1 /*
2  * $Source: /homes/cvs/ftape-stacked/ftape/compressor/lzrw3.c,v $
3  * $Revision: 1.1 $
4  * $Date: 1997/10/05 19:12:29 $
5  *
6  * Implementation of Ross Williams lzrw3 algorithm. Adaption for zftape.
7  *
8  */
9 
10 #include "../compressor/lzrw3.h"       /* Defines single exported function "compress".   */
11 
12 /******************************************************************************/
13 /*                                                                            */
14 /*                                    LZRW3.C                                 */
15 /*                                                                            */
16 /******************************************************************************/
17 /*                                                                            */
18 /* Author  : Ross Williams.                                                   */
19 /* Date    : 30-Jun-1991.                                                     */
20 /* Release : 1.                                                               */
21 /*                                                                            */
22 /******************************************************************************/
23 /*                                                                            */
24 /* This file contains an implementation of the LZRW3 data compression         */
25 /* algorithm in C.                                                            */
26 /*                                                                            */
27 /* The algorithm is a general purpose compression algorithm that runs fast    */
28 /* and gives reasonable compression. The algorithm is a member of the Lempel  */
29 /* Ziv family of algorithms and bases its compression on the presence in the  */
30 /* data of repeated substrings.                                               */
31 /*                                                                            */
32 /* This algorithm is unpatented and the code is public domain. As the         */
33 /* algorithm is based on the LZ77 class of algorithms, it is unlikely to be   */
34 /* the subject of a patent challenge.                                         */
35 /*                                                                            */
36 /* Unlike the LZRW1 and LZRW1-A algorithms, the LZRW3 algorithm is            */
37 /* deterministic and is guaranteed to yield the same compressed               */
38 /* representation for a given file each time it is run.                       */
39 /*                                                                            */
40 /* The LZRW3 algorithm was originally designed and implemented                */
41 /* by Ross Williams on 31-Dec-1990.                                           */
42 /*                                                                            */
43 /* Here are the results of applying this code, compiled under THINK C 4.0     */
44 /* and running on a Mac-SE (8MHz 68000), to the standard calgary corpus.      */
45 /*                                                                            */
46 /*    +----------------------------------------------------------------+      */
47 /*    | DATA COMPRESSION TEST                                          |      */
48 /*    | =====================                                          |      */
49 /*    | Time of run     : Sun 30-Jun-1991 09:31PM                      |      */
50 /*    | Timing accuracy : One part in 100                              |      */
51 /*    | Context length  : 262144 bytes (= 256.0000K)                   |      */
52 /*    | Test suite      : Calgary Corpus Suite                         |      */
53 /*    | Files in suite  : 14                                           |      */
54 /*    | Algorithm       : LZRW3                                        |      */
55 /*    | Note: All averages are calculated from the un-rounded values.  |      */
56 /*    +----------------------------------------------------------------+      */
57 /*    | File Name   Length  CxB  ComLen  %Remn  Bits  Com K/s  Dec K/s |      */
58 /*    | ----------  ------  ---  ------  -----  ----  -------  ------- |      */
59 /*    | rpus:Bib.D  111261    1   55033   49.5  3.96    19.46    32.27 |      */
60 /*    | us:Book1.D  768771    3  467962   60.9  4.87    17.03    31.07 |      */
61 /*    | us:Book2.D  610856    3  317102   51.9  4.15    19.39    34.15 |      */
62 /*    | rpus:Geo.D  102400    1   82424   80.5  6.44    11.65    18.18 |      */
63 /*    | pus:News.D  377109    2  205670   54.5  4.36    17.14    27.47 |      */
64 /*    | pus:Obj1.D   21504    1   13027   60.6  4.85    13.40    18.95 |      */
65 /*    | pus:Obj2.D  246814    1  116286   47.1  3.77    19.31    30.10 |      */
66 /*    | s:Paper1.D   53161    1   27522   51.8  4.14    18.60    31.15 |      */
67 /*    | s:Paper2.D   82199    1   45160   54.9  4.40    18.45    32.84 |      */
68 /*    | rpus:Pic.D  513216    2  122388   23.8  1.91    35.29    51.05 |      */
69 /*    | us:Progc.D   39611    1   19669   49.7  3.97    18.87    30.64 |      */
70 /*    | us:Progl.D   71646    1   28247   39.4  3.15    24.34    40.66 |      */
71 /*    | us:Progp.D   49379    1   19377   39.2  3.14    23.91    39.23 |      */
72 /*    | us:Trans.D   93695    1   33481   35.7  2.86    25.48    40.37 |      */
73 /*    +----------------------------------------------------------------+      */
