forked from madler/zlib
-
Notifications
You must be signed in to change notification settings - Fork 3
/
crc32.c
1049 lines (939 loc) · 30.9 KB
/
crc32.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/* crc32.c -- compute the CRC-32 of a data stream
* Copyright (C) 1995-2022 Mark Adler
* For conditions of distribution and use, see copyright notice in zlib.h
*
* This interleaved implementation of a CRC makes use of pipelined multiple
* arithmetic-logic units, commonly found in modern CPU cores. It is due to
* Kadatch and Jenkins (2010). See doc/crc-doc.1.0.pdf in this distribution.
*/
/* @(#) $Id$ */
/*
Note on the use of DYNAMIC_CRC_TABLE: there is no mutex or semaphore
protection on the static variables used to control the first-use generation
of the crc tables. Therefore, if you #define DYNAMIC_CRC_TABLE, you should
first call get_crc_table() to initialize the tables before allowing more than
one thread to use crc32().
MAKECRCH can be #defined to write out crc32.h. A main() routine is also
produced, so that this one source file can be compiled to an executable.
*/
#ifdef MAKECRCH
# include <stdio.h>
# ifndef DYNAMIC_CRC_TABLE
# define DYNAMIC_CRC_TABLE
# endif /* !DYNAMIC_CRC_TABLE */
#endif /* MAKECRCH */
#include "zutil.h" /* for Z_U4, Z_U8, z_crc_t, and FAR definitions */
/*
A CRC of a message is computed on N braids of words in the message, where
each word consists of W bytes (4 or 8). If N is 3, for example, then three
running sparse CRCs are calculated respectively on each braid, at these
indices in the array of words: 0, 3, 6, ..., 1, 4, 7, ..., and 2, 5, 8, ...
This is done starting at a word boundary, and continues until as many blocks
of N * W bytes as are available have been processed. The results are combined
into a single CRC at the end. For this code, N must be in the range 1..6 and
W must be 4 or 8. The upper limit on N can be increased if desired by adding
more #if blocks, extending the patterns apparent in the code. In addition,
crc32.h would need to be regenerated, if the maximum N value is increased.
N and W are chosen empirically by benchmarking the execution time on a given
processor. The choices for N and W below were based on testing on Intel Kaby
Lake i7, AMD Ryzen 7, ARM Cortex-A57, Sparc64-VII, PowerPC POWER9, and MIPS64
Octeon II processors. The Intel, AMD, and ARM processors were all fastest
with N=5, W=8. The Sparc, PowerPC, and MIPS64 were all fastest at N=5, W=4.
They were all tested with either gcc or clang, all using the -O3 optimization
level. Your mileage may vary.
*/
/* Define N */
#ifdef Z_TESTN
# define N Z_TESTN
#else
# define N 5
#endif
#if N < 1 || N > 6
# error N must be in 1..6
#endif
/*
z_crc_t must be at least 32 bits. z_word_t must be at least as long as
z_crc_t. It is assumed here that z_word_t is either 32 bits or 64 bits, and
that bytes are eight bits.
*/
/*
Define W and the associated z_word_t type. If W is not defined, then a
braided calculation is not used, and the associated tables and code are not
compiled.
*/
#ifdef Z_TESTW
# if Z_TESTW-1 != -1
# define W Z_TESTW
# endif
#else
# ifdef MAKECRCH
# define W 8 /* required for MAKECRCH */
# else
# if defined(__x86_64__) || defined(__aarch64__)
# define W 8
# else
# define W 4
# endif
# endif
#endif
#ifdef W
# if W == 8 && defined(Z_U8)
typedef Z_U8 z_word_t;
# elif defined(Z_U4)
# undef W
# define W 4
typedef Z_U4 z_word_t;
# else
# undef W
# endif
#endif
/* If available, use the ARM processor CRC32 instruction. */
#if defined(__aarch64__) && defined(__ARM_FEATURE_CRC32) && W == 8
# define ARMCRC32
#endif
#if defined(W) && (!defined(ARMCRC32) || defined(DYNAMIC_CRC_TABLE))
/*
Swap the bytes in a z_word_t to convert between little and big endian. Any
self-respecting compiler will optimize this to a single machine byte-swap
instruction, if one is available. This assumes that word_t is either 32 bits
or 64 bits.
