drivers/char/ftape/compressor/lzrw3.c

   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
 118 EXPORT void lzrw3_compress(
 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 {
 127  switch (action)
 128    {
 129     case COMPRESS_ACTION_IDENTITY:
 130        *((struct compress_identity **)p_dst_len)= &identity;
 131        break;
 132     case COMPRESS_ACTION_COMPRESS:
 133        compress_compress(wrk_mem,src_adr,src_len,dst_adr,(LONG *)p_dst_len);
 134        break;
 135     case COMPRESS_ACTION_DECOMPRESS:
 136        compress_decompress(wrk_mem,src_adr,src_len,dst_adr,(LONG *)p_dst_len);
 137        break;
 138    }
 139 }
 140
 141 /******************************************************************************/
 142 /*                                                                            */
 143 /* BRIEF DESCRIPTION OF THE LZRW3 ALGORITHM                                   */
 144 /* ========================================                                   */
 145 /* The LZRW3 algorithm is identical to the LZRW1-A algorithm except that      */
 146 /* instead of transmitting history offsets, it transmits hash table indexes.  */
 147 /* In order to decode the indexes, the decompressor must maintain an          */
 148 /* identical hash table. Copy items are straightforward:when the decompressor */
 149 /* receives a copy item, it simply looks up the hash table to translate the   */
 150 /* index into a pointer into the data already decompressed. To update the     */
 151 /* hash table, it replaces the same table entry with a pointer to the start   */
 152 /* of the newly decoded phrase. The tricky part is with literal items, for at */
 153 /* the time that the decompressor receives a literal item the decompressor    */
 154 /* does not have the three bytes in the Ziv (that the compressor has) to      */
 155 /* perform the three-byte hash. To solve this problem, in LZRW3, both the     */
 156 /* compressor and decompressor are wired up so that they "buffer" these       */
 157 /* literals and update their hash tables only when three bytes are available. */
 158 /* This makes the maximum buffering 2 bytes.                                  */
 159 /*                                                                            */
 160 /* Replacement of offsets by hash table indexes yields a few percent extra    */
 161 /* compression at the cost of some speed. LZRW3 is slower than LZRW1, LZRW1-A */
 162 /* and LZRW2, but yields better compression.                                  */
 163 /*                                                                            */
 164 /* Extra compression could be obtained by using a hash table of depth two.    */
 165 /* However, increasing the depth above one incurs a significant decrease in   */
 166 /* compression speed which was not considered worthwhile. Another reason for  */
 167 /* keeping the depth down to one was to allow easy comparison with the        */
 168 /* LZRW1-A and LZRW2 algorithms so as to demonstrate the exact effect of the  */
 169 /* use of direct hash indexes.                                                */
 170 /*                                                                            */
 171 /*                                  +---+                                     */
 172 /*                                  |___|4095                                 */
 173 /*                                  |___|                                     */
 174 /*              +---------------------*_|<---+   /----+---\                   */
 175 /*              |                   |___|    +---|Hash    |                   */
 176 /*              |                   |___|        |Function|                   */
 177 /*              |                   |___|        \--------/                   */
 178 /*              |                   |___|0            ^                       */
 179 /*              |                   +---+             |                       */
 180 /*              |                   Hash        +-----+                       */
 181 /*              |                   Table       |                             */
 182 /*              |                              ---                            */
 183 /*              v                              ^^^                            */
 184 /*      +-------------------------------------|----------------+              */
 185 /*      ||||||||||||||||||||||||||||||||||||||||||||||||||||||||              */
