2 /*-------------------------------------------------------------*/
3 /*--- Compression machinery (not incl block sorting) ---*/
5 /*-------------------------------------------------------------*/
7 /* ------------------------------------------------------------------
8 This file is part of bzip2/libbzip2, a program and library for
9 lossless, block-sorting data compression.
11 bzip2/libbzip2 version 1.0.4 of 20 December 2006
12 Copyright (C) 1996-2006 Julian Seward <jseward@bzip.org>
14 Please read the WARNING, DISCLAIMER and PATENTS sections in the
17 This program is released under the terms of the license contained
19 ------------------------------------------------------------------ */
23 0.9.0 -- original version.
24 0.9.0a/b -- no changes in this file.
25 0.9.0c -- changed setting of nGroups in sendMTFValues()
26 so as to do a bit better on small files
29 #include "bzlib_private.h"
32 /*---------------------------------------------------*/
33 /*--- Bit stream I/O ---*/
34 /*---------------------------------------------------*/
36 /*---------------------------------------------------*/
37 void BZ2_bsInitWrite ( EState
* s
)
44 /*---------------------------------------------------*/
46 void bsFinishWrite ( EState
* s
)
48 while (s
->bsLive
> 0) {
49 s
->zbits
[s
->numZ
] = (UChar
)(s
->bsBuff
>> 24);
57 /*---------------------------------------------------*/
60 while (s->bsLive >= 8) { \
62 = (UChar)(s->bsBuff >> 24); \
70 /*---------------------------------------------------*/
73 void bsW ( EState
* s
, Int32 n
, UInt32 v
)
76 s
->bsBuff
|= (v
<< (32 - s
->bsLive
- n
));
81 /*---------------------------------------------------*/
83 void bsPutUInt32 ( EState
* s
, UInt32 u
)
85 bsW ( s
, 8, (u
>> 24) & 0xffL
);
86 bsW ( s
, 8, (u
>> 16) & 0xffL
);
87 bsW ( s
, 8, (u
>> 8) & 0xffL
);
88 bsW ( s
, 8, u
& 0xffL
);
92 /*---------------------------------------------------*/
94 void bsPutUChar ( EState
* s
, UChar c
)
96 bsW( s
, 8, (UInt32
)c
);
100 /*---------------------------------------------------*/
101 /*--- The back end proper ---*/
102 /*---------------------------------------------------*/
104 /*---------------------------------------------------*/
106 void makeMaps_e ( EState
* s
)
110 for (i
= 0; i
< 256; i
++)
112 s
->unseqToSeq
[i
] = s
->nInUse
;
118 /*---------------------------------------------------*/
120 void generateMTFValues ( EState
* s
)
129 After sorting (eg, here),
130 s->arr1 [ 0 .. s->nblock-1 ] holds sorted order,
132 ((UChar*)s->arr2) [ 0 .. s->nblock-1 ]
133 holds the original block data.
135 The first thing to do is generate the MTF values,
137 ((UInt16*)s->arr1) [ 0 .. s->nblock-1 ].
138 Because there are strictly fewer or equal MTF values
139 than block values, ptr values in this area are overwritten
140 with MTF values only when they are no longer needed.