*/
local z_word_t byte_swap(z_word_t word) {
# if W == 8
return
(word & 0xff00000000000000) >> 56 |
(word & 0xff000000000000) >> 40 |
(word & 0xff0000000000) >> 24 |
(word & 0xff00000000) >> 8 |
(word & 0xff000000) << 8 |
(word & 0xff0000) << 24 |
(word & 0xff00) << 40 |
(word & 0xff) << 56;
# else /* W == 4 */
return
(word & 0xff000000) >> 24 |
(word & 0xff0000) >> 8 |
(word & 0xff00) << 8 |
(word & 0xff) << 24;
# endif
}
#endif
#ifdef DYNAMIC_CRC_TABLE
/* =========================================================================
* Table of powers of x for combining CRC-32s, filled in by make_crc_table()
* below.
*/
local z_crc_t FAR x2n_table[32];
#else
/* =========================================================================
* Tables for byte-wise and braided CRC-32 calculations, and a table of powers
* of x for combining CRC-32s, all made by make_crc_table().
*/
# include "crc32.h"
#endif
/* CRC polynomial. */
#define POLY 0xedb88320 /* p(x) reflected, with x^32 implied */
/*
Return a(x) multiplied by b(x) modulo p(x), where p(x) is the CRC polynomial,
reflected. For speed, this requires that a not be zero.
*/
local z_crc_t multmodp(z_crc_t a, z_crc_t b) {
z_crc_t m, p;
m = (z_crc_t)1 << 31;
p = 0;
for (;;) {
if (a & m) {
p ^= b;
if ((a & (m - 1)) == 0)
break;
}
m >>= 1;
b = b & 1 ? (b >> 1) ^ POLY : b >> 1;
}
return p;
}
/*
Return x^(n * 2^k) modulo p(x). Requires that x2n_table[] has been
initialized.
*/
local z_crc_t x2nmodp(z_off64_t n, unsigned k) {
z_crc_t p;
p = (z_crc_t)1 << 31; /* x^0 == 1 */
while (n) {
if (n & 1)
p = multmodp(x2n_table[k & 31], p);
n >>= 1;
k++;
}
return p;
}
#ifdef DYNAMIC_CRC_TABLE
/* =========================================================================
* Build the tables for byte-wise and braided CRC-32 calculations, and a table
* of powers of x for combining CRC-32s.
*/
local z_crc_t FAR crc_table[256];
#ifdef W
local z_word_t FAR crc_big_table[256];
local z_crc_t FAR crc_braid_table[W][256];
local z_word_t FAR crc_braid_big_table[W][256];
local void braid(z_crc_t [][256], z_word_t [][256], int, int);
#endif
#ifdef MAKECRCH
local void write_table(FILE *, const z_crc_t FAR *, int);
local void write_table32hi(FILE *, const z_word_t FAR *, int);
local void write_table64(FILE *, const z_word_t FAR *, int);
#endif /* MAKECRCH */
/*
Define a once() function depending on the availability of atomics. If this is
compiled with DYNAMIC_CRC_TABLE defined, and if CRCs will be computed in
multiple threads, and if atomics are not available, then get_crc_table() must
be called to initialize the tables and must return before any threads are
allowed to compute or combine CRCs.
*/
/* Definition of once functionality. */
typedef struct once_s once_t;
/* Check for the availability of atomics. */
#if defined(__STDC__) && __STDC_VERSION__ >= 201112L && \
!defined(__STDC_NO_ATOMICS__)
#include <stdatomic.h>
/* Structure for once(), which must be initialized with ONCE_INIT. */
struct once_s {
atomic_flag begun;
atomic_int done;
};
#define ONCE_INIT {ATOMIC_FLAG_INIT, 0}
/*
Run the provided init() function exactly once, even if multiple threads
invoke once() at the same time. The state must be a once_t initialized with
ONCE_INIT.