 186 /*      +-------------------------------------|----------------+              */
 187 /*      |                                     |1......18|      |              */
 188 /*      |<------- Lempel=History ------------>|<--Ziv-->|      |              */
 189 /*      |     (=bytes already processed)      |<-Still to go-->|              */
 190 /*      |<-------------------- INPUT BLOCK ------------------->|              */
 191 /*                                                                            */
 192 /* The diagram above for LZRW3 looks almost identical to the diagram for      */
 193 /* LZRW1. The difference is that in LZRW3, the compressor transmits hash      */
 194 /* table indices instead of Lempel offsets. For this to work, the             */
 195 /* decompressor must maintain a hash table as well as the compressor and both */
 196 /* compressor and decompressor must "buffer" literals, as the decompressor    */
 197 /* cannot hash phrases commencing with a literal until another two bytes have */
 198 /* arrived.                                                                   */
 199 /*                                                                            */
 200 /*  LZRW3 Algorithm Execution Summary                                         */
 201 /*  ---------------------------------                                         */
 202 /*  1. Hash the first three bytes of the Ziv to yield a hash table index h.   */
 203 /*  2. Look up the hash table yielding history pointer p.                     */
 204 /*  3. Match where p points with the Ziv. If there is a match of three or     */
 205 /*     more bytes, code those bytes (in the Ziv) as a copy item, otherwise    */
 206 /*     code the next byte in the Ziv as a literal item.                       */
 207 /*  4. Update the hash table as possible subject to the constraint that only  */
 208 /*     phrases commencing three bytes back from the Ziv can be hashed and     */
 209 /*     entered into the hash table. (This enables the decompressor to keep    */
 210 /*     pace). See the description and code for more details.                  */
 211 /*                                                                            */
 212 /******************************************************************************/
 213 /*                                                                            */
 214 /*                     DEFINITION OF COMPRESSED FILE FORMAT                   */
 215 /*                     ====================================                   */
 216 /*  * A compressed file consists of a COPY FLAG followed by a REMAINDER.      */
 217 /*  * The copy flag CF uses up four bytes with the first byte being the       */
 218 /*    least significant.                                                      */
 219 /*  * If CF=1, then the compressed file represents the remainder of the file  */
 220 /*    exactly. Otherwise CF=0 and the remainder of the file consists of zero  */
 221 /*    or more GROUPS, each of which represents one or more bytes.             */
 222 /*  * Each group consists of two bytes of CONTROL information followed by     */
 223 /*    sixteen ITEMs except for the last group which can contain from one      */
 224 /*    to sixteen items.                                                       */
 225 /*  * An item can be either a LITERAL item or a COPY item.                    */
 226 /*  * Each item corresponds to a bit in the control bytes.                    */
 227 /*  * The first control byte corresponds to the first 8 items in the group    */
 228 /*    with bit 0 corresponding to the first item in the group and bit 7 to    */
 229 /*    the eighth item in the group.                                           */
 230 /*  * The second control byte corresponds to the second 8 items in the group  */
 231 /*    with bit 0 corresponding to the ninth item in the group and bit 7 to    */
 232 /*    the sixteenth item in the group.                                        */
 233 /*  * A zero bit in a control word means that the corresponding item is a     */
 234 /*    literal item. A one bit corresponds to a copy item.                     */
 235 /*  * A literal item consists of a single byte which represents itself.       */
 236 /*  * A copy item consists of two bytes that represent from 3 to 18 bytes.    */
 237 /*  * The first  byte in a copy item will be denoted C1.                      */
