Coverage Report

Created: 2026-02-26 07:00

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
/src/bzip2/blocksort.c
Line
Count
Source
1
2
/*-------------------------------------------------------------*/
3
/*--- Block sorting machinery                               ---*/
4
/*---                                           blocksort.c ---*/
5
/*-------------------------------------------------------------*/
6
7
/* ------------------------------------------------------------------
8
   This file is part of bzip2/libbzip2, a program and library for
9
   lossless, block-sorting data compression.
10
11
   bzip2/libbzip2 version 1.0.8 of 13 July 2019
12
   Copyright (C) 1996-2019 Julian Seward <jseward@acm.org>
13
14
   Please read the WARNING, DISCLAIMER and PATENTS sections in the 
15
   README file.
16
17
   This program is released under the terms of the license contained
18
   in the file LICENSE.
19
   ------------------------------------------------------------------ */
20
21
22
#include "bzlib_private.h"
23
24
/*---------------------------------------------*/
25
/*--- Fallback O(N log(N)^2) sorting        ---*/
26
/*--- algorithm, for repetitive blocks      ---*/
27
/*---------------------------------------------*/
28
29
/*---------------------------------------------*/
30
static 
31
__inline__
32
void fallbackSimpleSort ( UInt32* fmap, 
33
                          UInt32* eclass, 
34
                          Int32   lo, 
35
                          Int32   hi )
36
204M
{
37
204M
   Int32 i, j, tmp;
38
204M
   UInt32 ec_tmp;
39
40
204M
   if (lo == hi) return;
41
42
200M
   if (hi - lo > 3) {
43
126M
      for ( i = hi-4; i >= lo; i-- ) {
44
94.5M
         tmp = fmap[i];
45
94.5M
         ec_tmp = eclass[tmp];
46
133M
         for ( j = i+4; j <= hi && ec_tmp > eclass[fmap[j]]; j += 4 )
47
38.8M
            fmap[j-4] = fmap[j];
48
94.5M
         fmap[j-4] = tmp;
49
94.5M
      }
50
32.2M
   }
51
52
610M
   for ( i = hi-1; i >= lo; i-- ) {
53
409M
      tmp = fmap[i];
54
409M
      ec_tmp = eclass[tmp];
55
599M
      for ( j = i+1; j <= hi && ec_tmp > eclass[fmap[j]]; j++ )
56
190M
         fmap[j-1] = fmap[j];
57
409M
      fmap[j-1] = tmp;
58
409M
   }
59
200M
}
60
61
62
/*---------------------------------------------*/
63
#define fswap(zz1, zz2) \
64
3.87G
   { Int32 zztmp = zz1; zz1 = zz2; zz2 = zztmp; }
65
66
76.2M
#define fvswap(zzp1, zzp2, zzn)       \
67
76.2M
{                                     \
68
76.2M
   Int32 yyp1 = (zzp1);               \
69
76.2M
   Int32 yyp2 = (zzp2);               \
70
76.2M
   Int32 yyn  = (zzn);                \
71
727M
   while (yyn > 0) {                  \
72
651M
      fswap(fmap[yyp1], fmap[yyp2]);  \
73
651M
      yyp1++; yyp2++; yyn--;          \
74
651M
   }                                  \
75
76.2M
}
76
77
78
76.2M
#define fmin(a,b) ((a) < (b)) ? (a) : (b)
79
80
243M
#define fpush(lz,hz) { stackLo[sp] = lz; \
81
243M
                       stackHi[sp] = hz; \
82
243M
                       sp++; }
83
84
243M
#define fpop(lz,hz) { sp--;              \
85
243M
                      lz = stackLo[sp];  \
86
243M
                      hz = stackHi[sp]; }
87
88
243M
#define FALLBACK_QSORT_SMALL_THRESH 10
89
#define FALLBACK_QSORT_STACK_SIZE   100
90
91
92
static
93
void fallbackQSort3 ( UInt32* fmap, 
94
                      UInt32* eclass,
95
                      Int32   loSt, 
96
                      Int32   hiSt )
97
167M
{
98
167M
   Int32 unLo, unHi, ltLo, gtHi, n, m;
99
167M
   Int32 sp, lo, hi;
100
167M
   UInt32 med, r, r3;
101
167M
   Int32 stackLo[FALLBACK_QSORT_STACK_SIZE];
102
167M
   Int32 stackHi[FALLBACK_QSORT_STACK_SIZE];
103
104
167M
   r = 0;
105
106
167M
   sp = 0;
107
167M
   fpush ( loSt, hiSt );
108
109
410M
   while (sp > 0) {
110
111
243M
      AssertH ( sp < FALLBACK_QSORT_STACK_SIZE - 1, 1004 );
112
113
243M
      fpop ( lo, hi );
114
243M
      if (hi - lo < FALLBACK_QSORT_SMALL_THRESH) {
115
204M
         fallbackSimpleSort ( fmap, eclass, lo, hi );
116
204M
         continue;
117
204M
      }
118
119
      /* Random partitioning.  Median of 3 sometimes fails to
120
         avoid bad cases.  Median of 9 seems to help but 
121
         looks rather expensive.  This too seems to work but
122
         is cheaper.  Guidance for the magic constants 
123
         7621 and 32768 is taken from Sedgewick's algorithms
124
         book, chapter 35.
