/src/opus/celt/entdec.c

Source
/* Copyright (c) 2001-2011 Timothy B. Terriberry
   Copyright (c) 2008-2009 Xiph.Org Foundation */
/*
   Redistribution and use in source and binary forms, with or without
   modification, are permitted provided that the following conditions
   are met:

   - Redistributions of source code must retain the above copyright
   notice, this list of conditions and the following disclaimer.

   - Redistributions in binary form must reproduce the above copyright
   notice, this list of conditions and the following disclaimer in the
   documentation and/or other materials provided with the distribution.

   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
   ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
   OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
   EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
   PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
   PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
   LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
   NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
   SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <stddef.h>
#include "os_support.h"
#include "arch.h"
#include "entdec.h"
#include "mfrngcod.h"

/*A range decoder.
  This is an entropy decoder based upon \cite{Mar79}, which is itself a
   rediscovery of the FIFO arithmetic code introduced by \cite{Pas76}.
  It is very similar to arithmetic encoding, except that encoding is done with
   digits in any base, instead of with bits, and so it is faster when using
   larger bases (i.e.: a byte).
  The author claims an average waste of $\frac{1}{2}\log_b(2b)$ bits, where $b$
   is the base, longer than the theoretical optimum, but to my knowledge there
   is no published justification for this claim.
  This only seems true when using near-infinite precision arithmetic so that
   the process is carried out with no rounding errors.

  An excellent description of implementation details is available at
   http://www.arturocampos.com/ac_range.html
  A recent work \cite{MNW98} which proposes several changes to arithmetic
   encoding for efficiency actually re-discovers many of the principles
   behind range encoding, and presents a good theoretical analysis of them.

  End of stream is handled by writing out the smallest number of bits that
   ensures that the stream will be correctly decoded regardless of the value of
   any subsequent bits.
  ec_tell() can be used to determine how many bits were needed to decode
   all the symbols thus far; other data can be packed in the remaining bits of
   the input buffer.
  @PHDTHESIS{Pas76,
    author="Richard Clark Pasco",
    title="Source coding algorithms for fast data compression",
    school="Dept. of Electrical Engineering, Stanford University",
    address="Stanford, CA",
    month=May,
    year=1976
  }
  @INPROCEEDINGS{Mar79,
   author="Martin, G.N.N.",
   title="Range encoding: an algorithm for removing redundancy from a digitised
    message",
   booktitle="Video & Data Recording Conference",
   year=1979,
   address="Southampton",
   month=Jul
  }
  @ARTICLE{MNW98,
   author="Alistair Moffat and Radford Neal and Ian H. Witten",
   title="Arithmetic Coding Revisited",
   journal="{ACM} Transactions on Information Systems",
   year=1998,
   volume=16,
   number=3,
   pages="256--294",
   month=Jul,
   URL="http://www.stanford.edu/class/ee398a/handouts/papers/Moffat98ArithmCoding.pdf"
  }*/

static int ec_read_byte(ec_dec *_this){
  return _this->offs<_this->storage?_this->buf[_this->offs++]:0;
}

static int ec_read_byte_from_end(ec_dec *_this){
  return _this->end_offs<_this->storage?
   _this->buf[_this->storage-++(_this->end_offs)]:0;
}

/*Normalizes the contents of val and rng so that rng lies entirely in the
   high-order symbol.*/
static void ec_dec_normalize(ec_dec *_this){
  /*If the range is too small, rescale it and input some bits.*/
  while(_this->rng<=EC_CODE_BOT){
    int sym;
    _this->nbits_total+=EC_SYM_BITS;
    _this->rng<<=EC_SYM_BITS;
    /*Use up the remaining bits from our last symbol.*/
    sym=_this->rem;
    /*Read the next value from the input.*/
    _this->rem=ec_read_byte(_this);
    /*Take the rest of the bits we need from this new symbol.*/
    sym=(sym<<EC_SYM_BITS|_this->rem)>>(EC_SYM_BITS-EC_CODE_EXTRA);
    /*And subtract them from val, capped to be less than EC_CODE_TOP.*/
    _this->val=((_this->val<<EC_SYM_BITS)+(EC_SYM_MAX&~sym))&(EC_CODE_TOP-1);
  }
}