 238 /*  * The second byte in a copy item will be denoted C2.                      */
 239 /*  * Bits will be selected using square brackets.                            */
 240 /*    For example: C1[0..3] is the low nibble of the first control byte.      */
 241 /*    of copy item C1.                                                        */
 242 /*  * The LENGTH of a copy item is defined to be C1[0..3]+3 which is a number */
 243 /*    in the range [3,18].                                                    */
 244 /*  * The INDEX of a copy item is defined to be C1[4..7]*256+C2[0..8] which   */
 245 /*    is a number in the range [0,4095].                                      */
 246 /*  * A copy item represents the sequence of bytes                            */
 247 /*       text[POS-OFFSET..POS-OFFSET+LENGTH-1] where                          */
 248 /*          text   is the entire text of the uncompressed string.             */
 249 /*          POS    is the index in the text of the character following the    */
 250 /*                   string represented by all the items preceeding the item  */
 251 /*                   being defined.                                           */
 252 /*          OFFSET is obtained from INDEX by looking up the hash table.       */
 253 /*                                                                            */
 254 /******************************************************************************/
 255
 256 /* The following #define defines the length of the copy flag that appears at  */
 257 /* the start of the compressed file. The value of four bytes was chosen       */
 258 /* because the fast_copy routine on my Macintosh runs faster if the source    */
 259 /* and destination blocks are relatively longword aligned.                    */
 260 /* The actual flag data appears in the first byte. The rest are zeroed so as  */
 261 /* to normalize the compressed representation (i.e. not non-deterministic).   */
 262 #define FLAG_BYTES 4
 263
 264 /* The following #defines define the meaning of the values of the copy        */
 265 /* flag at the start of the compressed file.                                  */
 266 #define FLAG_COMPRESS 0     /* Signals that output was result of compression. */
 267 #define FLAG_COPY     1     /* Signals that output was simply copied over.    */
 268
 269 /* The 68000 microprocessor (on which this algorithm was originally developed */
 270 /* is fussy about non-aligned arrays of words. To avoid these problems the    */
 271 /* following macro can be used to "waste" from 0 to 3 bytes so as to align    */
 272 /* the argument pointer.                                                      */
 273 #define ULONG_ALIGN_UP(X) ((((ULONG)X)+sizeof(ULONG)-1)&~(sizeof(ULONG)-1))
 274
 275
 276 /* The following constant defines the maximum length of an uncompressed item. */
 277 /* This definition must not be changed; its value is hardwired into the code. */
 278 /* The longest number of bytes that can be spanned by a single item is 18     */
 279 /* for the longest copy item.                                                 */
 280 #define MAX_RAW_ITEM (18)
 281
 282 /* The following constant defines the maximum length of an uncompressed group.*/
 283 /* This definition must not be changed; its value is hardwired into the code. */
 284 /* A group contains at most 16 items which explains this definition.          */
 285 #define MAX_RAW_GROUP (16*MAX_RAW_ITEM)
 286
 287 /* The following constant defines the maximum length of a compressed group.   */
 288 /* This definition must not be changed; its value is hardwired into the code. */
 289 /* A compressed group consists of two control bytes followed by up to 16      */
 290 /* compressed items each of which can have a maximum length of two bytes.     */
 291 #define MAX_CMP_GROUP (2+16*2)
 292
 293 /* The following constant defines the number of entries in the hash table.    */
 294 /* This definition must not be changed; its value is hardwired into the code. */
 295 #define HASH_TABLE_LENGTH (4096)
 296
 297 /* LZRW3, unlike LZRW1(-A), must initialize its hash table so as to enable    */
 298 /* the compressor and decompressor to stay in step maintaining identical hash */
 299 /* tables. In an early version of the algorithm, the tables were simply       */
 300 /* initialized to zero and a check for zero was included just before the      */
 301 /* matching code. However, this test costs time. A better solution is to      */