*/
local void once(once_t *state, void (*init)(void)) {
if (!atomic_load(&state->done)) {
if (atomic_flag_test_and_set(&state->begun))
while (!atomic_load(&state->done))
;
else {
init();
atomic_store(&state->done, 1);
}
}
}
#else /* no atomics */
/* Structure for once(), which must be initialized with ONCE_INIT. */
struct once_s {
volatile int begun;
volatile int done;
};
#define ONCE_INIT {0, 0}
/* Test and set. Alas, not atomic, but tries to minimize the period of
vulnerability. */
local int test_and_set(int volatile *flag) {
int was;
was = *flag;
*flag = 1;
return was;
}
/* Run the provided init() function once. This is not thread-safe. */
local void once(once_t *state, void (*init)(void)) {
if (!state->done) {
if (test_and_set(&state->begun))
while (!state->done)
;
else {
init();
state->done = 1;
}
}
}
#endif
/* State for once(). */
local once_t made = ONCE_INIT;
/*
Generate tables for a byte-wise 32-bit CRC calculation on the polynomial:
x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x+1.
Polynomials over GF(2) are represented in binary, one bit per coefficient,
with the lowest powers in the most significant bit. Then adding polynomials
is just exclusive-or, and multiplying a polynomial by x is a right shift by
one. If we call the above polynomial p, and represent a byte as the
polynomial q, also with the lowest power in the most significant bit (so the
byte 0xb1 is the polynomial x^7+x^3+x^2+1), then the CRC is (q*x^32) mod p,
where a mod b means the remainder after dividing a by b.
This calculation is done using the shift-register method of multiplying and
taking the remainder. The register is initialized to zero, and for each
incoming bit, x^32 is added mod p to the register if the bit is a one (where
x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by x
(which is shifting right by one and adding x^32 mod p if the bit shifted out
is a one). We start with the highest power (least significant bit) of q and
repeat for all eight bits of q.
The table is simply the CRC of all possible eight bit values. This is all the
information needed to generate CRCs on data a byte at a time for all
combinations of CRC register values and incoming bytes.
*/
local void make_crc_table(void) {
unsigned i, j, n;
z_crc_t p;
/* initialize the CRC of bytes tables */
for (i = 0; i < 256; i++) {
p = i;
for (j = 0; j < 8; j++)
p = p & 1 ? (p >> 1) ^ POLY : p >> 1;
crc_table[i] = p;
#ifdef W
crc_big_table[i] = byte_swap(p);
#endif
}
/* initialize the x^2^n mod p(x) table */
p = (z_crc_t)1 << 30; /* x^1 */
x2n_table[0] = p;
for (n = 1; n < 32; n++)
x2n_table[n] = p = multmodp(p, p);
#ifdef W
/* initialize the braiding tables -- needs x2n_table[] */
braid(crc_braid_table, crc_braid_big_table, N, W);
#endif
#ifdef MAKECRCH
{
/*
The crc32.h header file contains tables for both 32-bit and 64-bit
z_word_t's, and so requires a 64-bit type be available. In that case,
z_word_t must be defined to be 64-bits. This code then also generates
and writes out the tables for the case that z_word_t is 32 bits.
*/
#if !defined(W) || W != 8
# error Need a 64-bit integer type in order to generate crc32.h.
#endif
FILE *out;
int k, n;
z_crc_t ltl[8][256];
z_word_t big[8][256];
out = fopen("crc32.h", "w");
if (out == NULL) return;
/* write out little-endian CRC table to crc32.h */
fprintf(out,
"/* crc32.h -- tables for rapid CRC calculation\n"
" * Generated automatically by crc32.c\n */\n"
"\n"
"local const z_crc_t FAR crc_table[] = {\n"
" ");
write_table(out, crc_table, 256);
fprintf(out,
"};\n");
/* write out big-endian CRC table for 64-bit z_word_t to crc32.h */
fprintf(out,
"\n"
"#ifdef W\n"
"\n"
"#if W == 8\n"
"\n"
"local const z_word_t FAR crc_big_table[] = {\n"
" ");
write_table64(out, crc_big_table, 256);
fprintf(out,
"};\n");
/* write out big-endian CRC table for 32-bit z_word_t to crc32.h */