125
      */
126
38.6M
      r = ((r * 7621) + 1) % 32768;
127
38.6M
      r3 = r % 3;
128
38.6M
      if (r3 == 0) med = eclass[fmap[lo]]; else
129
31.9M
      if (r3 == 1) med = eclass[fmap[(lo+hi)>>1]]; else
130
12.9M
                   med = eclass[fmap[hi]];
131
132
38.6M
      unLo = ltLo = lo;
133
38.6M
      unHi = gtHi = hi;
134
135
403M
      while (1) {
136
2.87G
         while (1) {
137
2.87G
            if (unLo > unHi) break;
138
2.85G
            n = (Int32)eclass[fmap[unLo]] - (Int32)med;
139
2.85G
            if (n == 0) { 
140
1.32G
               fswap(fmap[unLo], fmap[ltLo]); 
141
1.32G
               ltLo++; unLo++; 
142
1.32G
               continue; 
143
1.52G
            };
144
1.52G
            if (n > 0) break;
145
1.14G
            unLo++;
146
1.14G
         }
147
3.06G
         while (1) {
148
3.06G
            if (unLo > unHi) break;
149
3.02G
            n = (Int32)eclass[fmap[unHi]] - (Int32)med;
150
3.02G
            if (n == 0) { 
151
1.53G
               fswap(fmap[unHi], fmap[gtHi]); 
152
1.53G
               gtHi--; unHi--; 
153
1.53G
               continue; 
154
1.53G
            };
155
1.49G
            if (n < 0) break;
156
1.12G
            unHi--;
157
1.12G
         }
158
403M
         if (unLo > unHi) break;
159
364M
         fswap(fmap[unLo], fmap[unHi]); unLo++; unHi--;
160
364M
      }
161
162
38.6M
      AssertD ( unHi == unLo-1, "fallbackQSort3(2)" );
163
164
38.6M
      if (gtHi < ltLo) continue;
165
166
38.1M
      n = fmin(ltLo-lo, unLo-ltLo); fvswap(lo, unLo-n, n);
167
38.1M
      m = fmin(hi-gtHi, gtHi-unHi); fvswap(unLo, hi-m+1, m);
168
169
38.1M
      n = lo + unLo - ltLo - 1;
170
38.1M
      m = hi - (gtHi - unHi) + 1;
171
172
38.1M
      if (n - lo > hi - m) {
173
17.5M
         fpush ( lo, n );
174
17.5M
         fpush ( m, hi );
175
20.6M
      } else {
176
20.6M
         fpush ( m, hi );
177
20.6M
         fpush ( lo, n );
178
20.6M
      }
179
38.1M
   }
180
167M
}
181
182
#undef fmin
183
#undef fpush
184
#undef fpop
185
#undef fswap
186
#undef fvswap
187
#undef FALLBACK_QSORT_SMALL_THRESH
188
#undef FALLBACK_QSORT_STACK_SIZE
189
190
191
/*---------------------------------------------*/
192
/* Pre:
193
      nblock > 0
194
      eclass exists for [0 .. nblock-1]
195
      ((UChar*)eclass) [0 .. nblock-1] holds block
196
      ptr exists for [0 .. nblock-1]
197
198
   Post:
199
      ((UChar*)eclass) [0 .. nblock-1] holds block
200
      All other areas of eclass destroyed
201
      fmap [0 .. nblock-1] holds sorted order
202
      bhtab [ 0 .. 2+(nblock/32) ] destroyed
203
*/
204
205
426M
#define       SET_BH(zz)  bhtab[(zz) >> 5] |= ((UInt32)1 << ((zz) & 31))
206
102k
#define     CLEAR_BH(zz)  bhtab[(zz) >> 5] &= ~((UInt32)1 << ((zz) & 31))
207
7.55G
#define     ISSET_BH(zz)  (bhtab[(zz) >> 5] & ((UInt32)1 << ((zz) & 31)))
208
142M
#define      WORD_BH(zz)  bhtab[(zz) >> 5]
209
907M
#define UNALIGNED_BH(zz)  ((zz) & 0x01f)
210
211
static
212
void fallbackSort ( UInt32* fmap, 
213
                    UInt32* eclass, 
214
                    UInt32* bhtab,
215
                    Int32   nblock,
216
                    Int32   verb )
217
3.19k
{
218
3.19k
   Int32 ftab[257];
219
3.19k
   Int32 ftabCopy[256];
220
3.19k
   Int32 H, i, j, k, l, r, cc, cc1;
221
3.19k
   Int32 nNotDone;
222
3.19k
   Int32 nBhtab;
223
3.19k
   UChar* eclass8 = (UChar*)eclass;
224
225
   /*--
226
      Initial 1-char radix sort to generate
227
      initial fmap and initial BH bits.
228
   --*/
229
3.19k
   if (verb >= 4)
230
0
      VPrintf0 ( "        bucket sorting ...\n" );
231
823k
   for (i = 0; i < 257;    i++) ftab[i] = 0;
232
260M
   for (i = 0; i < nblock; i++) ftab[eclass8[i]]++;
233
820k
   for (i = 0; i < 256;    i++) ftabCopy[i] = ftab[i];
234
820k
   for (i = 1; i < 257;    i++) ftab[i] += ftab[i-1];
235
236
260M
   for (i = 0; i < nblock; i++) {
237
260M
      j = eclass8[i];
238
260M
      k = ftab[j] - 1;
239
260M
      ftab[j] = k;
240
260M
      fmap[k] = i;
241
260M
   }
242
243
3.19k
   nBhtab = 2 + (nblock / 32);
244
8.13M
   for (i = 0; i < nBhtab; i++) bhtab[i] = 0;
245
820k
   for (i = 0; i < 256; i++) SET_BH(ftab[i]);
246
247
   /*--
248
      Inductively refine the buckets.  Kind-of an
249
      "exponential radix sort" (!), inspired by the
250
      Manber-Myers suffix array construction algorithm.