 302 /* initialize all the entries in the hash table to point to a constant        */
 303 /* string. The decompressor does the same. This solution requires no extra    */
 304 /* test. The contents of the string do not matter so long as the string is    */
 305 /* the same for the compressor and decompressor and contains at least         */
 306 /* MAX_RAW_ITEM bytes. I chose consecutive decimal digits because they do not */
 307 /* have white space problems (e.g. there is no chance that the compiler will  */
 308 /* replace more than one space by a TAB) and because they make the length of  */
 309 /* the string obvious by inspection.                                          */
 310 #define START_STRING_18 ((UBYTE *) "123456789012345678")
 311
 312 /* In this algorithm, hash values have to be calculated at more than one      */
 313 /* point. The following macro neatens the code up for this.                   */
 314 #define HASH(PTR) \
 315    (((40543*(((*(PTR))<<8)^((*((PTR)+1))<<4)^(*((PTR)+2))))>>4) & 0xFFF)
 316
 317 /******************************************************************************/
 318
 319 /* Input  : Hand over the required amount of working memory in p_wrk_mem.     */
 320 /* Input  : Specify input block using p_src_first and src_len.                */
 321 /* Input  : Point p_dst_first to the start of the output zone (OZ).           */
 322 /* Input  : Point p_dst_len to a ULONG to receive the output length.          */
 323 /* Input  : Input block and output zone must not overlap.                     */
 324 /* Output : Length of output block written to *p_dst_len.                     */
 325 /* Output : Output block in Mem[p_dst_first..p_dst_first+*p_dst_len-1]. May   */
 326 /* Output : write in OZ=Mem[p_dst_first..p_dst_first+src_len+MAX_CMP_GROUP-1].*/
 327 /* Output : Upon completion guaranteed *p_dst_len<=src_len+FLAG_BYTES.        */
 328 LOCAL void compress_compress(UBYTE *p_wrk_mem,
 329                              UBYTE *p_src_first, ULONG  src_len,
 330                              UBYTE *p_dst_first, LONG  *p_dst_len)
 331 {
 332  /* p_src and p_dst step through the source and destination blocks.           */
 333  register UBYTE *p_src = p_src_first;
 334  register UBYTE *p_dst = p_dst_first;
 335
 336  /* The following variables are never modified and are used in the            */
 337  /* calculations that determine when the main loop terminates.                */
 338  UBYTE *p_src_post  = p_src_first+src_len;
 339  UBYTE *p_dst_post  = p_dst_first+src_len;
 340  UBYTE *p_src_max1  = p_src_first+src_len-MAX_RAW_ITEM;
 341  UBYTE *p_src_max16 = p_src_first+src_len-MAX_RAW_ITEM*16;
 342
 343  /* The variables 'p_control' and 'control' are used to buffer control bits.  */
 344  /* Before each group is processed, the next two bytes of the output block    */
 345  /* are set aside for the control word for the group about to be processed.   */
 346  /* 'p_control' is set to point to the first byte of that word. Meanwhile,    */
 347  /* 'control' buffers the control bits being generated during the processing  */
 348  /* of the group. Instead of having a counter to keep track of how many items */
 349  /* have been processed (=the number of bits in the control word), at the     */
 350  /* start of each group, the top word of 'control' is filled with 1 bits.     */
 351  /* As 'control' is shifted for each item, the 1 bits in the top word are     */
 352  /* absorbed or destroyed. When they all run out (i.e. when the top word is   */
 353  /* all zero bits, we know that we are at the end of a group.                 */
 354 # define TOPWORD 0xFFFF0000
 355  UBYTE *p_control;
 356  register ULONG control=TOPWORD;
 357
 358  /* THe variable 'hash' always points to the first element of the hash table. */
 359  UBYTE **hash= (UBYTE **)  ULONG_ALIGN_UP(p_wrk_mem);
 360
 361  /* The following two variables represent the literal buffer. p_h1 points to  */
 362  /* the hash table entry corresponding to the youngest literal. p_h2 points   */
 363  /* to the hash table entry corresponding to the second youngest literal.     */
 364  /* Note: p_h1=0=>p_h2=0 because zero values denote absence of a pending      */
 365  /* literal. The variables are initialized to zero meaning an empty "buffer". */
 366  UBYTE **p_h1=NULL;
 367  UBYTE **p_h2=NULL;
 368
 369  /* To start, we write the flag bytes. Being optimistic, we set the flag to   */