142 The final compressed bitstream is generated into the
144 (UChar*) (&((UChar*)s->arr2)[s->nblock])
146 These storage aliases are set up in bzCompressInit(),
147 except for the last one, which is arranged in
150 UInt32
* ptr
= s
->ptr
;
151 UChar
* block
= s
->block
;
152 UInt16
* mtfv
= s
->mtfv
;
157 for (i
= 0; i
<= EOB
; i
++) s
->mtfFreq
[i
] = 0;
161 for (i
= 0; i
< s
->nInUse
; i
++) yy
[i
] = (UChar
) i
;
163 for (i
= 0; i
< s
->nblock
; i
++) {
165 AssertD ( wr
<= i
, "generateMTFValues(1)" );
166 j
= ptr
[i
]-1; if (j
< 0) j
+= s
->nblock
;
167 ll_i
= s
->unseqToSeq
[block
[j
]];
168 AssertD ( ll_i
< s
->nInUse
, "generateMTFValues(2a)" );
178 mtfv
[wr
] = BZ_RUNB
; wr
++;
179 s
->mtfFreq
[BZ_RUNB
]++;
181 mtfv
[wr
] = BZ_RUNA
; wr
++;
182 s
->mtfFreq
[BZ_RUNA
]++;
184 if (zPend
< 2) break;
185 zPend
= (zPend
- 2) / 2;
191 register UChar
* ryy_j
;
192 register UChar rll_i
;
197 while ( rll_i
!= rtmp
) {
198 register UChar rtmp2
;
205 j
= ryy_j
- &(yy
[0]);
206 mtfv
[wr
] = j
+1; wr
++; s
->mtfFreq
[j
+1]++;
216 mtfv
[wr
] = BZ_RUNB
; wr
++;
217 s
->mtfFreq
[BZ_RUNB
]++;
219 mtfv
[wr
] = BZ_RUNA
; wr
++;
220 s
->mtfFreq
[BZ_RUNA
]++;
222 if (zPend
< 2) break;
223 zPend
= (zPend
- 2) / 2;
228 mtfv
[wr
] = EOB
; wr
++; s
->mtfFreq
[EOB
]++;
234 /*---------------------------------------------------*/
235 #define BZ_LESSER_ICOST 0
236 #define BZ_GREATER_ICOST 15
239 void sendMTFValues ( EState
* s
)
241 Int32 v
, t
, i
, j
, gs
, ge
, totc
, bt
, bc
, iter
;
242 Int32 nSelectors
, alphaSize
, minLen
, maxLen
, selCtr
;
243 Int32 nGroups
, nBytes
;
246 UChar len [BZ_N_GROUPS][BZ_MAX_ALPHA_SIZE];
247 is a global since the decoder also needs it.
249 Int32 code[BZ_N_GROUPS][BZ_MAX_ALPHA_SIZE];
250 Int32 rfreq[BZ_N_GROUPS][BZ_MAX_ALPHA_SIZE];
251 are also globals only used in this proc.
252 Made global to keep stack frame size small.
256 UInt16 cost
[BZ_N_GROUPS
];
257 Int32 fave
[BZ_N_GROUPS
];
259 UInt16
* mtfv
= s
->mtfv
;
261 if (s
->verbosity
>= 3)
262 VPrintf3( " %d in block, %d after MTF & 1-2 coding, "
263 "%d+2 syms in use\n",
264 s
->nblock
, s
->nMTF
, s
->nInUse
);
266 alphaSize
= s
->nInUse
+2;
267 for (t
= 0; t
< BZ_N_GROUPS
; t
++)
268 for (v
= 0; v
< alphaSize
; v
++)
269 s
->len
[t
][v
] = BZ_GREATER_ICOST
;
271 /*--- Decide how many coding tables to use ---*/
272 AssertH ( s
->nMTF
> 0, 3001 );
273 if (s
->nMTF
< 200) nGroups
= 2; else
274 if (s
->nMTF
< 600) nGroups
= 3; else
275 if (s
->nMTF
< 1200) nGroups
= 4; else
276 if (s
->nMTF
< 2400) nGroups
= 5; else
279 /*--- Generate an initial set of coding tables ---*/
281 Int32 nPart
, remF
, tFreq
, aFreq
;
287 tFreq
= remF
/ nPart
;
290 while (aFreq
< tFreq
&& ge
< alphaSize
-1) {
292 aFreq
+= s
->mtfFreq
[ge
];
296 && nPart
!= nGroups
&& nPart
!= 1
297 && ((nGroups
-nPart
) % 2 == 1)) {
298 aFreq
-= s
->mtfFreq
[ge
];
302 if (s
->verbosity
>= 3)
303 VPrintf5( " initial group %d, [%d .. %d], "
304 "has %d syms (%4.1f%%)\n",
305 nPart
, gs
, ge
, aFreq
,
306 (100.0 * (float)aFreq
) / (float)(s
->nMTF
) );
308 for (v
= 0; v
< alphaSize
; v
++)
309 if (v
>= gs
&& v
<= ge
)
310 s
->len
[nPart
-1][v
] = BZ_LESSER_ICOST
; else
311 s
->len
[nPart
-1][v
] = BZ_GREATER_ICOST
;