fprintf(out,
"\n"
"#else /* W == 4 */\n"
"\n"
"local const z_word_t FAR crc_big_table[] = {\n"
" ");
write_table32hi(out, crc_big_table, 256);
fprintf(out,
"};\n"
"\n"
"#endif\n");
/* write out braid tables for each value of N */
for (n = 1; n <= 6; n++) {
fprintf(out,
"\n"
"#if N == %d\n", n);
/* compute braid tables for this N and 64-bit word_t */
braid(ltl, big, n, 8);
/* write out braid tables for 64-bit z_word_t to crc32.h */
fprintf(out,
"\n"
"#if W == 8\n"
"\n"
"local const z_crc_t FAR crc_braid_table[][256] = {\n");
for (k = 0; k < 8; k++) {
fprintf(out, " {");
write_table(out, ltl[k], 256);
fprintf(out, "}%s", k < 7 ? ",\n" : "");
}
fprintf(out,
"};\n"
"\n"
"local const z_word_t FAR crc_braid_big_table[][256] = {\n");
for (k = 0; k < 8; k++) {
fprintf(out, " {");
write_table64(out, big[k], 256);
fprintf(out, "}%s", k < 7 ? ",\n" : "");
}
fprintf(out,
"};\n");
/* compute braid tables for this N and 32-bit word_t */
braid(ltl, big, n, 4);
/* write out braid tables for 32-bit z_word_t to crc32.h */
fprintf(out,
"\n"
"#else /* W == 4 */\n"
"\n"
"local const z_crc_t FAR crc_braid_table[][256] = {\n");
for (k = 0; k < 4; k++) {
fprintf(out, " {");
write_table(out, ltl[k], 256);
fprintf(out, "}%s", k < 3 ? ",\n" : "");
}
fprintf(out,
"};\n"
"\n"
"local const z_word_t FAR crc_braid_big_table[][256] = {\n");
for (k = 0; k < 4; k++) {
fprintf(out, " {");
write_table32hi(out, big[k], 256);
fprintf(out, "}%s", k < 3 ? ",\n" : "");
}
fprintf(out,
"};\n"
"\n"
"#endif\n"
"\n"
"#endif\n");
}
fprintf(out,
"\n"
"#endif\n");
/* write out zeros operator table to crc32.h */
fprintf(out,
"\n"
"local const z_crc_t FAR x2n_table[] = {\n"
" ");
write_table(out, x2n_table, 32);
fprintf(out,
"};\n");
fclose(out);
}
#endif /* MAKECRCH */
}
#ifdef MAKECRCH
/*
Write the 32-bit values in table[0..k-1] to out, five per line in
hexadecimal separated by commas.
*/
local void write_table(FILE *out, const z_crc_t FAR *table, int k) {
int n;
for (n = 0; n < k; n++)
fprintf(out, "%s0x%08lx%s", n == 0 || n % 5 ? "" : " ",
(unsigned long)(table[n]),
n == k - 1 ? "" : (n % 5 == 4 ? ",\n" : ", "));
}
/*
Write the high 32-bits of each value in table[0..k-1] to out, five per line
in hexadecimal separated by commas.
*/
local void write_table32hi(FILE *out, const z_word_t FAR *table, int k) {
int n;
for (n = 0; n < k; n++)
fprintf(out, "%s0x%08lx%s", n == 0 || n % 5 ? "" : " ",
(unsigned long)(table[n] >> 32),
n == k - 1 ? "" : (n % 5 == 4 ? ",\n" : ", "));
}
/*
Write the 64-bit values in table[0..k-1] to out, three per line in
hexadecimal separated by commas. This assumes that if there is a 64-bit
type, then there is also a long long integer type, and it is at least 64
bits. If not, then the type cast and format string can be adjusted
accordingly.
*/
local void write_table64(FILE *out, const z_word_t FAR *table, int k) {
int n;
for (n = 0; n < k; n++)
fprintf(out, "%s0x%016llx%s", n == 0 || n % 3 ? "" : " ",
(unsigned long long)(table[n]),
n == k - 1 ? "" : (n % 3 == 2 ? ",\n" : ", "));
}
/* Actually do the deed. */
int main(void) {
make_crc_table();
return 0;
}
#endif /* MAKECRCH */
#ifdef W
/*
Generate the little and big-endian braid tables for the given n and z_word_t
size w. Each array must have room for w blocks of 256 elements.
*/
local void braid(z_crc_t ltl[][256], z_word_t big[][256], int n, int w) {
int k;
z_crc_t i, p, q;
for (k = 0; k < w; k++) {
p = x2nmodp((n * w + 3 - k) << 3, 0);
ltl[k][0] = 0;
big[w - 1 - k][0] = 0;
for (i = 1; i < 256; i++) {
ltl[k][i] = q = multmodp(i << 24, p);
big[w - 1 - k][i] = byte_swap(q);
}
}
}
#endif
#endif /* DYNAMIC_CRC_TABLE */
/* =========================================================================
* This function can be used by asm versions of crc32(), and to force the
* generation of the CRC tables in a threaded application.