74 /*    | Average     224401    1  110953   50.0  4.00    20.17    32.72 |      */
75 /*    +----------------------------------------------------------------+      */
76 /*                                                                            */
77 /******************************************************************************/
78 
79 /******************************************************************************/
80 
81 /* The following structure is returned by the "compress" function below when  */
82 /* the user asks the function to return identifying information.              */
83 /* The most important field in the record is the working memory field which   */
84 /* tells the calling program how much working memory should be passed to      */
85 /* "compress" when it is called to perform a compression or decompression.    */
86 /* LZRW3 uses the same amount of memory during compression and decompression. */
87 /* For more information on this structure see "compress.h".                   */
88 
89 #define U(X)            ((ULONG) X)
90 #define SIZE_P_BYTE     (U(sizeof(UBYTE *)))
91 #define SIZE_WORD       (U(sizeof(UWORD  )))
92 #define ALIGNMENT_FUDGE (U(16))
93 #define MEM_REQ ( U(4096)*(SIZE_P_BYTE) + ALIGNMENT_FUDGE )
94 
95 static struct compress_identity identity =
96 {
97  U(0x032DDEA8),                           /* Algorithm identification number. */
98  MEM_REQ,                                 /* Working memory (bytes) required. */
99  "LZRW3",                                 /* Name of algorithm.               */
100  "1.0",                                   /* Version number of algorithm.     */
101  "31-Dec-1990",                           /* Date of algorithm.               */
102  "Public Domain",                         /* Copyright notice.                */
103  "Ross N. Williams",                      /* Author of algorithm.             */
104  "Renaissance Software",                  /* Affiliation of author.           */
105  "Public Domain"                          /* Vendor of algorithm.             */
106 };
107 
108 LOCAL void compress_compress  (UBYTE *,UBYTE *,ULONG,UBYTE *, LONG *);
109 LOCAL void compress_decompress(UBYTE *,UBYTE *,LONG, UBYTE *, ULONG *);
110 
111 /******************************************************************************/
112 
113 /* This function is the only function exported by this module.                */
114 /* Depending on its first parameter, the function can be requested to         */
115 /* compress a block of memory, decompress a block of memory, or to identify   */
116 /* itself. For more information, see the specification file "compress.h".     */
117 
lzrw3_compress(action,wrk_mem,src_adr,src_len,dst_adr,p_dst_len)118 EXPORT void lzrw3_compress(action,wrk_mem,src_adr,src_len,dst_adr,p_dst_len)
119 UWORD     action;      /* Action to be performed.                             */
120 UBYTE   *wrk_mem;      /* Address of working memory we can use.               */
121 UBYTE   *src_adr;      /* Address of input data.                              */
122 LONG     src_len;      /* Length  of input data.                              */
123 UBYTE   *dst_adr;      /* Address to put output data.                         */
124 void  *p_dst_len;      /* Address of longword for length of output data.      */
125 {
126  switch (action)
127    {
128     case COMPRESS_ACTION_IDENTITY:
129        *((struct compress_identity **)p_dst_len)= &identity;
130        break;
131     case COMPRESS_ACTION_COMPRESS:
132        compress_compress(wrk_mem,src_adr,src_len,dst_adr,(LONG *)p_dst_len);
133        break;
134     case COMPRESS_ACTION_DECOMPRESS:
135        compress_decompress(wrk_mem,src_adr,src_len,dst_adr,(LONG *)p_dst_len);
136        break;
137    }
138 }
139 
140 /******************************************************************************/
141 /*                                                                            */
142 /* BRIEF DESCRIPTION OF THE LZRW3 ALGORITHM                                   */
143 /* ========================================                                   */
144 /* The LZRW3 algorithm is identical to the LZRW1-A algorithm except that      */
145 /* instead of transmitting history offsets, it transmits hash table indexes.  */
146 /* In order to decode the indexes, the decompressor must maintain an          */
147 /* identical hash table. Copy items are straightforward:when the decompressor */
148 /* receives a copy item, it simply looks up the hash table to translate the   */
149 /* index into a pointer into the data already decompressed. To update the     */
150 /* hash table, it replaces the same table entry with a pointer to the start   */
151 /* of the newly decoded phrase. The tricky part is with literal items, for at */
152 /* the time that the decompressor receives a literal item the decompressor    */
153 /* does not have the three bytes in the Ziv (that the compressor has) to      */
154 /* perform the three-byte hash. To solve this problem, in LZRW3, both the     */
155 /* compressor and decompressor are wired up so that they "buffer" these       */
156 /* literals and update their hash tables only when three bytes are available. */
157 /* This makes the maximum buffering 2 bytes.                                  */
158 /*                                                                            */
159 /* Replacement of offsets by hash table indexes yields a few percent extra    */