320 Iterate up to BZ_N_ITERS times to improve the tables.
322 for (iter
= 0; iter
< BZ_N_ITERS
; iter
++) {
324 for (t
= 0; t
< nGroups
; t
++) fave
[t
] = 0;
326 for (t
= 0; t
< nGroups
; t
++)
327 for (v
= 0; v
< alphaSize
; v
++)
331 Set up an auxiliary length table which is used to fast-track
332 the common case (nGroups == 6).
335 for (v
= 0; v
< alphaSize
; v
++) {
336 s
->len_pack
[v
][0] = (s
->len
[1][v
] << 16) | s
->len
[0][v
];
337 s
->len_pack
[v
][1] = (s
->len
[3][v
] << 16) | s
->len
[2][v
];
338 s
->len_pack
[v
][2] = (s
->len
[5][v
] << 16) | s
->len
[4][v
];
347 /*--- Set group start & end marks. --*/
348 if (gs
>= s
->nMTF
) break;
349 ge
= gs
+ BZ_G_SIZE
- 1;
350 if (ge
>= s
->nMTF
) ge
= s
->nMTF
-1;
353 Calculate the cost of this group as coded
354 by each of the coding tables.
356 for (t
= 0; t
< nGroups
; t
++) cost
[t
] = 0;
358 if (nGroups
== 6 && 50 == ge
-gs
+1) {
359 /*--- fast track the common case ---*/
360 register UInt32 cost01
, cost23
, cost45
;
362 cost01
= cost23
= cost45
= 0;
364 # define BZ_ITER(nn) \
365 icv = mtfv[gs+(nn)]; \
366 cost01 += s->len_pack[icv][0]; \
367 cost23 += s->len_pack[icv][1]; \
368 cost45 += s->len_pack[icv][2]; \
370 BZ_ITER(0); BZ_ITER(1); BZ_ITER(2); BZ_ITER(3); BZ_ITER(4);
371 BZ_ITER(5); BZ_ITER(6); BZ_ITER(7); BZ_ITER(8); BZ_ITER(9);
372 BZ_ITER(10); BZ_ITER(11); BZ_ITER(12); BZ_ITER(13); BZ_ITER(14);
373 BZ_ITER(15); BZ_ITER(16); BZ_ITER(17); BZ_ITER(18); BZ_ITER(19);
374 BZ_ITER(20); BZ_ITER(21); BZ_ITER(22); BZ_ITER(23); BZ_ITER(24);
375 BZ_ITER(25); BZ_ITER(26); BZ_ITER(27); BZ_ITER(28); BZ_ITER(29);
376 BZ_ITER(30); BZ_ITER(31); BZ_ITER(32); BZ_ITER(33); BZ_ITER(34);
377 BZ_ITER(35); BZ_ITER(36); BZ_ITER(37); BZ_ITER(38); BZ_ITER(39);
378 BZ_ITER(40); BZ_ITER(41); BZ_ITER(42); BZ_ITER(43); BZ_ITER(44);
379 BZ_ITER(45); BZ_ITER(46); BZ_ITER(47); BZ_ITER(48); BZ_ITER(49);
383 cost
[0] = cost01
& 0xffff; cost
[1] = cost01
>> 16;
384 cost
[2] = cost23
& 0xffff; cost
[3] = cost23
>> 16;
385 cost
[4] = cost45
& 0xffff; cost
[5] = cost45
>> 16;
388 /*--- slow version which correctly handles all situations ---*/
389 for (i
= gs
; i
<= ge
; i
++) {
390 UInt16 icv
= mtfv
[i
];
391 for (t
= 0; t
< nGroups
; t
++) cost
[t
] += s
->len
[t
][icv
];
396 Find the coding table which is best for this group,
397 and record its identity in the selector table.