*/
const z_crc_t FAR * ZEXPORT get_crc_table(void) {
#ifdef DYNAMIC_CRC_TABLE
once(&made, make_crc_table);
#endif /* DYNAMIC_CRC_TABLE */
return (const z_crc_t FAR *)crc_table;
}
/* =========================================================================
* Use ARM machine instructions if available. This will compute the CRC about
* ten times faster than the braided calculation. This code does not check for
* the presence of the CRC instruction at run time. __ARM_FEATURE_CRC32 will
* only be defined if the compilation specifies an ARM processor architecture
* that has the instructions. For example, compiling with -march=armv8.1-a or
* -march=armv8-a+crc, or -march=native if the compile machine has the crc32
* instructions.
*/
#ifdef ARMCRC32
/*
Constants empirically determined to maximize speed. These values are from
measurements on a Cortex-A57. Your mileage may vary.
*/
#define Z_BATCH 3990 /* number of words in a batch */
#define Z_BATCH_ZEROS 0xa10d3d0c /* computed from Z_BATCH = 3990 */
#define Z_BATCH_MIN 800 /* fewest words in a final batch */
unsigned long ZEXPORT crc32_z(unsigned long crc, const unsigned char FAR *buf,
z_size_t len) {
z_crc_t val;
z_word_t crc1, crc2;
const z_word_t *word;
z_word_t val0, val1, val2;
z_size_t last, last2, i;
z_size_t num;
/* Return initial CRC, if requested. */
if (buf == Z_NULL) return 0;
#ifdef DYNAMIC_CRC_TABLE
once(&made, make_crc_table);
#endif /* DYNAMIC_CRC_TABLE */
/* Pre-condition the CRC */
crc = (~crc) & 0xffffffff;
/* Compute the CRC up to a word boundary. */
while (len && ((z_size_t)buf & 7) != 0) {
len--;
val = *buf++;
__asm__ volatile("crc32b %w0, %w0, %w1" : "+r"(crc) : "r"(val));
}
/* Prepare to compute the CRC on full 64-bit words word[0..num-1]. */
word = (z_word_t const *)buf;
num = len >> 3;
len &= 7;
/* Do three interleaved CRCs to realize the throughput of one crc32x
instruction per cycle. Each CRC is calculated on Z_BATCH words. The
three CRCs are combined into a single CRC after each set of batches. */
while (num >= 3 * Z_BATCH) {
crc1 = 0;
crc2 = 0;
for (i = 0; i < Z_BATCH; i++) {
val0 = word[i];
val1 = word[i + Z_BATCH];
val2 = word[i + 2 * Z_BATCH];
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc) : "r"(val0));
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc1) : "r"(val1));
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc2) : "r"(val2));
}
word += 3 * Z_BATCH;
num -= 3 * Z_BATCH;
crc = multmodp(Z_BATCH_ZEROS, crc) ^ crc1;
crc = multmodp(Z_BATCH_ZEROS, crc) ^ crc2;
}
/* Do one last smaller batch with the remaining words, if there are enough
to pay for the combination of CRCs. */
last = num / 3;
if (last >= Z_BATCH_MIN) {
last2 = last << 1;
crc1 = 0;
crc2 = 0;
for (i = 0; i < last; i++) {
val0 = word[i];
val1 = word[i + last];
val2 = word[i + last2];
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc) : "r"(val0));
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc1) : "r"(val1));
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc2) : "r"(val2));
}
word += 3 * last;
num -= 3 * last;
val = x2nmodp(last, 6);
crc = multmodp(val, crc) ^ crc1;
crc = multmodp(val, crc) ^ crc2;
}
/* Compute the CRC on any remaining words. */
for (i = 0; i < num; i++) {
val0 = word[i];
__asm__ volatile("crc32x %w0, %w0, %x1" : "+r"(crc) : "r"(val0));
}
word += num;
/* Complete the CRC on any remaining bytes. */
buf = (const unsigned char FAR *)word;
while (len) {
len--;
val = *buf++;
__asm__ volatile("crc32b %w0, %w0, %w1" : "+r"(crc) : "r"(val));
}
/* Return the CRC, post-conditioned. */
return crc ^ 0xffffffff;
}
#else
#ifdef W
/*
Return the CRC of the W bytes in the word_t data, taking the
least-significant byte of the word as the first byte of data, without any pre
or post conditioning. This is used to combine the CRCs of each braid.