160 /* compression at the cost of some speed. LZRW3 is slower than LZRW1, LZRW1-A */
161 /* and LZRW2, but yields better compression.                                  */
162 /*                                                                            */
163 /* Extra compression could be obtained by using a hash table of depth two.    */
164 /* However, increasing the depth above one incurs a significant decrease in   */
165 /* compression speed which was not considered worthwhile. Another reason for  */
166 /* keeping the depth down to one was to allow easy comparison with the        */
167 /* LZRW1-A and LZRW2 algorithms so as to demonstrate the exact effect of the  */
168 /* use of direct hash indexes.                                                */
169 /*                                                                            */
170 /*                                  +---+                                     */
171 /*                                  |___|4095                                 */
172 /*                                  |___|                                     */
173 /*              +---------------------*_|<---+   /----+---\                   */
174 /*              |                   |___|    +---|Hash    |                   */
175 /*              |                   |___|        |Function|                   */
176 /*              |                   |___|        \--------/                   */
177 /*              |                   |___|0            ^                       */
178 /*              |                   +---+             |                       */
179 /*              |                   Hash        +-----+                       */
180 /*              |                   Table       |                             */
181 /*              |                              ---                            */
182 /*              v                              ^^^                            */
183 /*      +-------------------------------------|----------------+              */
184 /*      ||||||||||||||||||||||||||||||||||||||||||||||||||||||||              */
185 /*      +-------------------------------------|----------------+              */
186 /*      |                                     |1......18|      |              */
187 /*      |<------- Lempel=History ------------>|<--Ziv-->|      |              */
188 /*      |     (=bytes already processed)      |<-Still to go-->|              */
189 /*      |<-------------------- INPUT BLOCK ------------------->|              */
190 /*                                                                            */
191 /* The diagram above for LZRW3 looks almost identical to the diagram for      */
192 /* LZRW1. The difference is that in LZRW3, the compressor transmits hash      */
193 /* table indices instead of Lempel offsets. For this to work, the             */
194 /* decompressor must maintain a hash table as well as the compressor and both */
195 /* compressor and decompressor must "buffer" literals, as the decompressor    */
196 /* cannot hash phrases commencing with a literal until another two bytes have */
197 /* arrived.                                                                   */
198 /*                                                                            */
199 /*  LZRW3 Algorithm Execution Summary                                         */
200 /*  ---------------------------------                                         */
201 /*  1. Hash the first three bytes of the Ziv to yield a hash table index h.   */
202 /*  2. Look up the hash table yielding history pointer p.                     */
203 /*  3. Match where p points with the Ziv. If there is a match of three or     */
204 /*     more bytes, code those bytes (in the Ziv) as a copy item, otherwise    */
205 /*     code the next byte in the Ziv as a literal item.                       */
206 /*  4. Update the hash table as possible subject to the constraint that only  */
207 /*     phrases commencing three bytes back from the Ziv can be hashed and     */
208 /*     entered into the hash table. (This enables the decompressor to keep    */
209 /*     pace). See the description and code for more details.                  */
210 /*                                                                            */
211 /******************************************************************************/
212 /*                                                                            */
213 /*                     DEFINITION OF COMPRESSED FILE FORMAT                   */
214 /*                     ====================================                   */
215 /*  * A compressed file consists of a COPY FLAG followed by a REMAINDER.      */
216 /*  * The copy flag CF uses up four bytes with the first byte being the       */
217 /*    least significant.                                                      */
218 /*  * If CF=1, then the compressed file represents the remainder of the file  */
219 /*    exactly. Otherwise CF=0 and the remainder of the file consists of zero  */
220 /*    or more GROUPS, each of which represents one or more bytes.             */
221 /*  * Each group consists of two bytes of CONTROL information followed by     */
222 /*    sixteen ITEMs except for the last group which can contain from one      */
223 /*    to sixteen items.                                                       */
224 /*  * An item can be either a LITERAL item or a COPY item.                    */
225 /*  * Each item corresponds to a bit in the control bytes.                    */