251
   --*/
252
253
   /*-- set sentinel bits for block-end detection --*/
254
105k
   for (i = 0; i < 32; i++) { 
255
102k
      SET_BH(nblock + 2*i);
256
102k
      CLEAR_BH(nblock + 2*i + 1);
257
102k
   }
258
259
   /*-- the log(N) loop --*/
260
3.19k
   H = 1;
261
27.6k
   while (1) {
262
263
27.6k
      if (verb >= 4) 
264
0
         VPrintf1 ( "        depth %6d has ", H );
265
266
27.6k
      j = 0;
267
4.55G
      for (i = 0; i < nblock; i++) {
268
4.55G
         if (ISSET_BH(i)) j = i;
269
4.55G
         k = fmap[i] - H; if (k < 0) k += nblock;
270
4.55G
         eclass[k] = j;
271
4.55G
      }
272
273
27.6k
      nNotDone = 0;
274
27.6k
      r = -1;
275
167M
      while (1) {
276
277
   /*-- find the next non-singleton bucket --*/
278
167M
         k = r + 1;
279
719M
         while (ISSET_BH(k) && UNALIGNED_BH(k)) k++;
280
167M
         if (ISSET_BH(k)) {
281
39.1M
            while (WORD_BH(k) == 0xffffffff) k += 32;
282
160M
            while (ISSET_BH(k)) k++;
283
21.7M
         }
284
167M
         l = k - 1;
285
167M
         if (l >= nblock) break;
286
488M
         while (!ISSET_BH(k) && UNALIGNED_BH(k)) k++;
287
167M
         if (!ISSET_BH(k)) {
288
103M
            while (WORD_BH(k) == 0x00000000) k += 32;
289
90.4M
            while (!ISSET_BH(k)) k++;
290
12.5M
         }
291
167M
         r = k - 1;
292
167M
         if (r >= nblock) break;
293
294
         /*-- now [l, r] bracket current bucket --*/
295
167M
         if (r > l) {
296
167M
            nNotDone += (r - l + 1);
297
167M
            fallbackQSort3 ( fmap, eclass, l, r );
298
299
            /*-- scan bucket and generate header bits-- */
300
167M
            cc = -1;
301
3.63G
            for (i = l; i <= r; i++) {
302
3.46G
               cc1 = eclass[fmap[i]];
303
3.46G
               if (cc != cc1) { SET_BH(i); cc = cc1; };
304
3.46G
            }
305
167M
         }
306
167M
      }
307
308
27.6k
      if (verb >= 4) 
309
0
         VPrintf1 ( "%6d unresolved strings\n", nNotDone );
310
311
27.6k
      H *= 2;
312
27.6k
      if (H > nblock || nNotDone == 0) break;
313
27.6k
   }
314
315
   /*-- 
316
      Reconstruct the original block in
317
      eclass8 [0 .. nblock-1], since the
318
      previous phase destroyed it.
319
   --*/
320
3.19k
   if (verb >= 4)
321
0
      VPrintf0 ( "        reconstructing block ...\n" );
322
3.19k
   j = 0;
323
260M
   for (i = 0; i < nblock; i++) {
324
260M
      while (ftabCopy[j] == 0) j++;
325
260M
      ftabCopy[j]--;
326
260M
      eclass8[fmap[i]] = (UChar)j;
327
260M
   }
328
3.19k
   AssertH ( j < 256, 1005 );
329
3.19k
}
330
331
#undef       SET_BH
332
#undef     CLEAR_BH
333
#undef     ISSET_BH
334
#undef      WORD_BH
335
#undef UNALIGNED_BH
336
337
338
/*---------------------------------------------*/
339
/*--- The main, O(N^2 log(N)) sorting       ---*/
340
/*--- algorithm.  Faster for "normal"       ---*/
341
/*--- non-repetitive blocks.                ---*/
342
/*---------------------------------------------*/
343
344
/*---------------------------------------------*/
345
static
346
__inline__
347
Bool mainGtU ( UInt32  i1, 
348
               UInt32  i2,
349
               UChar*  block, 
350
               UInt16* quadrant,
351
               UInt32  nblock,
352
               Int32*  budget )
353
317M
{
354
317M
   Int32  k;
355
317M
   UChar  c1, c2;
356
317M
   UInt16 s1, s2;
357
358
317M
   AssertD ( i1 != i2, "mainGtU" );
359
   /* 1 */
360
317M
   c1 = block[i1]; c2 = block[i2];
361
317M
   if (c1 != c2) return (c1 > c2);
362
142M
   i1++; i2++;
363
   /* 2 */
364
142M
   c1 = block[i1]; c2 = block[i2];
365
142M
   if (c1 != c2) return (c1 > c2);
366
126M
   i1++; i2++;
367
   /* 3 */
368
126M
   c1 = block[i1]; c2 = block[i2];
369
126M
   if (c1 != c2) return (c1 > c2);
370
120M
   i1++; i2++;
371
   /* 4 */
372
120M
   c1 = block[i1]; c2 = block[i2];
373
120M
   if (c1 != c2) return (c1 > c2);
374
114M
   i1++; i2++;
375
   /* 5 */
376
114M
   c1 = block[i1]; c2 = block[i2];
377
114M
   if (c1 != c2) return (c1 > c2);
378
109M
   i1++; i2++;
379
   /* 6 */
380
109M
   c1 = block[i1]; c2 = block[i2];
381
109M
   if (c1 != c2) return (c1 > c2);
382
104M
   i1++; i2++;
383
   /* 7 */
384
104M
   c1 = block[i1]; c2 = block[i2];
385
104M
   if (c1 != c2) return (c1 > c2);
386
100M
   i1++; i2++;
387
   /* 8 */
388
100M
   c1 = block[i1]; c2 = block[i2];
389
100M
   if (c1 != c2) return (c1 > c2);
390
97.6M
   i1++; i2++;
391
   /* 9 */
392
97.6M
   c1 = block[i1]; c2 = block[i2];
393
97.6M
   if (c1 != c2) return (c1 > c2);
394
95.6M
   i1++; i2++;
395
   /* 10 */
396
95.6M
   c1 = block[i1]; c2 = block[i2];
397
95.6M
   if (c1 != c2) return (c1 > c2);
398
94.2M
   i1++; i2++;
399
   /* 11 */
400
94.2M
   c1 = block[i1]; c2 = block[i2];
401
94.2M
   if (c1 != c2) return (c1 > c2);
402
93.1M
   i1++; i2++;
403
   /* 12 */
404
93.1M
   c1 = block[i1]; c2 = block[i2];
405
93.1M
   if (c1 != c2) return (c1 > c2);
406
91.8M
   i1++; i2++;
407
408
91.8M
   k = nblock + 8;
409
410
2.60G
   do {
411
      /* 1 */
412
2.60G
      c1 = block[i1]; c2 = block[i2];