void ec_dec_init(ec_dec *_this,unsigned char *_buf,opus_uint32 _storage){
  _this->buf=_buf;
  _this->storage=_storage;
  _this->end_offs=0;
  _this->end_window=0;
  _this->nend_bits=0;
  /*This is the offset from which ec_tell() will subtract partial bits.
    The final value after the ec_dec_normalize() call will be the same as in
     the encoder, but we have to compensate for the bits that are added there.*/
  _this->nbits_total=EC_CODE_BITS+1
   -((EC_CODE_BITS-EC_CODE_EXTRA)/EC_SYM_BITS)*EC_SYM_BITS;
  _this->offs=0;
  _this->rng=1U<<EC_CODE_EXTRA;
  _this->rem=ec_read_byte(_this);
  _this->val=_this->rng-1-(_this->rem>>(EC_SYM_BITS-EC_CODE_EXTRA));
  _this->error=0;
  /*Normalize the interval.*/
  ec_dec_normalize(_this);
}

unsigned ec_decode(ec_dec *_this,unsigned _ft){
  unsigned s;
  _this->ext=celt_udiv(_this->rng,_ft);
  s=(unsigned)(_this->val/_this->ext);
  return _ft-EC_MINI(s+1,_ft);
}

unsigned ec_decode_bin(ec_dec *_this,unsigned _bits){
   unsigned s;
   _this->ext=_this->rng>>_bits;
   s=(unsigned)(_this->val/_this->ext);
   return (1U<<_bits)-EC_MINI(s+1U,1U<<_bits);
}

void ec_dec_update(ec_dec *_this,unsigned _fl,unsigned _fh,unsigned _ft){
  opus_uint32 s;
  s=IMUL32(_this->ext,_ft-_fh);
  _this->val-=s;
  _this->rng=_fl>0?IMUL32(_this->ext,_fh-_fl):_this->rng-s;
  ec_dec_normalize(_this);
}

/*The probability of having a "one" is 1/(1<<_logp).*/
int ec_dec_bit_logp(ec_dec *_this,unsigned _logp){
  opus_uint32 r;
  opus_uint32 d;
  opus_uint32 s;
  int         ret;
  r=_this->rng;
  d=_this->val;
  s=r>>_logp;
  ret=d<s;
  if(!ret)_this->val=d-s;
  _this->rng=ret?s:r-s;
  ec_dec_normalize(_this);
  return ret;
}

int ec_dec_icdf(ec_dec *_this,const unsigned char *_icdf,unsigned _ftb){
  opus_uint32 r;
  opus_uint32 d;
  opus_uint32 s;
  opus_uint32 t;
  int         ret;
  s=_this->rng;
  d=_this->val;
  r=s>>_ftb;
  ret=-1;
  do{
    t=s;
    s=IMUL32(r,_icdf[++ret]);
  }
  while(d<s);
  _this->val=d-s;
  _this->rng=t-s;
  ec_dec_normalize(_this);
  return ret;
}

int ec_dec_icdf16(ec_dec *_this,const opus_uint16 *_icdf,unsigned _ftb){
  opus_uint32 r;
  opus_uint32 d;
  opus_uint32 s;
  opus_uint32 t;
  int         ret;
  s=_this->rng;
  d=_this->val;
  r=s>>_ftb;
  ret=-1;
  do{
    t=s;
    s=IMUL32(r,_icdf[++ret]);
  }
  while(d<s);
  _this->val=d-s;
  _this->rng=t-s;
  ec_dec_normalize(_this);
  return ret;
}

opus_uint32 ec_dec_uint(ec_dec *_this,opus_uint32 _ft){
  unsigned ft;
  unsigned s;
  int      ftb;
  /*In order to optimize EC_ILOG(), it is undefined for the value 0.*/
  celt_assert(_ft>1);
  _ft--;
  ftb=EC_ILOG(_ft);
  if(ftb>EC_UINT_BITS){
    opus_uint32 t;
    ftb-=EC_UINT_BITS;
    ft=(unsigned)(_ft>>ftb)+1;
    s=ec_decode(_this,ft);
    ec_dec_update(_this,s,s+1,ft);
    t=(opus_uint32)s<<ftb|ec_dec_bits(_this,ftb);
    if(t<=_ft)return t;
    _this->error=1;
    return _ft;
  }
  else{
    _ft++;
    s=ec_decode(_this,(unsigned)_ft);
    ec_dec_update(_this,s,s+1,(unsigned)_ft);
    return s;
  }
}