226 /*  * The first control byte corresponds to the first 8 items in the group    */
227 /*    with bit 0 corresponding to the first item in the group and bit 7 to    */
228 /*    the eighth item in the group.                                           */
229 /*  * The second control byte corresponds to the second 8 items in the group  */
230 /*    with bit 0 corresponding to the ninth item in the group and bit 7 to    */
231 /*    the sixteenth item in the group.                                        */
232 /*  * A zero bit in a control word means that the corresponding item is a     */
233 /*    literal item. A one bit corresponds to a copy item.                     */
234 /*  * A literal item consists of a single byte which represents itself.       */
235 /*  * A copy item consists of two bytes that represent from 3 to 18 bytes.    */
236 /*  * The first  byte in a copy item will be denoted C1.                      */
237 /*  * The second byte in a copy item will be denoted C2.                      */
238 /*  * Bits will be selected using square brackets.                            */
239 /*    For example: C1[0..3] is the low nibble of the first control byte.      */
240 /*    of copy item C1.                                                        */
241 /*  * The LENGTH of a copy item is defined to be C1[0..3]+3 which is a number */
242 /*    in the range [3,18].                                                    */
243 /*  * The INDEX of a copy item is defined to be C1[4..7]*256+C2[0..8] which   */
244 /*    is a number in the range [0,4095].                                      */
245 /*  * A copy item represents the sequence of bytes                            */
246 /*       text[POS-OFFSET..POS-OFFSET+LENGTH-1] where                          */
247 /*          text   is the entire text of the uncompressed string.             */
248 /*          POS    is the index in the text of the character following the    */
249 /*                   string represented by all the items preceeding the item  */
250 /*                   being defined.                                           */
251 /*          OFFSET is obtained from INDEX by looking up the hash table.       */
252 /*                                                                            */
253 /******************************************************************************/
254 
255 /* The following #define defines the length of the copy flag that appears at  */
256 /* the start of the compressed file. The value of four bytes was chosen       */
257 /* because the fast_copy routine on my Macintosh runs faster if the source    */
258 /* and destination blocks are relatively longword aligned.                    */
259 /* The actual flag data appears in the first byte. The rest are zeroed so as  */
260 /* to normalize the compressed representation (i.e. not non-deterministic).   */
261 #define FLAG_BYTES 4
262 
263 /* The following #defines define the meaning of the values of the copy        */
264 /* flag at the start of the compressed file.                                  */
265 #define FLAG_COMPRESS 0     /* Signals that output was result of compression. */
266 #define FLAG_COPY     1     /* Signals that output was simply copied over.    */
267 
268 /* The 68000 microprocessor (on which this algorithm was originally developed */
269 /* is fussy about non-aligned arrays of words. To avoid these problems the    */
270 /* following macro can be used to "waste" from 0 to 3 bytes so as to align    */
271 /* the argument pointer.                                                      */
272 #define ULONG_ALIGN_UP(X) ((((ULONG)X)+sizeof(ULONG)-1)&~(sizeof(ULONG)-1))
273 
274 
275 /* The following constant defines the maximum length of an uncompressed item. */
276 /* This definition must not be changed; its value is hardwired into the code. */
277 /* The longest number of bytes that can be spanned by a single item is 18     */
278 /* for the longest copy item.                                                 */
279 #define MAX_RAW_ITEM (18)
280 
281 /* The following constant defines the maximum length of an uncompressed group.*/
282 /* This definition must not be changed; its value is hardwired into the code. */
283 /* A group contains at most 16 items which explains this definition.          */
284 #define MAX_RAW_GROUP (16*MAX_RAW_ITEM)
285 
286 /* The following constant defines the maximum length of a compressed group.   */
287 /* This definition must not be changed; its value is hardwired into the code. */
288 /* A compressed group consists of two control bytes followed by up to 16      */
289 /* compressed items each of which can have a maximum length of two bytes.     */
290 #define MAX_CMP_GROUP (2+16*2)
291 
292 /* The following constant defines the number of entries in the hash table.    */
293 /* This definition must not be changed; its value is hardwired into the code. */
294 #define HASH_TABLE_LENGTH (4096)
295 
296 /* LZRW3, unlike LZRW1(-A), must initialize its hash table so as to enable    */
297 /* the compressor and decompressor to stay in step maintaining identical hash */
298 /* tables. In an early version of the algorithm, the tables were simply       */
299 /* initialized to zero and a check for zero was included just before the      */
300 /* matching code. However, this test costs time. A better solution is to      */
301 /* initialize all the entries in the hash table to point to a constant        */