 370  /* FLAG_COMPRESS. The remaining flag bytes are zeroed so as to keep the      */
 371  /* algorithm deterministic.                                                  */
 372  *p_dst++=FLAG_COMPRESS;
 373  {UWORD i; for (i=2;i<=FLAG_BYTES;i++) *p_dst++=0;}
 374
 375  /* Reserve the first word of output as the control word for the first group. */
 376  /* Note: This is undone at the end if the input block is empty.              */
 377  p_control=p_dst; p_dst+=2;
 378
 379  /* Initialize all elements of the hash table to point to a constant string.  */
 380  /* Use of an unrolled loop speeds this up considerably.                      */
 381  {UWORD i; UBYTE **p_h=hash;
 382 #  define ZH *p_h++=START_STRING_18
 383   for (i=0;i<256;i++)     /* 256=HASH_TABLE_LENGTH/16. */
 384     {ZH;ZH;ZH;ZH;
 385      ZH;ZH;ZH;ZH;
 386      ZH;ZH;ZH;ZH;
 387      ZH;ZH;ZH;ZH;}
 388  }
 389
 390  /* The main loop processes either 1 or 16 items per iteration. As its        */
 391  /* termination logic is complicated, I have opted for an infinite loop       */
 392  /* structure containing 'break' and 'goto' statements.                       */
 393  while (TRUE)
 394    {/* Begin main processing loop. */
 395
 396     /* Note: All the variables here except unroll should be defined within    */
 397     /*       the inner loop. Unfortunately the loop hasn't got a block.       */
 398      register UBYTE *p;         /* Scans through targ phrase during matching. */
 399      register UBYTE *p_ziv= NULL ;     /* Points to first byte of current Ziv.       */
 400      register UWORD unroll;     /* Loop counter for unrolled inner loop.      */
 401      register UWORD index;      /* Index of current hash table entry.         */
 402      register UBYTE **p_h0 = NULL ;     /* Pointer to current hash table entry.       */
 403
 404     /* Test for overrun and jump to overrun code if necessary.                */
 405     if (p_dst>p_dst_post)
 406        goto overrun;
 407
 408     /* The following cascade of if statements efficiently catches and deals   */
 409     /* with varying degrees of closeness to the end of the input block.       */
 410     /* When we get very close to the end, we stop updating the table and      */
 411     /* code the remaining bytes as literals. This makes the code simpler.     */
 412     unroll=16;
 413     if (p_src>p_src_max16)
 414       {
 415        unroll=1;
 416        if (p_src>p_src_max1)
 417          {
 418           if (p_src==p_src_post)
 419              break;
 420           else
 421              goto literal;
 422          }
 423       }
 424
 425     /* This inner unrolled loop processes 'unroll' (whose value is either 1   */
 426     /* or 16) items. I have chosen to implement this loop with labels and     */
 427     /* gotos to heighten the ease with which the loop may be implemented with */
 428     /* a single decrement and branch instruction in assembly language and     */
 429     /* also because the labels act as highly readable place markers.          */
 430     /* (Also because we jump into the loop for endgame literals (see above)). */
 431
 432     begin_unrolled_loop:
 433
 434        /* To process the next phrase, we hash the next three bytes and use    */
 435        /* the resultant hash table index to look up the hash table. A pointer */
 436        /* to the entry is stored in p_h0 so as to avoid an array lookup. The  */
 437        /* hash table entry *p_h0 is looked up yielding a pointer p to a       */
 438        /* potential match of the Ziv in the history.                          */
 439        index=HASH(p_src);
 440        p_h0=&hash[index];
 441        p=*p_h0;
 442
 443        /* Having looked up the candidate position, we are in a position to    */
 444        /* attempt a match. The match loop has been unrolled using the PS      */
 445        /* macro so that failure within the first three bytes automatically    */
 446        /* results in the literal branch being taken. The coding is simple.    */
 447        /* p_ziv saves p_src so we can let p_src wander.                       */
 448 #       define PS *p++!=*p_src++
 449        p_ziv=p_src;
 450        if (PS || PS || PS)
 451          {
 452           /* Literal. */
 453
 454           /* Code the literal byte as itself and a zero control bit.          */
 455           p_src=p_ziv; literal: *p_dst++=*p_src++; control&=0xFFFEFFFF;
 456
 457           /* We have just coded a literal. If we had two pending ones, that   */