opus_uint32 ec_dec_bits(ec_dec *_this,unsigned _bits){
  ec_window   window;
  int         available;
  opus_uint32 ret;
  window=_this->end_window;
  available=_this->nend_bits;
  if((unsigned)available<_bits){
    do{
      window|=(ec_window)ec_read_byte_from_end(_this)<<available;
      available+=EC_SYM_BITS;
    }
    while(available<=EC_WINDOW_SIZE-EC_SYM_BITS);
  }
  ret=(opus_uint32)window&(((opus_uint32)1<<_bits)-1U);
  window>>=_bits;
  available-=_bits;
  _this->end_window=window;
  _this->nend_bits=available;
  _this->nbits_total+=_bits;
  return ret;
}

Coverage Report

Created: 2026-07-25 07:52

Line	Count	Source
1		/* Copyright (c) 2001-2011 Timothy B. Terriberry
2		Copyright (c) 2008-2009 Xiph.Org Foundation */
3		/*
4		Redistribution and use in source and binary forms, with or without
5		modification, are permitted provided that the following conditions
6		are met:
7
8		- Redistributions of source code must retain the above copyright
9		notice, this list of conditions and the following disclaimer.
10
11		- Redistributions in binary form must reproduce the above copyright
12		notice, this list of conditions and the following disclaimer in the
13		documentation and/or other materials provided with the distribution.
14
15		THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
16		``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
17		LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
18		A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
19		OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
20		EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
21		PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
22		PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
23		LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
24		NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
25		SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26		*/
27
28		#ifdef HAVE_CONFIG_H
29		#include "config.h"
30		#endif
31
32		#include <stddef.h>
33		#include "os_support.h"
34		#include "arch.h"
35		#include "entdec.h"
36		#include "mfrngcod.h"
37
38		/*A range decoder.
39		This is an entropy decoder based upon \cite{Mar79}, which is itself a
40		rediscovery of the FIFO arithmetic code introduced by \cite{Pas76}.
41		It is very similar to arithmetic encoding, except that encoding is done with
42		digits in any base, instead of with bits, and so it is faster when using
43		larger bases (i.e.: a byte).
44		The author claims an average waste of $\frac{1}{2}\log_b(2b)$ bits, where $b$
45		is the base, longer than the theoretical optimum, but to my knowledge there
46		is no published justification for this claim.
47		This only seems true when using near-infinite precision arithmetic so that
48		the process is carried out with no rounding errors.
49
50		An excellent description of implementation details is available at
51		http://www.arturocampos.com/ac_range.html
52		A recent work \cite{MNW98} which proposes several changes to arithmetic
53		encoding for efficiency actually re-discovers many of the principles
54		behind range encoding, and presents a good theoretical analysis of them.
55
56		End of stream is handled by writing out the smallest number of bits that
57		ensures that the stream will be correctly decoded regardless of the value of
58		any subsequent bits.