302 /* string. The decompressor does the same. This solution requires no extra    */
303 /* test. The contents of the string do not matter so long as the string is    */
304 /* the same for the compressor and decompressor and contains at least         */
305 /* MAX_RAW_ITEM bytes. I chose consecutive decimal digits because they do not */
306 /* have white space problems (e.g. there is no chance that the compiler will  */
307 /* replace more than one space by a TAB) and because they make the length of  */
308 /* the string obvious by inspection.                                          */
309 #define START_STRING_18 ((UBYTE *) "123456789012345678")
310 
311 /* In this algorithm, hash values have to be calculated at more than one      */
312 /* point. The following macro neatens the code up for this.                   */
313 #define HASH(PTR) \
314    (((40543*(((*(PTR))<<8)^((*((PTR)+1))<<4)^(*((PTR)+2))))>>4) & 0xFFF)
315 
316 /******************************************************************************/
317 
compress_compress(p_wrk_mem,p_src_first,src_len,p_dst_first,p_dst_len)318 LOCAL void compress_compress
319            (p_wrk_mem,p_src_first,src_len,p_dst_first,p_dst_len)
320 /* Input  : Hand over the required amount of working memory in p_wrk_mem.     */
321 /* Input  : Specify input block using p_src_first and src_len.                */
322 /* Input  : Point p_dst_first to the start of the output zone (OZ).           */
323 /* Input  : Point p_dst_len to a ULONG to receive the output length.          */
324 /* Input  : Input block and output zone must not overlap.                     */
325 /* Output : Length of output block written to *p_dst_len.                     */
326 /* Output : Output block in Mem[p_dst_first..p_dst_first+*p_dst_len-1]. May   */
327 /* Output : write in OZ=Mem[p_dst_first..p_dst_first+src_len+MAX_CMP_GROUP-1].*/
328 /* Output : Upon completion guaranteed *p_dst_len<=src_len+FLAG_BYTES.        */
329 UBYTE *p_wrk_mem;
330 UBYTE *p_src_first;
331 ULONG  src_len;
332 UBYTE *p_dst_first;
333 LONG  *p_dst_len;
334 {
335  /* p_src and p_dst step through the source and destination blocks.           */
336  register UBYTE *p_src = p_src_first;
337  register UBYTE *p_dst = p_dst_first;
338 
339  /* The following variables are never modified and are used in the            */
340  /* calculations that determine when the main loop terminates.                */
341  UBYTE *p_src_post  = p_src_first+src_len;
342  UBYTE *p_dst_post  = p_dst_first+src_len;
343  UBYTE *p_src_max1  = p_src_first+src_len-MAX_RAW_ITEM;
344  UBYTE *p_src_max16 = p_src_first+src_len-MAX_RAW_ITEM*16;
345 
346  /* The variables 'p_control' and 'control' are used to buffer control bits.  */
347  /* Before each group is processed, the next two bytes of the output block    */
348  /* are set aside for the control word for the group about to be processed.   */
349  /* 'p_control' is set to point to the first byte of that word. Meanwhile,    */
350  /* 'control' buffers the control bits being generated during the processing  */
351  /* of the group. Instead of having a counter to keep track of how many items */
352  /* have been processed (=the number of bits in the control word), at the     */
353  /* start of each group, the top word of 'control' is filled with 1 bits.     */
354  /* As 'control' is shifted for each item, the 1 bits in the top word are     */
355  /* absorbed or destroyed. When they all run out (i.e. when the top word is   */
356  /* all zero bits, we know that we are at the end of a group.                 */
357 # define TOPWORD 0xFFFF0000
358  UBYTE *p_control;
359  register ULONG control=TOPWORD;
360 
361  /* THe variable 'hash' always points to the first element of the hash table. */
362  UBYTE **hash= (UBYTE **)  ULONG_ALIGN_UP(p_wrk_mem);
363 
364  /* The following two variables represent the literal buffer. p_h1 points to  */
365  /* the hash table entry corresponding to the youngest literal. p_h2 points   */
366  /* to the hash table entry corresponding to the second youngest literal.     */
367  /* Note: p_h1=0=>p_h2=0 because zero values denote absence of a pending      */
368  /* literal. The variables are initialized to zero meaning an empty "buffer". */
369  UBYTE **p_h1=0;
370  UBYTE **p_h2=0;
371 
372  /* To start, we write the flag bytes. Being optimistic, we set the flag to   */
373  /* FLAG_COMPRESS. The remaining flag bytes are zeroed so as to keep the      */
374  /* algorithm deterministic.                                                  */
375  *p_dst++=FLAG_COMPRESS;
376  {UWORD i; for (i=2;i<=FLAG_BYTES;i++) *p_dst++=0;}
377 
378  /* Reserve the first word of output as the control word for the first group. */
379  /* Note: This is undone at the end if the input block is empty.              */
380  p_control=p_dst; p_dst+=2;
381 
382  /* Initialize all elements of the hash table to point to a constant string.  */
383  /* Use of an unrolled loop speeds this up considerably.                      */
384  {UWORD i; UBYTE **p_h=hash;
385 #  define ZH *p_h++=START_STRING_18
386   for (i=0;i<256;i++)     /* 256=HASH_TABLE_LENGTH/16. */
387     {ZH;ZH;ZH;ZH;
388      ZH;ZH;ZH;ZH;
389      ZH;ZH;ZH;ZH;
390      ZH;ZH;ZH;ZH;}
391  }
392 
393  /* The main loop processes either 1 or 16 items per iteration. As its        */
394  /* termination logic is complicated, I have opted for an infinite loop       */
395  /* structure containing 'break' and 'goto' statements.                       */
396  while (TRUE)