413
2.60G
      if (c1 != c2) return (c1 > c2);
414
2.60G
      s1 = quadrant[i1]; s2 = quadrant[i2];
415
2.60G
      if (s1 != s2) return (s1 > s2);
416
2.58G
      i1++; i2++;
417
      /* 2 */
418
2.58G
      c1 = block[i1]; c2 = block[i2];
419
2.58G
      if (c1 != c2) return (c1 > c2);
420
2.58G
      s1 = quadrant[i1]; s2 = quadrant[i2];
421
2.58G
      if (s1 != s2) return (s1 > s2);
422
2.56G
      i1++; i2++;
423
      /* 3 */
424
2.56G
      c1 = block[i1]; c2 = block[i2];
425
2.56G
      if (c1 != c2) return (c1 > c2);
426
2.56G
      s1 = quadrant[i1]; s2 = quadrant[i2];
427
2.56G
      if (s1 != s2) return (s1 > s2);
428
2.55G
      i1++; i2++;
429
      /* 4 */
430
2.55G
      c1 = block[i1]; c2 = block[i2];
431
2.55G
      if (c1 != c2) return (c1 > c2);
432
2.54G
      s1 = quadrant[i1]; s2 = quadrant[i2];
433
2.54G
      if (s1 != s2) return (s1 > s2);
434
2.54G
      i1++; i2++;
435
      /* 5 */
436
2.54G
      c1 = block[i1]; c2 = block[i2];
437
2.54G
      if (c1 != c2) return (c1 > c2);
438
2.54G
      s1 = quadrant[i1]; s2 = quadrant[i2];
439
2.54G
      if (s1 != s2) return (s1 > s2);
440
2.53G
      i1++; i2++;
441
      /* 6 */
442
2.53G
      c1 = block[i1]; c2 = block[i2];
443
2.53G
      if (c1 != c2) return (c1 > c2);
444
2.53G
      s1 = quadrant[i1]; s2 = quadrant[i2];
445
2.53G
      if (s1 != s2) return (s1 > s2);
446
2.52G
      i1++; i2++;
447
      /* 7 */
448
2.52G
      c1 = block[i1]; c2 = block[i2];
449
2.52G
      if (c1 != c2) return (c1 > c2);
450
2.52G
      s1 = quadrant[i1]; s2 = quadrant[i2];
451
2.52G
      if (s1 != s2) return (s1 > s2);
452
2.52G
      i1++; i2++;
453
      /* 8 */
454
2.52G
      c1 = block[i1]; c2 = block[i2];
455
2.52G
      if (c1 != c2) return (c1 > c2);
456
2.52G
      s1 = quadrant[i1]; s2 = quadrant[i2];
457
2.52G
      if (s1 != s2) return (s1 > s2);
458
2.51G
      i1++; i2++;
459
460
2.51G
      if (i1 >= nblock) i1 -= nblock;
461
2.51G
      if (i2 >= nblock) i2 -= nblock;
462
463
2.51G
      k -= 8;
464
2.51G
      (*budget)--;
465
2.51G
   }
466
2.51G
      while (k >= 0);
467
468
550
   return False;
469
91.8M
}
470
471
472
/*---------------------------------------------*/
473
/*--
474
   Knuth's increments seem to work better
475
   than Incerpi-Sedgewick here.  Possibly
476
   because the number of elems to sort is
477
   usually small, typically <= 20.
478
--*/
479
static
480
Int32 incs[14] = { 1, 4, 13, 40, 121, 364, 1093, 3280,
481
                   9841, 29524, 88573, 265720,
482
                   797161, 2391484 };
483
484
static
485
void mainSimpleSort ( UInt32* ptr,
486
                      UChar*  block,
487
                      UInt16* quadrant,
488
                      Int32   nblock,
489
                      Int32   lo, 
490
                      Int32   hi, 
491
                      Int32   d,
492
                      Int32*  budget )
493
11.0M
{
494
11.0M
   Int32 i, j, h, bigN, hp;
495
11.0M
   UInt32 v;
496
497
11.0M
   bigN = hi - lo + 1;
498
11.0M
   if (bigN < 2) return;
499
500
10.5M
   hp = 0;
501
28.9M
   while (incs[hp] < bigN) hp++;
502
10.5M
   hp--;
503
504
28.9M
   for (; hp >= 0; hp--) {
505
18.4M
      h = incs[hp];
506
507
18.4M
      i = lo + h;
508
72.3M
      while (True) {
509
510
         /*-- copy 1 --*/
511
72.3M
         if (i > hi) break;
512
67.1M
         v = ptr[i];
513
67.1M
         j = i;
514
109M
         while ( mainGtU ( 
515
109M
                    ptr[j-h]+d, v+d, block, quadrant, nblock, budget 
516
109M
                 ) ) {
517
55.7M
            ptr[j] = ptr[j-h];
518
55.7M
            j = j - h;
519
55.7M
            if (j <= (lo + h - 1)) break;
520
55.7M
         }
521
67.1M
         ptr[j] = v;
522
67.1M
         i++;
523
524
         /*-- copy 2 --*/
525
67.1M
         if (i > hi) break;
526
59.6M
         v = ptr[i];
527
59.6M
         j = i;
528
105M
         while ( mainGtU ( 
529
105M
                    ptr[j-h]+d, v+d, block, quadrant, nblock, budget 
530
105M
                 ) ) {
531
55.7M
            ptr[j] = ptr[j-h];
532
55.7M
            j = j - h;
533
55.7M
            if (j <= (lo + h - 1)) break;
534
55.7M
         }
535
59.6M
         ptr[j] = v;
536
59.6M
         i++;
537
538
         /*-- copy 3 --*/
539
59.6M
         if (i > hi) break;
540
53.9M
         v = ptr[i];
541
53.9M
         j = i;
542
102M
         while ( mainGtU ( 
543
102M
                    ptr[j-h]+d, v+d, block, quadrant, nblock, budget 
544
102M
                 ) ) {
545
55.5M
            ptr[j] = ptr[j-h];
546
55.5M
            j = j - h;
547
55.5M
            if (j <= (lo + h - 1)) break;
548
55.5M
         }
549
53.9M
         ptr[j] = v;
550
53.9M
         i++;
551
552
53.9M
         if (*budget < 0) return;
553
53.9M
      }
554
18.4M
   }
555
10.5M
}
556
557
558
/*---------------------------------------------*/
559
/*--
560
   The following is an implementation of
561
   an elegant 3-way quicksort for strings,
562
   described in a paper "Fast Algorithms for
563
   Sorting and Searching Strings", by Robert
564
   Sedgewick and Jon L. Bentley.