 458           /* makes three and we can update the hash table.                    */
 459           if (p_h2!=0)
 460              {*p_h2=p_ziv-2;}
 461
 462           /* In any case, rotate the hash table pointers for next time. */
 463           p_h2=p_h1; p_h1=p_h0;
 464
 465          }
 466        else
 467          {
 468           /* Copy */
 469
 470           /* Match up to 15 remaining bytes using an unrolled loop and code. */
 471 #if 0
 472           PS || PS || PS || PS || PS || PS || PS || PS ||
 473           PS || PS || PS || PS || PS || PS || PS || p_src++;
 474 #else
 475           if (
 476                !( PS || PS || PS || PS || PS || PS || PS || PS ||
 477                   PS || PS || PS || PS || PS || PS || PS )
 478              ) p_src++;
 479 #endif
 480           *p_dst++=((index&0xF00)>>4)|(--p_src-p_ziv-3);
 481           *p_dst++=index&0xFF;
 482
 483           /* As we have just coded three bytes, we are now in a position to   */
 484           /* update the hash table with the literal bytes that were pending   */
 485           /* upon the arrival of extra context bytes.                         */
 486           if (p_h1!=0)
 487             {
 488              if (p_h2)
 489                {*p_h2=p_ziv-2; p_h2=NULL;}
 490              *p_h1=p_ziv-1; p_h1=NULL;
 491             }
 492
 493           /* In any case, we can update the hash table based on the current   */
 494           /* position as we just coded at least three bytes in a copy items.  */
 495           *p_h0=p_ziv;
 496
 497          }
 498        control>>=1;
 499
 500        /* This loop is all set up for a decrement and jump instruction! */
 501 #ifndef linux
 502 `    end_unrolled_loop: if (--unroll) goto begin_unrolled_loop;
 503 #else
 504     /* end_unrolled_loop: */ if (--unroll) goto begin_unrolled_loop;
 505 #endif
 506
 507     /* At this point it will nearly always be the end of a group in which     */
 508     /* case, we have to do some control-word processing. However, near the    */
 509     /* end of the input block, the inner unrolled loop is only executed once. */
 510     /* This necessitates the 'if' test.                                       */
 511     if ((control&TOPWORD)==0)
 512       {
 513        /* Write the control word to the place we saved for it in the output. */
 514        *p_control++=  control     &0xFF;
 515        *p_control  = (control>>8) &0xFF;
 516
 517        /* Reserve the next word in the output block for the control word */
 518        /* for the group about to be processed.                           */
 519        p_control=p_dst; p_dst+=2;
 520
 521        /* Reset the control bits buffer. */
 522        control=TOPWORD;
 523       }
 524
 525    } /* End main processing loop. */
 526
 527  /* After the main processing loop has executed, all the input bytes have     */
 528  /* been processed. However, the control word has still to be written to the  */
 529  /* word reserved for it in the output at the start of the most recent group. */
 530  /* Before writing, the control word has to be shifted so that all the bits   */
 531  /* are in the right place. The "empty" bit positions are filled with 1s      */
 532  /* which partially fill the top word.                                        */
 533  while(control&TOPWORD) control>>=1;
 534  *p_control++= control     &0xFF;
 535  *p_control++=(control>>8) &0xFF;
 536
 537  /* If the last group contained no items, delete the control word too.        */
 538  if (p_control==p_dst) p_dst-=2;
 539
 540  /* Write the length of the output block to the dst_len parameter and return. */
 541  *p_dst_len=p_dst-p_dst_first;
 542  return;
 543
 544  /* Jump here as soon as an overrun is detected. An overrun is defined to     */
 545  /* have occurred if p_dst>p_dst_first+src_len. That is, the moment the       */
 546  /* length of the output written so far exceeds the length of the input block.*/
 547  /* The algorithm checks for overruns at least at the end of each group       */
 548  /* which means that the maximum overrun is MAX_CMP_GROUP bytes.              */
 549  /* Once an overrun occurs, the only thing to do is to set the copy flag and  */
 550  /* copy the input over.                                                      */
 551  overrun:
 552 #if 0
 553  *p_dst_first=FLAG_COPY;
 554  fast_copy(p_src_first,p_dst_first+FLAG_BYTES,src_len);
 555  *p_dst_len=src_len+FLAG_BYTES;
 556 #else
 557  fast_copy(p_src_first,p_dst_first,src_len);