397    {/* Begin main processing loop. */
398 
399     /* Note: All the variables here except unroll should be defined within    */
400     /*       the inner loop. Unfortunately the loop hasn't got a block.       */
401      register UBYTE *p;         /* Scans through targ phrase during matching. */
402      register UBYTE *p_ziv= NULL ;     /* Points to first byte of current Ziv.       */
403      register UWORD unroll;     /* Loop counter for unrolled inner loop.      */
404      register UWORD index;      /* Index of current hash table entry.         */
405      register UBYTE **p_h0 = NULL ;     /* Pointer to current hash table entry.       */
406 
407     /* Test for overrun and jump to overrun code if necessary.                */
408     if (p_dst>p_dst_post)
409        goto overrun;
410 
411     /* The following cascade of if statements efficiently catches and deals   */
412     /* with varying degrees of closeness to the end of the input block.       */
413     /* When we get very close to the end, we stop updating the table and      */
414     /* code the remaining bytes as literals. This makes the code simpler.     */
415     unroll=16;
416     if (p_src>p_src_max16)
417       {
418        unroll=1;
419        if (p_src>p_src_max1)
420          {
421           if (p_src==p_src_post)
422              break;
423           else
424              goto literal;
425          }
426       }
427 
428     /* This inner unrolled loop processes 'unroll' (whose value is either 1   */
429     /* or 16) items. I have chosen to implement this loop with labels and     */
430     /* gotos to heighten the ease with which the loop may be implemented with */
431     /* a single decrement and branch instruction in assembly language and     */
432     /* also because the labels act as highly readable place markers.          */
433     /* (Also because we jump into the loop for endgame literals (see above)). */
434 
435     begin_unrolled_loop:
436 
437        /* To process the next phrase, we hash the next three bytes and use    */
438        /* the resultant hash table index to look up the hash table. A pointer */
439        /* to the entry is stored in p_h0 so as to avoid an array lookup. The  */
440        /* hash table entry *p_h0 is looked up yielding a pointer p to a       */
441        /* potential match of the Ziv in the history.                          */
442        index=HASH(p_src);
443        p_h0=&hash[index];
444        p=*p_h0;
445 
446        /* Having looked up the candidate position, we are in a position to    */
447        /* attempt a match. The match loop has been unrolled using the PS      */
448        /* macro so that failure within the first three bytes automatically    */
449        /* results in the literal branch being taken. The coding is simple.    */
450        /* p_ziv saves p_src so we can let p_src wander.                       */
451 #       define PS *p++!=*p_src++
452        p_ziv=p_src;
453        if (PS || PS || PS)
454          {
455           /* Literal. */
456 
457           /* Code the literal byte as itself and a zero control bit.          */
458           p_src=p_ziv; literal: *p_dst++=*p_src++; control&=0xFFFEFFFF;
459 
460           /* We have just coded a literal. If we had two pending ones, that   */
461           /* makes three and we can update the hash table.                    */
462           if (p_h2!=0)
463              {*p_h2=p_ziv-2;}
464 
465           /* In any case, rotate the hash table pointers for next time. */
466           p_h2=p_h1; p_h1=p_h0;
467 
468          }
469        else
470          {
471           /* Copy */
472 
473           /* Match up to 15 remaining bytes using an unrolled loop and code. */
474 #if 0
475           PS || PS || PS || PS || PS || PS || PS || PS ||
476           PS || PS || PS || PS || PS || PS || PS || p_src++;
477 #else
478           if (
479                !( PS || PS || PS || PS || PS || PS || PS || PS ||
480                   PS || PS || PS || PS || PS || PS || PS )
481              ) p_src++;
482 #endif
483           *p_dst++=((index&0xF00)>>4)|(--p_src-p_ziv-3);
484           *p_dst++=index&0xFF;
485 
486           /* As we have just coded three bytes, we are now in a position to   */
487           /* update the hash table with the literal bytes that were pending   */
488           /* upon the arrival of extra context bytes.                         */
489           if (p_h1!=0)
490             {
491              if (p_h2!=0)
492                {*p_h2=p_ziv-2; p_h2=0;}
493              *p_h1=p_ziv-1; p_h1=0;
494             }
495 
496           /* In any case, we can update the hash table based on the current   */
497           /* position as we just coded at least three bytes in a copy items.  */
498           *p_h0=p_ziv;
499 
500          }
501        control>>=1;
502 
503        /* This loop is all set up for a decrement and jump instruction! */
504 #ifndef linux
505 `    end_unrolled_loop: if (--unroll) goto begin_unrolled_loop;
506 #else
507     /* end_unrolled_loop: */ if (--unroll) goto begin_unrolled_loop;
508 #endif
509 
510     /* At this point it will nearly always be the end of a group in which     */
511     /* case, we have to do some control-word processing. However, near the    */
512     /* end of the input block, the inner unrolled loop is only executed once. */
513     /* This necessitates the 'if' test.                                       */
514     if ((control&TOPWORD)==0)
515       {