565
--*/
566
567
#define mswap(zz1, zz2) \
568
412M
   { Int32 zztmp = zz1; zz1 = zz2; zz2 = zztmp; }
569
570
2.20M
#define mvswap(zzp1, zzp2, zzn)       \
571
2.20M
{                                     \
572
2.20M
   Int32 yyp1 = (zzp1);               \
573
2.20M
   Int32 yyp2 = (zzp2);               \
574
2.20M
   Int32 yyn  = (zzn);                \
575
20.1M
   while (yyn > 0) {                  \
576
17.9M
      mswap(ptr[yyp1], ptr[yyp2]);    \
577
17.9M
      yyp1++; yyp2++; yyn--;          \
578
17.9M
   }                                  \
579
2.20M
}
580
581
static 
582
__inline__
583
UChar mmed3 ( UChar a, UChar b, UChar c )
584
1.53M
{
585
1.53M
   UChar t;
586
1.53M
   if (a > b) { t = a; a = b; b = t; };
587
1.53M
   if (b > c) { 
588
629k
      b = c;
589
629k
      if (a > b) b = a;
590
629k
   }
591
1.53M
   return b;
592
1.53M
}
593
594
2.20M
#define mmin(a,b) ((a) < (b)) ? (a) : (b)
595
596
12.5M
#define mpush(lz,hz,dz) { stackLo[sp] = lz; \
597
12.5M
                          stackHi[sp] = hz; \
598
12.5M
                          stackD [sp] = dz; \
599
12.5M
                          sp++; }
600
601
12.5M
#define mpop(lz,hz,dz) { sp--;             \
602
12.5M
                         lz = stackLo[sp]; \
603
12.5M
                         hz = stackHi[sp]; \
604
12.5M
                         dz = stackD [sp]; }
605
606
607
6.62M
#define mnextsize(az) (nextHi[az]-nextLo[az])
608
609
#define mnextswap(az,bz)                                        \
610
1.04M
   { Int32 tz;                                                  \
611
1.04M
     tz = nextLo[az]; nextLo[az] = nextLo[bz]; nextLo[bz] = tz; \
612
1.04M
     tz = nextHi[az]; nextHi[az] = nextHi[bz]; nextHi[bz] = tz; \
613
1.04M
     tz = nextD [az]; nextD [az] = nextD [bz]; nextD [bz] = tz; }
614
615
616
25.1M
#define MAIN_QSORT_SMALL_THRESH 20
617
1.57M
#define MAIN_QSORT_DEPTH_THRESH (BZ_N_RADIX + BZ_N_QSORT)
618
#define MAIN_QSORT_STACK_SIZE 100
619
620
static
621
void mainQSort3 ( UInt32* ptr,
622
                  UChar*  block,
623
                  UInt16* quadrant,
624
                  Int32   nblock,
625
                  Int32   loSt, 
626
                  Int32   hiSt, 
627
                  Int32   dSt,
628
                  Int32*  budget )
629
8.81M
{
630
8.81M
   Int32 unLo, unHi, ltLo, gtHi, n, m, med;
631
8.81M
   Int32 sp, lo, hi, d;
632
633
8.81M
   Int32 stackLo[MAIN_QSORT_STACK_SIZE];
634
8.81M
   Int32 stackHi[MAIN_QSORT_STACK_SIZE];
635
8.81M
   Int32 stackD [MAIN_QSORT_STACK_SIZE];
636
637
8.81M
   Int32 nextLo[3];
638
8.81M
   Int32 nextHi[3];
639
8.81M
   Int32 nextD [3];
640
641
8.81M
   sp = 0;
642
8.81M
   mpush ( loSt, hiSt, dSt );
643
644
21.3M
   while (sp > 0) {
645
646
12.5M
      AssertH ( sp < MAIN_QSORT_STACK_SIZE - 2, 1001 );
647
648
12.5M
      mpop ( lo, hi, d );
649
12.5M
      if (hi - lo < MAIN_QSORT_SMALL_THRESH || 
650
11.0M
          d > MAIN_QSORT_DEPTH_THRESH) {
651
11.0M
         mainSimpleSort ( ptr, block, quadrant, nblock, lo, hi, d, budget );
652
11.0M
         if (*budget < 0) return;
653
11.0M
         continue;
654
11.0M
      }
655
656
1.53M
      med = (Int32) 
657
1.53M
            mmed3 ( block[ptr[ lo         ]+d],
658
1.53M
                    block[ptr[ hi         ]+d],
659
1.53M
                    block[ptr[ (lo+hi)>>1 ]+d] );
660
661
1.53M
      unLo = ltLo = lo;
662
1.53M
      unHi = gtHi = hi;
663
664
12.4M
      while (True) {
665
282M
         while (True) {
666
282M
            if (unLo > unHi) break;
667
281M
            n = ((Int32)block[ptr[unLo]+d]) - med;
668
281M
            if (n == 0) { 
669
239M
               mswap(ptr[unLo], ptr[ltLo]); 
670
239M
               ltLo++; unLo++; continue; 
671
239M
            };
672
42.2M
            if (n >  0) break;
673
30.7M
            unLo++;
674
30.7M
         }
675
184M
         while (True) {
676
184M
            if (unLo > unHi) break;
677
182M
            n = ((Int32)block[ptr[unHi]+d]) - med;
678
182M
            if (n == 0) { 
679
144M
               mswap(ptr[unHi], ptr[gtHi]); 
680
144M
               gtHi--; unHi--; continue; 
681
144M
            };
682
38.7M
            if (n <  0) break;
683
27.8M
            unHi--;
684
27.8M
         }
685
12.4M
         if (unLo > unHi) break;
686
10.9M
         mswap(ptr[unLo], ptr[unHi]); unLo++; unHi--;
687
10.9M
      }
688
689
1.53M
      AssertD ( unHi == unLo-1, "mainQSort3(2)" );
690
691
1.53M
      if (gtHi < ltLo) {
692
426k
         mpush(lo, hi, d+1 );
693
426k
         continue;
694
426k
      }
695
696
1.10M
      n = mmin(ltLo-lo, unLo-ltLo); mvswap(lo, unLo-n, n);
697
1.10M