399 bc
= 999999999; bt
= -1;
400 for (t
= 0; t
< nGroups
; t
++)
401 if (cost
[t
] < bc
) { bc
= cost
[t
]; bt
= t
; };
404 s
->selector
[nSelectors
] = bt
;
408 Increment the symbol frequencies for the selected table.
410 if (nGroups
== 6 && 50 == ge
-gs
+1) {
411 /*--- fast track the common case ---*/
413 # define BZ_ITUR(nn) s->rfreq[bt][ mtfv[gs+(nn)] ]++
415 BZ_ITUR(0); BZ_ITUR(1); BZ_ITUR(2); BZ_ITUR(3); BZ_ITUR(4);
416 BZ_ITUR(5); BZ_ITUR(6); BZ_ITUR(7); BZ_ITUR(8); BZ_ITUR(9);
417 BZ_ITUR(10); BZ_ITUR(11); BZ_ITUR(12); BZ_ITUR(13); BZ_ITUR(14);
418 BZ_ITUR(15); BZ_ITUR(16); BZ_ITUR(17); BZ_ITUR(18); BZ_ITUR(19);
419 BZ_ITUR(20); BZ_ITUR(21); BZ_ITUR(22); BZ_ITUR(23); BZ_ITUR(24);
420 BZ_ITUR(25); BZ_ITUR(26); BZ_ITUR(27); BZ_ITUR(28); BZ_ITUR(29);
421 BZ_ITUR(30); BZ_ITUR(31); BZ_ITUR(32); BZ_ITUR(33); BZ_ITUR(34);
422 BZ_ITUR(35); BZ_ITUR(36); BZ_ITUR(37); BZ_ITUR(38); BZ_ITUR(39);
423 BZ_ITUR(40); BZ_ITUR(41); BZ_ITUR(42); BZ_ITUR(43); BZ_ITUR(44);
424 BZ_ITUR(45); BZ_ITUR(46); BZ_ITUR(47); BZ_ITUR(48); BZ_ITUR(49);
429 /*--- slow version which correctly handles all situations ---*/
430 for (i
= gs
; i
<= ge
; i
++)
431 s
->rfreq
[bt
][ mtfv
[i
] ]++;
436 if (s
->verbosity
>= 3) {
437 VPrintf2 ( " pass %d: size is %d, grp uses are ",
439 for (t
= 0; t
< nGroups
; t
++)
440 VPrintf1 ( "%d ", fave
[t
] );
445 Recompute the tables based on the accumulated frequencies.
447 /* maxLen was changed from 20 to 17 in bzip2-1.0.3. See
448 comment in huffman.c for details. */
449 for (t
= 0; t
< nGroups
; t
++)
450 BZ2_hbMakeCodeLengths ( &(s
->len
[t
][0]), &(s
->rfreq
[t
][0]),
451 alphaSize
, 17 /*20*/ );
455 AssertH( nGroups
< 8, 3002 );
456 AssertH( nSelectors
< 32768 &&
457 nSelectors
<= (2 + (900000 / BZ_G_SIZE
)),
461 /*--- Compute MTF values for the selectors. ---*/
463 UChar pos
[BZ_N_GROUPS
], ll_i
, tmp2
, tmp
;
464 for (i
= 0; i
< nGroups
; i
++) pos
[i
] = i
;
465 for (i
= 0; i
< nSelectors
; i
++) {
466 ll_i
= s
->selector
[i
];
469 while ( ll_i
!= tmp
) {
476 s
->selectorMtf
[i
] = j
;
480 /*--- Assign actual codes for the tables. --*/
481 for (t
= 0; t
< nGroups
; t
++) {
484 for (i
= 0; i
< alphaSize
; i
++) {
485 if (s
->len
[t
][i
] > maxLen
) maxLen
= s
->len
[t
][i
];
486 if (s
->len
[t
][i
] < minLen
) minLen
= s
->len
[t
][i
];
488 AssertH ( !(maxLen
> 17 /*20*/ ), 3004 );
489 AssertH ( !(minLen
< 1), 3005 );
490 BZ2_hbAssignCodes ( &(s
->code