516        /* Write the control word to the place we saved for it in the output. */
517        *p_control++=  control     &0xFF;
518        *p_control  = (control>>8) &0xFF;
519 
520        /* Reserve the next word in the output block for the control word */
521        /* for the group about to be processed.                           */
522        p_control=p_dst; p_dst+=2;
523 
524        /* Reset the control bits buffer. */
525        control=TOPWORD;
526       }
527 
528    } /* End main processing loop. */
529 
530  /* After the main processing loop has executed, all the input bytes have     */
531  /* been processed. However, the control word has still to be written to the  */
532  /* word reserved for it in the output at the start of the most recent group. */
533  /* Before writing, the control word has to be shifted so that all the bits   */
534  /* are in the right place. The "empty" bit positions are filled with 1s      */
535  /* which partially fill the top word.                                        */
536  while(control&TOPWORD) control>>=1;
537  *p_control++= control     &0xFF;
538  *p_control++=(control>>8) &0xFF;
539 
540  /* If the last group contained no items, delete the control word too.        */
541  if (p_control==p_dst) p_dst-=2;
542 
543  /* Write the length of the output block to the dst_len parameter and return. */
544  *p_dst_len=p_dst-p_dst_first;
545  return;
546 
547  /* Jump here as soon as an overrun is detected. An overrun is defined to     */
548  /* have occurred if p_dst>p_dst_first+src_len. That is, the moment the       */
549  /* length of the output written so far exceeds the length of the input block.*/
550  /* The algorithm checks for overruns at least at the end of each group       */
551  /* which means that the maximum overrun is MAX_CMP_GROUP bytes.              */
552  /* Once an overrun occurs, the only thing to do is to set the copy flag and  */
553  /* copy the input over.                                                      */
554  overrun:
555 #if 0
556  *p_dst_first=FLAG_COPY;
557  fast_copy(p_src_first,p_dst_first+FLAG_BYTES,src_len);
558  *p_dst_len=src_len+FLAG_BYTES;
559 #else
560  fast_copy(p_src_first,p_dst_first,src_len);
561  *p_dst_len= -src_len; /* return a negative number to indicate uncompressed data */
562 #endif
563 }
564 
565 /******************************************************************************/
566 
compress_decompress(p_wrk_mem,p_src_first,src_len,p_dst_first,p_dst_len)567 LOCAL void compress_decompress
568            (p_wrk_mem,p_src_first,src_len,p_dst_first,p_dst_len)
569 /* Input  : Hand over the required amount of working memory in p_wrk_mem.     */
570 /* Input  : Specify input block using p_src_first and src_len.                */
571 /* Input  : Point p_dst_first to the start of the output zone.                */
572 /* Input  : Point p_dst_len to a ULONG to receive the output length.          */
573 /* Input  : Input block and output zone must not overlap. User knows          */
574 /* Input  : upperbound on output block length from earlier compression.       */
575 /* Input  : In any case, maximum expansion possible is nine times.            */
576 /* Output : Length of output block written to *p_dst_len.                     */
577 /* Output : Output block in Mem[p_dst_first..p_dst_first+*p_dst_len-1].       */
578 /* Output : Writes only  in Mem[p_dst_first..p_dst_first+*p_dst_len-1].       */
579 UBYTE *p_wrk_mem;
580 UBYTE *p_src_first;
581 LONG   src_len;
582 UBYTE *p_dst_first;
583 ULONG *p_dst_len;
584 {
585  /* Byte pointers p_src and p_dst scan through the input and output blocks.   */
586  register UBYTE *p_src = p_src_first+FLAG_BYTES;
587  register UBYTE *p_dst = p_dst_first;
588  /* we need to avoid a SEGV when trying to uncompress corrupt data */
589  register UBYTE *p_dst_post = p_dst_first + *p_dst_len;
590 
591  /* The following two variables are never modified and are used to control    */
592  /* the main loop.                                                            */
593  UBYTE *p_src_post  = p_src_first+src_len;
594  UBYTE *p_src_max16 = p_src_first+src_len-(MAX_CMP_GROUP-2);
595 
596  /* The hash table is the only resident of the working memory. The hash table */
597  /* contains HASH_TABLE_LENGTH=4096 pointers to positions in the history. To  */
598  /* keep Macintoshes happy, it is longword aligned.                           */
599  UBYTE **hash = (UBYTE **) ULONG_ALIGN_UP(p_wrk_mem);
600 
601  /* The variable 'control' is used to buffer the control bits which appear in */
602  /* groups of 16 bits (control words) at the start of each compressed group.  */
603  /* When each group is read, bit 16 of the register is set to one. Whenever   */
604  /* a new bit is needed, the register is shifted right. When the value of the */
605  /* register becomes 1, we know that we have reached the end of a group.      */
606  /* Initializing the register to 1 thus instructs the code to follow that it  */
607  /* should read a new control word immediately.                               */
608  register ULONG control=1;
609 
610  /* The value of 'literals' is always in the range 0..3. It is the number of  */
611  /* consecutive literal items just seen. We have to record this number so as  */
612  /* to know when to update the hash table. When literals gets to 3, there     */
613  /* have been three consecutive literals and we can update at the position of */
614  /* the oldest of the three.                                                  */
615  register UWORD literals=0;