*/
local z_crc_t crc_word(z_word_t data) {
int k;
for (k = 0; k < W; k++)
data = (data >> 8) ^ crc_table[data & 0xff];
return (z_crc_t)data;
}
local z_word_t crc_word_big(z_word_t data) {
int k;
for (k = 0; k < W; k++)
data = (data << 8) ^
crc_big_table[(data >> ((W - 1) << 3)) & 0xff];
return data;
}
#endif
/* ========================================================================= */
unsigned long ZEXPORT crc32_z(unsigned long crc, const unsigned char FAR *buf,
z_size_t len) {
/* Return initial CRC, if requested. */
if (buf == Z_NULL) return 0;
#ifdef DYNAMIC_CRC_TABLE
once(&made, make_crc_table);
#endif /* DYNAMIC_CRC_TABLE */
/* Pre-condition the CRC */
crc = (~crc) & 0xffffffff;
#ifdef W
/* If provided enough bytes, do a braided CRC calculation. */
if (len >= N * W + W - 1) {
z_size_t blks;
z_word_t const *words;
unsigned endian;
int k;
/* Compute the CRC up to a z_word_t boundary. */
while (len && ((z_size_t)buf & (W - 1)) != 0) {
len--;
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
}
/* Compute the CRC on as many N z_word_t blocks as are available. */
blks = len / (N * W);
len -= blks * N * W;
words = (z_word_t const *)buf;
/* Do endian check at execution time instead of compile time, since ARM
processors can change the endianness at execution time. If the
compiler knows what the endianness will be, it can optimize out the
check and the unused branch. */
endian = 1;
if (*(unsigned char *)&endian) {
/* Little endian. */
z_crc_t crc0;
z_word_t word0;
#if N > 1
z_crc_t crc1;
z_word_t word1;
#if N > 2
z_crc_t crc2;
z_word_t word2;
#if N > 3
z_crc_t crc3;
z_word_t word3;
#if N > 4
z_crc_t crc4;
z_word_t word4;
#if N > 5
z_crc_t crc5;
z_word_t word5;
#endif
#endif
#endif
#endif
#endif
/* Initialize the CRC for each braid. */
crc0 = crc;
#if N > 1
crc1 = 0;
#if N > 2
crc2 = 0;
#if N > 3
crc3 = 0;
#if N > 4
crc4 = 0;
#if N > 5
crc5 = 0;
#endif
#endif
#endif
#endif
#endif
/*
Process the first blks-1 blocks, computing the CRCs on each braid
independently.
*/
while (--blks) {
/* Load the word for each braid into registers. */
word0 = crc0 ^ words[0];
#if N > 1
word1 = crc1 ^ words[1];
#if N > 2
word2 = crc2 ^ words[2];
#if N > 3
word3 = crc3 ^ words[3];
#if N > 4
word4 = crc4 ^ words[4];
#if N > 5
word5 = crc5 ^ words[5];
#endif
#endif
#endif
#endif
#endif
words += N;
/* Compute and update the CRC for each word. The loop should
get unrolled. */
crc0 = crc_braid_table[0][word0 & 0xff];
#if N > 1
crc1 = crc_braid_table[0][word1 & 0xff];
#if N > 2
crc2 = crc_braid_table[0][word2 & 0xff];
#if N > 3
crc3 = crc_braid_table[0][word3 & 0xff];
#if N > 4
crc4 = crc_braid_table[0][word4 & 0xff];
#if N > 5
crc5 = crc_braid_table[0][word5 & 0xff];
#endif
#endif
#endif
#endif
#endif
for (k = 1; k < W; k++) {
crc0 ^= crc_braid_table[k][(word0 >> (k << 3)) & 0xff];
#if N > 1
crc1 ^= crc_braid_table[k][(word1 >> (k << 3)) & 0xff];
#if N > 2
crc2 ^= crc_braid_table[k][(word2 >> (k << 3)) & 0xff];
#if N > 3
crc3 ^= crc_braid_table[k][(word3 >> (k << 3)) & 0xff];
#if N > 4
crc4 ^= crc_braid_table[k][(word4 >> (k << 3)) & 0xff];
#if N > 5
crc5 ^= crc_braid_table[k][(word5 >> (k << 3)) & 0xff];
#endif
#endif
#endif
#endif
#endif
}
}
/*
Process the last block, combining the CRCs of the N braids at the
same time.