59		ec_tell() can be used to determine how many bits were needed to decode
60		all the symbols thus far; other data can be packed in the remaining bits of
61		the input buffer.
62		@PHDTHESIS{Pas76,
63		author="Richard Clark Pasco",
64		title="Source coding algorithms for fast data compression",
65		school="Dept. of Electrical Engineering, Stanford University",
66		address="Stanford, CA",
67		month=May,
68		year=1976
69		}
70		@INPROCEEDINGS{Mar79,
71		author="Martin, G.N.N.",
72		title="Range encoding: an algorithm for removing redundancy from a digitised
73		message",
74		booktitle="Video & Data Recording Conference",
75		year=1979,
76		address="Southampton",
77		month=Jul
78		}
79		@ARTICLE{MNW98,
80		author="Alistair Moffat and Radford Neal and Ian H. Witten",
81		title="Arithmetic Coding Revisited",
82		journal="{ACM} Transactions on Information Systems",
83		year=1998,
84		volume=16,
85		number=3,
86		pages="256--294",
87		month=Jul,
88		URL="http://www.stanford.edu/class/ee398a/handouts/papers/Moffat98ArithmCoding.pdf"
89		}*/
90
91	7.13M	static int ec_read_byte(ec_dec *_this){
92	7.13M	return _this->offs<_this->storage?_this->buf[_this->offs++]:0;
93	7.13M	}
94
95	668k	static int ec_read_byte_from_end(ec_dec *_this){
96	668k	return _this->end_offs<_this->storage?
97	664k	_this->buf[_this->storage-++(_this->end_offs)]:0;
98	668k	}
99
100		/*Normalizes the contents of val and rng so that rng lies entirely in the
101		high-order symbol.*/
102	35.1M	static void ec_dec_normalize(ec_dec *_this){
103		/If the range is too small, rescale it and input some bits./
104	42.1M	while(_this->rng<=EC_CODE_BOT){
105	7.00M	int sym;
106	7.00M	_this->nbits_total+=EC_SYM_BITS;
107	7.00M	_this->rng<<=EC_SYM_BITS;
108		/Use up the remaining bits from our last symbol./
109	7.00M	sym=_this->rem;
110		/Read the next value from the input./
111	7.00M	_this->rem=ec_read_byte(_this);
112		/Take the rest of the bits we need from this new symbol./
113	7.00M	sym=(sym<<EC_SYM_BITS\|_this->rem)>>(EC_SYM_BITS-EC_CODE_EXTRA);
114		/And subtract them from val, capped to be less than EC_CODE_TOP./
115	7.00M	_this->val=((_this->val<<EC_SYM_BITS)+(EC_SYM_MAX&~sym))&(EC_CODE_TOP-1);
116	7.00M	}
117	35.1M	}
118
119	135k	void ec_dec_init(ec_dec _this,unsigned char _buf,opus_uint32 _storage){
120	135k	_this->buf=_buf;
121	135k	_this->storage=_storage;
122	135k	_this->end_offs=0;
123	135k	_this->end_window=0;
124	135k	_this->nend_bits=0;
125		/*This is the offset from which ec_tell() will subtract partial bits.
126		The final value after the ec_dec_normalize() call will be the same as in
127		the encoder, but we have to compensate for the bits that are added there.*/
128	135k	_this->nbits_total=EC_CODE_BITS+1
129	135k	-((EC_CODE_BITS-EC_CODE_EXTRA)/EC_SYM_BITS)*EC_SYM_BITS;
130	135k	_this->offs=0;
131	135k	_this->rng=1U<<EC_CODE_EXTRA;
132	135k	_this->rem=ec_read_byte(_this);
133	135k	_this->val=_this->rng-1-(_this->rem>>(EC_SYM_BITS-EC_CODE_EXTRA));
134	135k	_this->error=0;
135		/Normalize the interval./
136	135k	ec_dec_normalize(_this);
137	135k	}
138
139	756k	unsigned ec_decode(ec_dec *_this,unsigned _ft){
140	756k	unsigned s;
141	756k	_this->ext=celt_udiv(_this->rng,_ft);
142	756k	s=(unsigned)(_this->val/_this->ext);
143	756k	return _ft-EC_MINI(s+1,_ft);
144	756k	}
145
146	889k	unsigned ec_decode_bin(ec_dec *_this,unsigned _bits){