616 
617  /* Check the leading copy flag to see if the compressor chose to use a copy  */
618  /* operation instead of a compression operation. If a copy operation was     */
619  /* used, then all we need to do is copy the data over, set the output length */
620  /* and return.                                                               */
621 #if 0
622  if (*p_src_first==FLAG_COPY)
623    {
624     fast_copy(p_src_first+FLAG_BYTES,p_dst_first,src_len-FLAG_BYTES);
625     *p_dst_len=src_len-FLAG_BYTES;
626     return;
627    }
628 #else
629   if ( src_len < 0 )
630   {
631    fast_copy(p_src_first,p_dst_first,-src_len );
632    *p_dst_len = (ULONG)-src_len;
633    return;
634   }
635 #endif
636 
637  /* Initialize all elements of the hash table to point to a constant string.  */
638  /* Use of an unrolled loop speeds this up considerably.                      */
639  {UWORD i; UBYTE **p_h=hash;
640 #  define ZJ *p_h++=START_STRING_18
641   for (i=0;i<256;i++)     /* 256=HASH_TABLE_LENGTH/16. */
642     {ZJ;ZJ;ZJ;ZJ;
643      ZJ;ZJ;ZJ;ZJ;
644      ZJ;ZJ;ZJ;ZJ;
645      ZJ;ZJ;ZJ;ZJ;}
646  }
647 
648  /* The outer loop processes either 1 or 16 items per iteration depending on  */
649  /* how close p_src is to the end of the input block.                         */
650  while (p_src!=p_src_post)
651    {/* Start of outer loop */
652 
653     register UWORD unroll;   /* Counts unrolled loop executions.              */
654 
655     /* When 'control' has the value 1, it means that the 16 buffered control  */
656     /* bits that were read in at the start of the current group have all been */
657     /* shifted out and that all that is left is the 1 bit that was injected   */
658     /* into bit 16 at the start of the current group. When we reach the end   */
659     /* of a group, we have to load a new control word and inject a new 1 bit. */
660     if (control==1)
661       {
662        control=0x10000|*p_src++;
663        control|=(*p_src++)<<8;
664       }
665 
666     /* If it is possible that we are within 16 groups from the end of the     */
667     /* input, execute the unrolled loop only once, else process a whole group */
668     /* of 16 items by looping 16 times.                                       */
669     unroll= p_src<=p_src_max16 ? 16 : 1;
670 
671     /* This inner loop processes one phrase (item) per iteration. */
672     while (unroll--)
673       { /* Begin unrolled inner loop. */
674 
675        /* Process a literal or copy item depending on the next control bit. */
676        if (control&1)
677          {
678           /* Copy item. */
679 
680           register UBYTE *p;           /* Points to place from which to copy. */
681           register UWORD lenmt;        /* Length of copy item minus three.    */
682           register UBYTE **p_hte;      /* Pointer to current hash table entry.*/
683           register UBYTE *p_ziv=p_dst; /* Pointer to start of current Ziv.    */
684 
685           /* Read and dismantle the copy word. Work out from where to copy.   */
686           lenmt=*p_src++;
687           p_hte=&hash[((lenmt&0xF0)<<4)|*p_src++];
688           p=*p_hte;
689           lenmt&=0xF;
690 
691           /* Now perform the copy using a half unrolled loop. */
692           *p_dst++=*p++;
693           *p_dst++=*p++;
694           *p_dst++=*p++;
695           while (lenmt--)
696              *p_dst++=*p++;
697 
698           /* Because we have just received 3 or more bytes in a copy item     */
699           /* (whose bytes we have just installed in the output), we are now   */
700           /* in a position to flush all the pending literal hashings that had */
701           /* been postponed for lack of bytes.                                */
702           if (literals>0)
703             {
704              register UBYTE *r=p_ziv-literals;;
705              hash[HASH(r)]=r;
706              if (literals==2)
707                 {r++; hash[HASH(r)]=r;}
708              literals=0;
709             }
710 
711           /* In any case, we can immediately update the hash table with the   */
712           /* current position. We don't need to do a HASH(...) to work out    */
713           /* where to put the pointer, as the compressor just told us!!!      */
714           *p_hte=p_ziv;
715 
716          }
717        else
718          {
719           /* Literal item. */
720 
721           /* Copy over the literal byte. */
722           *p_dst++=*p_src++;
723 
724           /* If we now have three literals waiting to be hashed into the hash */
725           /* table, we can do one of them now (because there are three).      */
726           if (++literals == 3)
727              {register UBYTE *p=p_dst-3; hash[HASH(p)]=p; literals=2;}
728          }
729 
730        /* Shift the control buffer so the next control bit is in bit 0. */
731        control>>=1;
732 #if 1
733        if (p_dst > p_dst_post)
734        {
735 	       /* Shit: we tried to decompress corrupt data */
736 	       *p_dst_len = 0;
737 	       return;
738        }
739 #endif
740       } /* End unrolled inner loop. */
741 
742    } /* End of outer loop */
743 
744  /* Write the length of the decompressed data before returning. */
745   *p_dst_len=p_dst-p_dst_first;
746 }
747 
748 /******************************************************************************/
749 /*                               End of LZRW3.C                               */
750 /******************************************************************************/
751