      m = mmin(hi-gtHi, gtHi-unHi); mvswap(unLo, hi-m+1, m);
698
699
1.10M
      n = lo + unLo - ltLo - 1;
700
1.10M
      m = hi - (gtHi - unHi) + 1;
701
702
1.10M
      nextLo[0] = lo;  nextHi[0] = n;   nextD[0] = d;
703
1.10M
      nextLo[1] = m;   nextHi[1] = hi;  nextD[1] = d;
704
1.10M
      nextLo[2] = n+1; nextHi[2] = m-1; nextD[2] = d+1;
705
706
1.10M
      if (mnextsize(0) < mnextsize(1)) mnextswap(0,1);
707
1.10M
      if (mnextsize(1) < mnextsize(2)) mnextswap(1,2);
708
1.10M
      if (mnextsize(0) < mnextsize(1)) mnextswap(0,1);
709
710
1.10M
      AssertD (mnextsize(0) >= mnextsize(1), "mainQSort3(8)" );
711
1.10M
      AssertD (mnextsize(1) >= mnextsize(2), "mainQSort3(9)" );
712
713
1.10M
      mpush (nextLo[0], nextHi[0], nextD[0]);
714
1.10M
      mpush (nextLo[1], nextHi[1], nextD[1]);
715
1.10M
      mpush (nextLo[2], nextHi[2], nextD[2]);
716
1.10M
   }
717
8.81M
}
718
719
#undef mswap
720
#undef mvswap
721
#undef mpush
722
#undef mpop
723
#undef mmin
724
#undef mnextsize
725
#undef mnextswap
726
#undef MAIN_QSORT_SMALL_THRESH
727
#undef MAIN_QSORT_DEPTH_THRESH
728
#undef MAIN_QSORT_STACK_SIZE
729
730
731
/*---------------------------------------------*/
732
/* Pre:
733
      nblock > N_OVERSHOOT
734
      block32 exists for [0 .. nblock-1 +N_OVERSHOOT]
735
      ((UChar*)block32) [0 .. nblock-1] holds block
736
      ptr exists for [0 .. nblock-1]
737
738
   Post:
739
      ((UChar*)block32) [0 .. nblock-1] holds block
740
      All other areas of block32 destroyed
741
      ftab [0 .. 65536 ] destroyed
742
      ptr [0 .. nblock-1] holds sorted order
743
      if (*budget < 0), sorting was abandoned
744
*/
745
746
5.49M
#define BIGFREQ(b) (ftab[((b)+1) << 8] - ftab[(b) << 8])
747
646M
#define SETMASK (1 << 21)
748
324M
#define CLEARMASK (~(SETMASK))
749
750
static
751
void mainSort ( UInt32* ptr, 
752
                UChar*  block,
753
                UInt16* quadrant, 
754
                UInt32* ftab,
755
                Int32   nblock,
756
                Int32   verb,
757
                Int32*  budget )
758
1.66k
{
759
1.66k
   Int32  i, j, k, ss, sb;
760
1.66k
   Int32  runningOrder[256];
761
1.66k
   Bool   bigDone[256];
762
1.66k
   Int32  copyStart[256];
763
1.66k
   Int32  copyEnd  [256];
764
1.66k
   UChar  c1;
765
1.66k
   Int32  numQSorted;
766
1.66k
   UInt16 s;
767
1.66k
   if (verb >= 4) VPrintf0 ( "        main sort initialise ...\n" );
768
769
   /*-- set up the 2-byte frequency table --*/
770
109M
   for (i = 65536; i >= 0; i--) ftab[i] = 0;
771
772
1.66k
   j = block[0] << 8;
773
1.66k
   i = nblock-1;
774
109M
   for (; i >= 3; i -= 4) {
775
109M
      quadrant[i] = 0;
776
109M
      j = (j >> 8) | ( ((UInt16)block[i]) << 8);
777
109M
      ftab[j]++;
778
109M
      quadrant[i-1] = 0;
779
109M
      j = (j >> 8) | ( ((UInt16)block[i-1]) << 8);
780
109M
      ftab[j]++;
781
109M
      quadrant[i-2] = 0;
782
109M
      j = (j >> 8) | ( ((UInt16)block[i-2]) << 8);
783
109M
      ftab[j]++;
784
109M
      quadrant[i-3] = 0;
785
109M
      j = (j >> 8) | ( ((UInt16)block[i-3]) << 8);
786
109M
      ftab[j]++;
787
109M
   }
788
3.58k
   for (; i >= 0; i--) {
789
1.91k
      quadrant[i] = 0;
790
1.91k
      j = (j >> 8) | ( ((UInt16)block[i]) << 8);
791
1.91k
      ftab[j]++;
792
1.91k
   }
793
794
   /*-- (emphasises close relationship of block & quadrant) --*/
795
58.2k
   for (i = 0; i < BZ_N_OVERSHOOT; i++) {
796
56.5k
      block   [nblock+i] = block[i];
797
56.5k
      quadrant[nblock+i] = 0;
798
56.5k
   }
799
800
1.66k
   if (verb >= 4) VPrintf0 ( "        bucket sorting ...\n" );
801
802
   /*-- Complete the initial radix sort --*/
803
109M
   for (i = 1; i <= 65536; i++) ftab[i] += ftab[i-1];
804
805
1.66k
   s = block[0] << 8;
806
1.66k
   i = nblock-1;
807
109M
   for (; i >= 3; i -= 4) {
808
109M
      s = (s >> 8) | (block[i] << 8);
809
109M
      j = ftab[s] -1;
810
109M
      ftab[s] = j;
811
109M
      ptr[j] = i;
812
109M
      s = (s >> 8) | (block[i-1] << 8);
813
109M
      j = ftab[s] -1;
814
109M
      ftab[s] = j;
815
109M
      ptr[j] = i-1;
816
109M
      s = (s >> 8) | (block[i-2] << 8);
817
109M
      j = ftab[s] -1;
818
109M
      ftab[s] = j;
819
109M
      ptr[j] = i-2;
820
109M
      s = (s >> 8) | (block[i-3] << 8);
821
109M
      j = ftab[s] -1;
822
109M
      ftab[s] = j;
823
109M
      ptr[j] = i-3;
824
109M
   }
825
3.58k
   for (; i >= 0; i--) {
826
1.91k
      s = (s >> 8) | (block[i] << 8);
827
1.91k
      j = ftab[s] -1;
828
1.91k
      ftab[s] = j;
829
1.91k
      ptr[j] = i;
830
1.91k
   }
831
832
   /*--
833
      Now ftab contains the first loc of every small bucket.
