diff --git a/libentcode/entdec.h b/libentcode/entdec.h index b78b85a3..ddcbb6f7 100644 --- a/libentcode/entdec.h +++ b/libentcode/entdec.h @@ -49,6 +49,7 @@ unsigned ec_decode(ec_dec *_this,unsigned _ft); _fl: The cumulative frequency of all symbols that come before the symbol decoded. _fh: The cumulative frequency of all symbols up to and including the symbol + decoded. Together with _fl, this defines the range [_fl,_fh) in which the value returned above must fall. _ft: The total frequency of the symbols in the alphabet the symbol decoded diff --git a/libentcode/entenc.h b/libentcode/entenc.h index feb08a56..04ed0c14 100644 --- a/libentcode/entenc.h +++ b/libentcode/entenc.h @@ -15,7 +15,7 @@ struct ec_enc{ ec_byte_buffer *buf; /*A buffered output symbol, awaiting carry propagation.*/ int rem; - /*Number of extra carry propogating symbols.*/ + /*Number of extra carry propagating symbols.*/ size_t ext; /*The number of values in the current range.*/ ec_uint32 rng; @@ -34,10 +34,14 @@ void ec_enc_init(ec_enc *_this,ec_byte_buffer *_buf); It is allowable to change the frequency information, or even the entire source alphabet, so long as the decoder can tell from the context of the previously encoded information that it is supposed to do so as well. - _fl: The sum of the frequencies of symbols before the one to be encoded. - _fs: The frequency of the symbol to be encoded. + _fl: The cumulative frequency of all symbols that come before the one to be + encoded. + _fh: The cumulative frequency of all symbols up to and including the one to + be encoded. + Together with _fl, this defines the range [_fl,_fh) in which the + decoded value will fall. _ft: The sum of the frequencies of all the symbols*/ -void ec_encode(ec_enc *_this,unsigned _fl,unsigned _fs,unsigned _ft); +void ec_encode(ec_enc *_this,unsigned _fl,unsigned _fh,unsigned _ft); /*Encodes a sequence of raw bits in the stream. _fl: The bits to encode. _ftb: The number of bits to encode. diff --git a/libentcode/mfrngcod.h b/libentcode/mfrngcod.h index ccded287..a45e5b10 100644 --- a/libentcode/mfrngcod.h +++ b/libentcode/mfrngcod.h @@ -16,7 +16,7 @@ # define EC_CODE_TOP (1U<>EC_SYM_BITS) -/*Code for which propogating carries are possible.*/ +/*Code for which propagating carries are possible.*/ # define EC_CODE_CARRY (EC_SYM_MAX<rng<=EC_CODE_BOT){ @@ -176,7 +176,7 @@ static void ec_dec_normalize(ec_dec *_this){ /*dif can never be larger than EC_CODE_TOP. This is equivalent to the slightly more readable: if(_this->dif>EC_CODE_TOP)_this->dif-=EC_CODE_TOP;*/ - _this->dif^=(_this->dif&_this->dif-1)&EC_CODE_TOP; + _this->dif^=_this->dif&_this->dif-1&EC_CODE_TOP; } } @@ -248,7 +248,7 @@ int ec_dec_done(ec_dec *_this){ msk=EC_CODE_TOP-1; do{ msk>>=1; - end=(low+msk)&~msk|msk+1; + end=low+msk&~msk|msk+1; } while(end-low>=_this->rng); } diff --git a/libentcode/mfrngenc.c b/libentcode/mfrngenc.c index 126bc47a..31b4d607 100644 --- a/libentcode/mfrngenc.c +++ b/libentcode/mfrngenc.c @@ -41,15 +41,18 @@ /*Outputs a symbol, with a carry bit. - If there is a potential to propogate a carry over several symbols, they are + If there is a potential to propagate a carry over several symbols, they are buffered until it can be determined whether or not an actual carry will occur. - If the counter for the buffered symbols overflows, then the range is - truncated to force a carry to occur, towards whichever side maximizes the - remaining range.*/ + If the counter for the buffered symbols overflows, then the stream becomes + undecodable. + This gives a theoretical limit of a few billion symbols in a single packet on + 32-bit systems. + The alternative is to truncate the range in order to force a carry, but + requires similar carry tracking in the decoder, needlessly slowing it down.*/ static void ec_enc_carry_out(ec_enc *_this,int _c){ if(_c!=EC_SYM_MAX){ - /*No further carry propogation possible, flush buffer.*/ + /*No further carry propagation possible, flush buffer.*/ int carry; carry=_c>>EC_SYM_BITS; /*Don't output a byte on the first write. @@ -129,7 +132,7 @@ void ec_enc_done(ec_enc *_this){ msk=EC_CODE_TOP-1; do{ msk>>=1; - end=(_this->low+msk)&~msk|msk+1; + end=_this->low+msk&~msk|msk+1; } while(end-_this->low>=_this->rng); } diff --git a/libentcode/rangedec.c b/libentcode/rangedec.c index be93fcfd..5b2d6e12 100644 --- a/libentcode/rangedec.c +++ b/libentcode/rangedec.c @@ -90,7 +90,7 @@ @PHDTHESIS{Pas76, author="Richard Clark Pasco", - title="Sorce coding algorithms for fast data compression", + title="Source coding algorithms for fast data compression", school="Dept. of Electrical Engineering, Stanford University", address="Stanford, CA", month=May, @@ -159,8 +159,8 @@ static int ec_dec_in(ec_dec *_this){ else return ret; } -/*Normalizes the contents of low and rng so that rng is contained in the - high-order symbol of low.*/ +/*Normalizes the contents of dif 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){ @@ -176,7 +176,7 @@ static void ec_dec_normalize(ec_dec *_this){ /*dif can never be larger than EC_CODE_TOP. This is equivalent to the slightly more readable: if(_this->dif>EC_CODE_TOP)_this->dif-=EC_CODE_TOP;*/ - _this->dif^=(_this->dif&_this->dif-1)&EC_CODE_TOP; + _this->dif^=_this->dif&_this->dif-1&EC_CODE_TOP; } } diff --git a/libentcode/rangeenc.c b/libentcode/rangeenc.c index 29ea2919..bb5384b4 100644 --- a/libentcode/rangeenc.c +++ b/libentcode/rangeenc.c @@ -41,15 +41,18 @@ /*Outputs a symbol, with a carry bit. - If there is a potential to propogate a carry over several symbols, they are + If there is a potential to propagate a carry over several symbols, they are buffered until it can be determined whether or not an actual carry will occur. - If the counter for the buffered symbols overflows, then the range is - truncated to force a carry to occur, towards whichever side maximizes the - remaining range.*/ + If the counter for the buffered symbols overflows, then the stream becomes + undecodable. + This gives a theoretical limit of a few billion symbols in a single packet on + 32-bit systems. + The alternative is to truncate the range in order to force a carry, but + requires similar carry tracking in the decoder, needlessly slowing it down.*/ static void ec_enc_carry_out(ec_enc *_this,int _c){ if(_c!=EC_SYM_MAX){ - /*No further carry propogation possible, flush buffer.*/ + /*No further carry propagation possible, flush buffer.*/ int carry; carry=_c>>EC_SYM_BITS; /*Don't output a byte on the first write. @@ -110,7 +113,7 @@ void ec_enc_done(ec_enc *_this){ msk=EC_CODE_TOP-1; do{ msk>>=1; - end=(_this->low+msk)&~msk|msk+1; + end=_this->low+msk&~msk|msk+1; } while(end-_this->low>=_this->rng); }