147	889k	unsigned s;
148	889k	_this->ext=_this->rng>>_bits;
149	889k	s=(unsigned)(_this->val/_this->ext);
150	889k	return (1U<<_bits)-EC_MINI(s+1U,1U<<_bits);
151	889k	}
152
153	1.64M	void ec_dec_update(ec_dec *_this,unsigned _fl,unsigned _fh,unsigned _ft){
154	1.64M	opus_uint32 s;
155	1.64M	s=IMUL32(_this->ext,_ft-_fh);
156	1.64M	_this->val-=s;
157	1.64M	_this->rng=_fl>0?IMUL32(_this->ext,_fh-_fl):_this->rng-s;
158	1.64M	ec_dec_normalize(_this);
159	1.64M	}
160
161		/The probability of having a "one" is 1/(1<<_logp)./
162	2.00M	int ec_dec_bit_logp(ec_dec *_this,unsigned _logp){
163	2.00M	opus_uint32 r;
164	2.00M	opus_uint32 d;
165	2.00M	opus_uint32 s;
166	2.00M	int ret;
167	2.00M	r=_this->rng;
168	2.00M	d=_this->val;
169	2.00M	s=r>>_logp;
170	2.00M	ret=d<s;
171	2.00M	if(!ret)_this->val=d-s;
172	2.00M	_this->rng=ret?s:r-s;
173	2.00M	ec_dec_normalize(_this);
174	2.00M	return ret;
175	2.00M	}
176
177	31.3M	int ec_dec_icdf(ec_dec _this,const unsigned char _icdf,unsigned _ftb){
178	31.3M	opus_uint32 r;
179	31.3M	opus_uint32 d;
180	31.3M	opus_uint32 s;
181	31.3M	opus_uint32 t;
182	31.3M	int ret;
183	31.3M	s=_this->rng;
184	31.3M	d=_this->val;
185	31.3M	r=s>>_ftb;
186	31.3M	ret=-1;
187	79.7M	do{
188	79.7M	t=s;
189	79.7M	s=IMUL32(r,_icdf[++ret]);
190	79.7M	}
191	79.7M	while(d<s);
192	31.3M	_this->val=d-s;
193	31.3M	_this->rng=t-s;
194	31.3M	ec_dec_normalize(_this);
195	31.3M	return ret;
196	31.3M	}
197
198	0	int ec_dec_icdf16(ec_dec _this,const opus_uint16 _icdf,unsigned _ftb){
199	0	opus_uint32 r;
200	0	opus_uint32 d;
201	0	opus_uint32 s;
202	0	opus_uint32 t;
203	0	int ret;
204	0	s=_this->rng;
205	0	d=_this->val;
206	0	r=s>>_ftb;
207	0	ret=-1;
208	0	do{
209	0	t=s;
210	0	s=IMUL32(r,_icdf[++ret]);
211	0	}
212	0	while(d<s);
213	0	_this->val=d-s;
214	0	_this->rng=t-s;
215	0	ec_dec_normalize(_this);
216	0	return ret;
217	0	}
218
219	619k	opus_uint32 ec_dec_uint(ec_dec *_this,opus_uint32 _ft){
220	619k	unsigned ft;
221	619k	unsigned s;
222	619k	int ftb;
223		/In order to optimize EC_ILOG(), it is undefined for the value 0./
224	619k	celt_assert(_ft>1);
225	619k	_ft--;
226	619k	ftb=EC_ILOG(_ft);
227	619k	if(ftb>EC_UINT_BITS){
228	389k	opus_uint32 t;
229	389k	ftb-=EC_UINT_BITS;
230	389k	ft=(unsigned)(_ft>>ftb)+1;
231	389k	s=ec_decode(_this,ft);
232	389k	ec_dec_update(_this,s,s+1,ft);
233	389k	t=(opus_uint32)s<<ftb\|ec_dec_bits(_this,ftb);
234	389k	if(t<=_ft)return t;
235	770	_this->error=1;
236	770	return _ft;
237	389k	}
238	230k	else{
239	230k	_ft++;
240	230k	s=ec_decode(_this,(unsigned)_ft);
241	230k	ec_dec_update(_this,s,s+1,(unsigned)_ft);
242	230k	return s;
243	230k	}
244	619k	}
245
246	1.16M	opus_uint32 ec_dec_bits(ec_dec *_this,unsigned _bits){
247	1.16M	ec_window window;
248	1.16M	int available;
249	1.16M	opus_uint32 ret;
250	1.16M	window=_this->end_window;
251	1.16M	available=_this->nend_bits;
252	1.16M	if((unsigned)available<_bits){
253	668k	do{
254	668k	window\|=(ec_window)ec_read_byte_from_end(_this)<<available;
255	668k	available+=EC_SYM_BITS;
256	668k	}
257	668k	while(available<=EC_WINDOW_SIZE-EC_SYM_BITS);
258	201k	}
259	1.16M	ret=(opus_uint32)window&(((opus_uint32)1<<_bits)-1U);
260	1.16M	window>>=_bits;
261	1.16M	available-=_bits;
262	1.16M	_this->end_window=window;
263	1.16M	_this->nend_bits=available;
264	1.16M	_this->nbits_total+=_bits;
265	1.16M	return ret;
266	1.16M	}