*/
crc = crc_word(crc0 ^ words[0]);
#if N > 1
crc = crc_word(crc1 ^ words[1] ^ crc);
#if N > 2
crc = crc_word(crc2 ^ words[2] ^ crc);
#if N > 3
crc = crc_word(crc3 ^ words[3] ^ crc);
#if N > 4
crc = crc_word(crc4 ^ words[4] ^ crc);
#if N > 5
crc = crc_word(crc5 ^ words[5] ^ crc);
#endif
#endif
#endif
#endif
#endif
words += N;
}
else {
/* Big endian. */
z_word_t crc0, word0, comb;
#if N > 1
z_word_t crc1, word1;
#if N > 2
z_word_t crc2, word2;
#if N > 3
z_word_t crc3, word3;
#if N > 4
z_word_t crc4, word4;
#if N > 5
z_word_t crc5, word5;
#endif
#endif
#endif
#endif
#endif
/* Initialize the CRC for each braid. */
crc0 = byte_swap(crc);
#if N > 1
crc1 = 0;
#if N > 2
crc2 = 0;
#if N > 3
crc3 = 0;
#if N > 4
crc4 = 0;
#if N > 5
crc5 = 0;
#endif
#endif
#endif
#endif
#endif
/*
Process the first blks-1 blocks, computing the CRCs on each braid
independently.
*/
while (--blks) {
/* Load the word for each braid into registers. */
word0 = crc0 ^ words[0];
#if N > 1
word1 = crc1 ^ words[1];
#if N > 2
word2 = crc2 ^ words[2];
#if N > 3
word3 = crc3 ^ words[3];
#if N > 4
word4 = crc4 ^ words[4];
#if N > 5
word5 = crc5 ^ words[5];
#endif
#endif
#endif
#endif
#endif
words += N;
/* Compute and update the CRC for each word. The loop should
get unrolled. */
crc0 = crc_braid_big_table[0][word0 & 0xff];
#if N > 1
crc1 = crc_braid_big_table[0][word1 & 0xff];
#if N > 2
crc2 = crc_braid_big_table[0][word2 & 0xff];
#if N > 3
crc3 = crc_braid_big_table[0][word3 & 0xff];
#if N > 4
crc4 = crc_braid_big_table[0][word4 & 0xff];
#if N > 5
crc5 = crc_braid_big_table[0][word5 & 0xff];
#endif
#endif
#endif
#endif
#endif
for (k = 1; k < W; k++) {
crc0 ^= crc_braid_big_table[k][(word0 >> (k << 3)) & 0xff];
#if N > 1
crc1 ^= crc_braid_big_table[k][(word1 >> (k << 3)) & 0xff];
#if N > 2
crc2 ^= crc_braid_big_table[k][(word2 >> (k << 3)) & 0xff];
#if N > 3
crc3 ^= crc_braid_big_table[k][(word3 >> (k << 3)) & 0xff];
#if N > 4
crc4 ^= crc_braid_big_table[k][(word4 >> (k << 3)) & 0xff];
#if N > 5
crc5 ^= crc_braid_big_table[k][(word5 >> (k << 3)) & 0xff];
#endif
#endif
#endif
#endif
#endif
}
}
/*
Process the last block, combining the CRCs of the N braids at the
same time.
*/
comb = crc_word_big(crc0 ^ words[0]);
#if N > 1
comb = crc_word_big(crc1 ^ words[1] ^ comb);
#if N > 2
comb = crc_word_big(crc2 ^ words[2] ^ comb);
#if N > 3
comb = crc_word_big(crc3 ^ words[3] ^ comb);
#if N > 4
comb = crc_word_big(crc4 ^ words[4] ^ comb);
#if N > 5
comb = crc_word_big(crc5 ^ words[5] ^ comb);
#endif
#endif
#endif
#endif
#endif
words += N;
crc = byte_swap(comb);
}
/*
Update the pointer to the remaining bytes to process.
*/
buf = (unsigned char const *)words;
}
#endif /* W */
/* Complete the computation of the CRC on any remaining bytes. */
while (len >= 8) {
len -= 8;
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];
crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff];