* If there was no match at the previous position, output a
* single literal. If there was a match but the current match
* is longer, truncate the previous match to a single literal.
*/
Tracevv((stderr,"%c", s->window[s->strstart-1]));
if (zlib_tr_tally (s, 0, s->window[s->strstart-1])) {
FLUSH_BLOCK_ONLY(s, 0);
}
s->strstart++;
s->lookahead--;
if (s->strm->avail_out == 0) return need_more;
} else {
/* There is no previous match to compare with, wait for
* the next step to decide.
*/
s->match_available = 1;
s->strstart++;
s->lookahead--;
}
}
Assert (flush != Z_NO_FLUSH, "no flush?");
if (s->match_available) {
Tracevv((stderr,"%c", s->window[s->strstart-1]));
zlib_tr_tally (s, 0, s->window[s->strstart-1]);
s->match_available = 0;
}
FLUSH_BLOCK(s, flush == Z_FINISH);
return flush == Z_FINISH ? finish_done : block_done;
}
int zlib_deflate_workspacesize(void)
{
return sizeof(deflate_workspace);
}