834
      Calculate the running order, from smallest to largest
835
      big bucket.
836
   --*/
837
427k
   for (i = 0; i <= 255; i++) {
838
425k
      bigDone     [i] = False;
839
425k
      runningOrder[i] = i;
840
425k
   }
841
842
1.66k
   {
843
1.66k
      Int32 vv;
844
1.66k
      Int32 h = 1;
845
8.32k
      do h = 3 * h + 1; while (h <= 256);
846
8.32k
      do {
847
8.32k
         h = h / 3;
848
1.84M
         for (i = h; i <= 255; i++) {
849
1.83M
            vv = runningOrder[i];
850
1.83M
            j = i;
851
2.74M
            while ( BIGFREQ(runningOrder[j-h]) > BIGFREQ(vv) ) {
852
1.02M
               runningOrder[j] = runningOrder[j-h];
853
1.02M
               j = j - h;
854
1.02M
               if (j <= (h - 1)) goto zero;
855
1.02M
            }
856
1.83M
            zero:
857
1.83M
            runningOrder[j] = vv;
858
1.83M
         }
859
8.32k
      } while (h != 1);
860
1.66k
   }
861
862
   /*--
863
      The main sorting loop.
864
   --*/
865
866
1.66k
   numQSorted = 0;
867
868
421k
   for (i = 0; i <= 255; i++) {
869
870
      /*--
871
         Process big buckets, starting with the least full.
872
         Basically this is a 3-step process in which we call
873
         mainQSort3 to sort the small buckets [ss, j], but
874
         also make a big effort to avoid the calls if we can.
875
      --*/
876
421k
      ss = runningOrder[i];
877
878
      /*--
879
         Step 1:
880
         Complete the big bucket [ss] by quicksorting
881
         any unsorted small buckets [ss, j], for j != ss.  
882
         Hopefully previous pointer-scanning phases have already
883
         completed many of the small buckets [ss, j], so
884
         we don't have to sort them at all.
885
      --*/
886
108M
      for (j = 0; j <= 255; j++) {
887
107M
         if (j != ss) {
888
107M
            sb = (ss << 8) + j;
889
107M
            if ( ! (ftab[sb] & SETMASK) ) {
890
53.9M
               Int32 lo = ftab[sb]   & CLEARMASK;
891
53.9M
               Int32 hi = (ftab[sb+1] & CLEARMASK) - 1;
892
53.9M
               if (hi > lo) {
893
8.81M
                  if (verb >= 4)
894
0
                     VPrintf4 ( "        qsort [0x%x, 0x%x]   "
895
8.81M
                                "done %d   this %d\n",
896
8.81M
                                ss, j, numQSorted, hi - lo + 1 );
897
8.81M
                  mainQSort3 ( 
898
8.81M
                     ptr, block, quadrant, nblock, 
899
8.81M
                     lo, hi, BZ_N_RADIX, budget 
900
8.81M
                  );   
901
8.81M
                  numQSorted += (hi - lo + 1);
902
8.81M
                  if (*budget < 0) return;
903
8.81M
               }
904
53.9M
            }
905
107M
            ftab[sb] |= SETMASK;
906
107M
         }
907
107M
      }
908
909
420k
      AssertH ( !bigDone[ss], 1006 );
910
911
      /*--
912
         Step 2:
913
         Now scan this big bucket [ss] so as to synthesise the
914
         sorted order for small buckets [t, ss] for all t,
915
         including, magically, the bucket [ss,ss] too.
916
         This will avoid doing Real Work in subsequent Step 1's.
917
      --*/
918
420k
      {
919
107M
         for (j = 0; j <= 255; j++) {
920
107M
            copyStart[j] =  ftab[(j << 8) + ss]     & CLEARMASK;
921
107M
            copyEnd  [j] = (ftab[(j << 8) + ss + 1] & CLEARMASK) - 1;
922
107M
         }
923
95.8M
         for (j = ftab[ss << 8] & CLEARMASK; j < copyStart[ss]; j++) {
924
95.4M
            k = ptr[j]-1; if (k < 0) k += nblock;
925
95.4M
            c1 = block[k];
926
95.4M
            if (!bigDone[c1])
927
53.6M
               ptr[ copyStart[c1]++ ] = k;
928
95.4M
         }
929
102M
         for (j = (ftab[(ss+1) << 8] & CLEARMASK) - 1; j > copyEnd[ss]; j--) {
930
102M
            k = ptr[j]-1; if (k < 0) k += nblock;
931
102M
            c1 = block[k];
932
102M
            if (!bigDone[c1]) 
933
53.8M
               ptr[ copyEnd[c1]-- ] = k;
934
102M
         }
935
420k
      }
936
937
420k
      AssertH ( (copyStart[ss]-1 == copyEnd[ss])
938
420k
                || 
939
                /* Extremely rare case missing in bzip2-1.0.0 and 1.0.1.
940
                   Necessity for this case is demonstrated by compressing 
941
                   a sequence of approximately 48.5 million of character 
942
                   251; 1.0.0/1.0.1 will then die here. */
943
420k
                (copyStart[ss] == 0 && copyEnd[ss] == nblock-1),
944
420k
                1007 )
945
946
107M
      for (j = 0; j <= 255; j++) ftab[(j << 8) + ss] |= SETMASK;
947
948
      /*--
949
         Step 3:
950
         The [ss] big bucket is now done.  Record this fact,
951
         and update the quadrant descriptors.  Remember to
952
         update quadrants in the overshoot area too, if
953
         necessary.  The "if (i < 255)" test merely skips
954
         this updating for the last bucket processed, since
955
         updating for the last bucket is pointless.