 558  *p_dst_len= -src_len; /* return a negative number to indicate uncompressed data */
 559 #endif
 560 }
 561
 562 /******************************************************************************/
 563
 564 /* Input  : Hand over the required amount of working memory in p_wrk_mem.     */
 565 /* Input  : Specify input block using p_src_first and src_len.                */
 566 /* Input  : Point p_dst_first to the start of the output zone.                */
 567 /* Input  : Point p_dst_len to a ULONG to receive the output length.          */
 568 /* Input  : Input block and output zone must not overlap. User knows          */
 569 /* Input  : upperbound on output block length from earlier compression.       */
 570 /* Input  : In any case, maximum expansion possible is nine times.            */
 571 /* Output : Length of output block written to *p_dst_len.                     */
 572 /* Output : Output block in Mem[p_dst_first..p_dst_first+*p_dst_len-1].       */
 573 /* Output : Writes only  in Mem[p_dst_first..p_dst_first+*p_dst_len-1].       */
 574 LOCAL void compress_decompress( UBYTE *p_wrk_mem,
 575                                 UBYTE *p_src_first, LONG   src_len,
 576                                 UBYTE *p_dst_first, ULONG *p_dst_len)
 577 {
 578  /* Byte pointers p_src and p_dst scan through the input and output blocks.   */
 579  register UBYTE *p_src = p_src_first+FLAG_BYTES;
 580  register UBYTE *p_dst = p_dst_first;
 581  /* we need to avoid a SEGV when trying to uncompress corrupt data */
 582  register UBYTE *p_dst_post = p_dst_first + *p_dst_len;
 583
 584  /* The following two variables are never modified and are used to control    */
 585  /* the main loop.                                                            */
 586  UBYTE *p_src_post  = p_src_first+src_len;
 587  UBYTE *p_src_max16 = p_src_first+src_len-(MAX_CMP_GROUP-2);
 588
 589  /* The hash table is the only resident of the working memory. The hash table */
 590  /* contains HASH_TABLE_LENGTH=4096 pointers to positions in the history. To  */
 591  /* keep Macintoshes happy, it is longword aligned.                           */
 592  UBYTE **hash = (UBYTE **) ULONG_ALIGN_UP(p_wrk_mem);
 593
 594  /* The variable 'control' is used to buffer the control bits which appear in */
 595  /* groups of 16 bits (control words) at the start of each compressed group.  */
 596  /* When each group is read, bit 16 of the register is set to one. Whenever   */
 597  /* a new bit is needed, the register is shifted right. When the value of the */
 598  /* register becomes 1, we know that we have reached the end of a group.      */
 599  /* Initializing the register to 1 thus instructs the code to follow that it  */
 600  /* should read a new control word immediately.                               */
 601  register ULONG control=1;
 602
 603  /* The value of 'literals' is always in the range 0..3. It is the number of  */
 604  /* consecutive literal items just seen. We have to record this number so as  */
 605  /* to know when to update the hash table. When literals gets to 3, there     */
 606  /* have been three consecutive literals and we can update at the position of */
 607  /* the oldest of the three.                                                  */
 608  register UWORD literals=0;
 609
 610  /* Check the leading copy flag to see if the compressor chose to use a copy  */
 611  /* operation instead of a compression operation. If a copy operation was     */
 612  /* used, then all we need to do is copy the data over, set the output length */
 613  /* and return.                                                               */
 614 #if 0
 615  if (*p_src_first==FLAG_COPY)
 616    {
 617     fast_copy(p_src_first+FLAG_BYTES,p_dst_first,src_len-FLAG_BYTES);
 618     *p_dst_len=src_len-FLAG_BYTES;
 619     return;
 620    }
 621 #else
 622   if ( src_len < 0 )
 623   {
 624    fast_copy(p_src_first,p_dst_first,-src_len );
 625    *p_dst_len = (ULONG)-src_len;
 626    return;
 627   }
 628 #endif
 629
 630  /* Initialize all elements of the hash table to point to a constant string.  */
 631  /* Use of an unrolled loop speeds this up considerably.                      */
 632  {UWORD i; UBYTE **p_h=hash;
 633 #  define ZJ *p_h++=START_STRING_18
 634   for (i=0;i<256;i++)     /* 256=HASH_TABLE_LENGTH/16. */
 635     {ZJ;ZJ;ZJ;ZJ;
 636      ZJ;ZJ;ZJ;ZJ;
 637      ZJ;ZJ;ZJ;ZJ;
 638      ZJ;ZJ;ZJ;ZJ;}
 639  }
 640
 641  /* The outer loop processes either 1 or 16 items per iteration depending on  */
 642  /* how close p_src is to the end of the input block.                         */
 643  while (p_src!=p_src_post)
 644    {/* Start of outer loop */
 645
 646     register UWORD unroll;   /* Counts unrolled loop executions.              */
 647
 648     /* When 'control' has the value 1, it means that the 16 buffered control  */
 649     /* bits that were read in at the start of the current group have all been */
 650     /* shifted out and that all that is left is the 1 bit that was injected   */
 651     /* into bit 16 at the start of the current group. When we reach the end   */
 652     /* of a group, we have to load a new control word and inject a new 1 bit. */
 653     if (control==1)
 654       {
 655        control=0x10000|*p_src++;
 656        control|=(*p_src++)<<8;
 657       }
 658
 659     /* If it is possible that we are within 16 groups from the end of the     */
 660     /* input, execute the unrolled loop only once, else process a whole group */
 661     /* of 16 items by looping 16 times.                                       */
 662     unroll= p_src<=p_src_max16 ? 16 : 1;
 663
 664     /* This inner loop processes one phrase (item) per iteration. */
 665     while (unroll--)
 666       { /* Begin unrolled inner loop. */
 667
 668        /* Process a literal or copy item depending on the next control bit. */
 669        if (control&1)
 670          {
 671           /* Copy item. */
 672
 673           register UBYTE *p;           /* Points to place from which to copy. */
 674           register UWORD lenmt;        /* Length of copy item minus three.    */
 675           register UBYTE **p_hte;      /* Pointer to current hash table entry.*/
 676           register UBYTE *p_ziv=p_dst; /* Pointer to start of current Ziv.    */
 677
 678           /* Read and dismantle the copy word. Work out from where to copy.   */
 679           lenmt=*p_src++;
 680           p_hte=&hash[((lenmt&0xF0)<<4)|*p_src++];
 681           p=*p_hte;
 682           lenmt&=0xF;
 683
 684           /* Now perform the copy using a half unrolled loop. */
 685           *p_dst++=*p++;
 686           *p_dst++=*p++;
 687           *p_dst++=*p++;
 688           while (lenmt--)
 689              *p_dst++=*p++;
 690
 691           /* Because we have just received 3 or more bytes in a copy item     */
 692           /* (whose bytes we have just installed in the output), we are now   */
 693           /* in a position to flush all the pending literal hashings that had */
 694           /* been postponed for lack of bytes.                                */
 695           if (literals>0)
 696             {
 697              register UBYTE *r=p_ziv-literals;
 698              hash[HASH(r)]=r;
 699              if (literals==2)
 700                 {r++; hash[HASH(r)]=r;}
 701              literals=0;
 702             }
 703
 704           /* In any case, we can immediately update the hash table with the   */
 705           /* current position. We don't need to do a HASH(...) to work out    */
 706           /* where to put the pointer, as the compressor just told us!!!      */
 707           *p_hte=p_ziv;
 708
 709          }
 710        else
 711          {
 712           /* Literal item. */
 713
 714           /* Copy over the literal byte. */
 715           *p_dst++=*p_src++;
 716
 717           /* If we now have three literals waiting to be hashed into the hash */
 718           /* table, we can do one of them now (because there are three).      */
 719           if (++literals == 3)
 720              {register UBYTE *p=p_dst-3; hash[HASH(p)]=p; literals=2;}
 721          }
 722
 723        /* Shift the control buffer so the next control bit is in bit 0. */
 724        control>>=1;
 725 #if 1
 726        if (p_dst > p_dst_post)
 727        {
 728                /* Shit: we tried to decompress corrupt data */
 729                *p_dst_len = 0;
 730                return;
 731        }
 732 #endif
 733       } /* End unrolled inner loop. */
 734
 735    } /* End of outer loop */
 736
 737  /* Write the length of the decompressed data before returning. */
 738   *p_dst_len=p_dst-p_dst_first;
 739 }
 740
 741 /******************************************************************************/
 742 /*                               End of LZRW3.C                               */
 743 /******************************************************************************/