[t
][0]), &(s
->len
[t
][0]),
491 minLen
, maxLen
, alphaSize
);
494 /*--- Transmit the mapping table. ---*/
497 for (i
= 0; i
< 16; i
++) {
499 for (j
= 0; j
< 16; j
++)
500 if (s
->inUse
[i
* 16 + j
]) inUse16
[i
] = True
;
504 for (i
= 0; i
< 16; i
++)
505 if (inUse16
[i
]) bsW(s
,1,1); else bsW(s
,1,0);
507 for (i
= 0; i
< 16; i
++)
509 for (j
= 0; j
< 16; j
++) {
510 if (s
->inUse
[i
* 16 + j
]) bsW(s
,1,1); else bsW(s
,1,0);
513 if (s
->verbosity
>= 3)
514 VPrintf1( " bytes: mapping %d, ", s
->numZ
-nBytes
);
517 /*--- Now the selectors. ---*/
519 bsW ( s
, 3, nGroups
);
520 bsW ( s
, 15, nSelectors
);
521 for (i
= 0; i
< nSelectors
; i
++) {
522 for (j
= 0; j
< s
->selectorMtf
[i
]; j
++) bsW(s
,1,1);
525 if (s
->verbosity
>= 3)
526 VPrintf1( "selectors %d, ", s
->numZ
-nBytes
);
528 /*--- Now the coding tables. ---*/
531 for (t
= 0; t
< nGroups
; t
++) {
532 Int32 curr
= s
->len
[t
][0];
534 for (i
= 0; i
< alphaSize
; i
++) {
535 while (curr
< s
->len
[t
][i
]) { bsW(s
,2,2); curr
++; /* 10 */ };
536 while (curr
> s
->len
[t
][i
]) { bsW(s
,2,3); curr
--; /* 11 */ };
541 if (s
->verbosity
>= 3)
542 VPrintf1 ( "code lengths %d, ", s
->numZ
-nBytes
);
544 /*--- And finally, the block data proper ---*/
549 if (gs
>= s
->nMTF
) break;
550 ge
= gs
+ BZ_G_SIZE
- 1;
551 if (ge
>= s
->nMTF
) ge
= s
->nMTF
-1;
552 AssertH ( s
->selector
[selCtr
] < nGroups
, 3006 );
554 if (nGroups
== 6 && 50 == ge
-gs
+1) {
555 /*--- fast track the common case ---*/
557 UChar
* s_len_sel_selCtr
558 = &(s
->len
[s
->selector
[selCtr
]][0]);
559 Int32
* s_code_sel_selCtr
560 = &(s
->code
[s
->selector
[selCtr
]][0]);
562 # define BZ_ITAH(nn) \
563 mtfv_i = mtfv[gs+(nn)]; \
565 s_len_sel_selCtr[mtfv_i], \
566 s_code_sel_selCtr[mtfv_i] )
568 BZ_ITAH(0); BZ_ITAH(1); BZ_ITAH(2); BZ_ITAH(3); BZ_ITAH(4);
569 BZ_ITAH(5); BZ_ITAH(6); BZ_ITAH(7); BZ_ITAH(8); BZ_ITAH(9);
570 BZ_ITAH(10); BZ_ITAH(11); BZ_ITAH(12); BZ_ITAH(13); BZ_ITAH(14);
571 BZ_ITAH(15); BZ_ITAH(16); BZ_ITAH(17); BZ_ITAH(18); BZ_ITAH(19);
572 BZ_ITAH(20); BZ_ITAH(21); BZ_ITAH(22); BZ_ITAH(23); BZ_ITAH(24);
573 BZ_ITAH(25); BZ_ITAH(26); BZ_ITAH(27); BZ_ITAH(28); BZ_ITAH(29);
574 BZ_ITAH(30); BZ_ITAH(31); BZ_ITAH(32); BZ_ITAH(33); BZ_ITAH(34);
575 BZ_ITAH(35); BZ_ITAH(36); BZ_ITAH(37); BZ_ITAH(38); BZ_ITAH(39);
576 BZ_ITAH(40); BZ_ITAH(41); BZ_ITAH(42); BZ_ITAH(43); BZ_ITAH(44);
577 BZ_ITAH(45); BZ_ITAH(46); BZ_ITAH(47); BZ_ITAH(48); BZ_ITAH(49);
582 /*--- slow version which correctly handles all situations ---*/
583 for (i
= gs
; i
<= ge
; i
++) {
585 s
->len
[s
->selector
[selCtr
]] [mtfv
[i
]],
586 s
->code
[s
->selector
[selCtr
]] [mtfv
[i
]] );
594 AssertH( selCtr
== nSelectors
, 3007 );
596 if (s
->verbosity
>= 3)
597 VPrintf1( "codes %d\n", s
->numZ
-nBytes
);
601 /*---------------------------------------------------*/
602 void BZ2_compressBlock ( EState
* s
, Bool is_last_block
)
606 BZ_FINALISE_CRC ( s
->blockCRC
);
607 s
->combinedCRC
= (s
->combinedCRC
<< 1) | (s
->combinedCRC
>> 31);
608 s
->combinedCRC
^= s
->blockCRC
;
609 if (s
->blockNo
> 1) s
->numZ
= 0;
611 if (s
->verbosity
>= 2)
612 VPrintf4( " block %d: crc = 0x%08x, "
613 "combined CRC = 0x%08x, size = %d\n",
614 s
->blockNo
, s
->blockCRC
, s
->combinedCRC
, s
->nblock
);
619 s
->zbits
= (UChar
*) (&((UChar
*)s
->arr2
)[s
->nblock
]);
621 /*-- If this is the first block, create the stream header. --*/
622 if (s
->blockNo
== 1) {
623 BZ2_bsInitWrite ( s
);
624 bsPutUChar ( s
, BZ_HDR_B
);
625 bsPutUChar ( s
, BZ_HDR_Z
);
626 bsPutUChar ( s
, BZ_HDR_h
);
627 bsPutUChar ( s
, (UChar
)(BZ_HDR_0
+ s
->blockSize100k
) );
632 bsPutUChar ( s
, 0x31 ); bsPutUChar ( s
, 0x41 );
633 bsPutUChar ( s
, 0x59 ); bsPutUChar ( s
, 0x26 );
634 bsPutUChar ( s
, 0x53 ); bsPutUChar ( s
, 0x59 );
636 /*-- Now the block's CRC, so it is in a known place. --*/
637 bsPutUInt32 ( s
, s
->blockCRC
);
640 Now a single bit indicating (non-)randomisation.
641 As of version 0.9.5, we use a better sorting algorithm
642 which makes randomisation unnecessary. So always set
643 the randomised bit to 'no'. Of course, the decoder
644 still needs to be able to handle randomised blocks
645 so as to maintain backwards compatibility with
646 older versions of bzip2.
650 bsW ( s
, 24, s
->origPtr
);
651 generateMTFValues ( s
);
656 /*-- If this is the last block, add the stream trailer. --*/
659 bsPutUChar ( s
, 0x17 ); bsPutUChar ( s
, 0x72 );
660 bsPutUChar ( s
, 0x45 ); bsPutUChar ( s
, 0x38 );
661 bsPutUChar ( s
, 0x50 ); bsPutUChar ( s
, 0x90 );
662 bsPutUInt32 ( s
, s
->combinedCRC
);
663 if (s
->verbosity
>= 2)
664 VPrintf1( " final combined CRC = 0x%08x\n ", s
->combinedCRC
);
670 /*-------------------------------------------------------------*/
671 /*--- end compress.c ---*/
672 /*-------------------------------------------------------------*/