956
957
         The quadrant array provides a way to incrementally
958
         cache sort orderings, as they appear, so as to 
959
         make subsequent comparisons in fullGtU() complete
960
         faster.  For repetitive blocks this makes a big
961
         difference (but not big enough to be able to avoid
962
         the fallback sorting mechanism, exponential radix sort).
963
964
         The precise meaning is: at all times:
965
966
            for 0 <= i < nblock and 0 <= j <= nblock
967
968
            if block[i] != block[j], 
969
970
               then the relative values of quadrant[i] and 
971
                    quadrant[j] are meaningless.
972
973
               else {
974
                  if quadrant[i] < quadrant[j]
975
                     then the string starting at i lexicographically
976
                     precedes the string starting at j
977
978
                  else if quadrant[i] > quadrant[j]
979
                     then the string starting at j lexicographically
980
                     precedes the string starting at i
981
982
                  else
983
                     the relative ordering of the strings starting
984
                     at i and j has not yet been determined.
985
               }
986
      --*/
987
420k
      bigDone[ss] = True;
988
989
420k
      if (i < 255) {
990
419k
         Int32 bbStart  = ftab[ss << 8] & CLEARMASK;
991
419k
         Int32 bbSize   = (ftab[(ss+1) << 8] & CLEARMASK) - bbStart;
992
419k
         Int32 shifts   = 0;
993
994
419k
         while ((bbSize >> shifts) > 65534) shifts++;
995
996
180M
         for (j = bbSize-1; j >= 0; j--) {
997
179M
            Int32 a2update     = ptr[bbStart + j];
998
179M
            UInt16 qVal        = (UInt16)(j >> shifts);
999
179M
            quadrant[a2update] = qVal;
1000
179M
            if (a2update < BZ_N_OVERSHOOT)
1001
27.2k
               quadrant[a2update + nblock] = qVal;
1002
179M
         }
1003
419k
         AssertH ( ((bbSize-1) >> shifts) <= 65535, 1002 );
1004
419k
      }
1005
1006
420k
   }
1007
1008
725
   if (verb >= 4)
1009
0
      VPrintf3 ( "        %d pointers, %d sorted, %d scanned\n",
1010
725
                 nblock, numQSorted, nblock - numQSorted );
1011
725
}
1012
1013
#undef BIGFREQ
1014
#undef SETMASK
1015
#undef CLEARMASK
1016
1017
1018
/*---------------------------------------------*/
1019
/* Pre:
1020
      nblock > 0
1021
      arr2 exists for [0 .. nblock-1 +N_OVERSHOOT]
1022
      ((UChar*)arr2)  [0 .. nblock-1] holds block
1023
      arr1 exists for [0 .. nblock-1]
1024
1025
   Post:
1026
      ((UChar*)arr2) [0 .. nblock-1] holds block
1027
      All other areas of block destroyed
1028
      ftab [ 0 .. 65536 ] destroyed
1029
      arr1 [0 .. nblock-1] holds sorted order
1030
*/
1031
void BZ2_blockSort ( EState* s )
1032
3.91k
{
1033
3.91k
   UInt32* ptr    = s->ptr; 
1034
3.91k
   UChar*  block  = s->block;
1035
3.91k
   UInt32* ftab   = s->ftab;
1036
3.91k
   Int32   nblock = s->nblock;
1037
3.91k
   Int32   verb   = s->verbosity;
1038
3.91k
   Int32   wfact  = s->workFactor;
1039
3.91k
   UInt16* quadrant;
1040
3.91k
   Int32   budget;
1041
3.91k
   Int32   budgetInit;
1042
3.91k
   Int32   i;
1043
1044
3.91k
   if (nblock < 10000) {
1045
2.25k
      fallbackSort ( s->arr1, s->arr2, ftab, nblock, verb );
1046
2.25k
   } else {
1047
      /* Calculate the location for quadrant, remembering to get
1048
         the alignment right.  Assumes that &(block[0]) is at least
1049
         2-byte aligned -- this should be ok since block is really
1050
         the first section of arr2.
1051
      */
1052
1.66k
      i = nblock+BZ_N_OVERSHOOT;
1053
1.66k
      if (i & 1) i++;
1054
1.66k
      quadrant = (UInt16*)(&(block[i]));
1055
1056
      /* (wfact-1) / 3 puts the default-factor-30
1057
         transition point at very roughly the same place as 
1058
         with v0.1 and v0.9.0.  
1059
         Not that it particularly matters any more, since the
1060
         resulting compressed stream is now the same regardless
1061
         of whether or not we use the main sort or fallback sort.
1062
      */
1063
1.66k
      if (wfact < 1  ) wfact = 1;
1064
1.66k
      if (wfact > 100) wfact = 100;
1065
1.66k
      budgetInit = nblock * ((wfact-1) / 3);
1066
1.66k
      budget = budgetInit;
1067
1068
1.66k
      mainSort ( ptr, block, quadrant, ftab, nblock, verb, &budget );
1069
1.66k
      if (verb >= 3) 
1070
0
         VPrintf3 ( "      %d work, %d block, ratio %5.2f\n",
1071
1.66k
                    budgetInit - budget,
1072
1.66k
                    nblock, 
1073
1.66k
                    (float)(budgetInit - budget) /
1074
1.66k
                    (float)(nblock==0 ? 1 : nblock) ); 
1075
1.66k
      if (budget < 0) {
1076
939
         if (verb >= 2) 
1077
0
            VPrintf0 ( "    too repetitive; using fallback"
1078
939
                       " sorting algorithm\n" );
1079
939
         fallbackSort ( s->arr1, s->arr2, ftab, nblock, verb );
1080
939
      }
1081
1.66k
   }
1082
1083
3.91k
   s->origPtr = -1;
1084
253M
   for (i = 0; i < s->nblock; i++)
1085
253M
      if (ptr[i] == 0)
1086
3.91k
         { s->origPtr = i; break; };
1087
1088
3.91k
   AssertH( s->origPtr != -1, 1003 );
1089
3.91k
}
1090
1091
1092
/*-------------------------------------------------------------*/
1093
/*--- end                                       blocksort.c ---*/
1094
/*-------------------------------------------------------------*/