1 | /* crc32.c -- compute the CRC-32 of a data stream |
2 | * Copyright (C) 1995-2022 Mark Adler |
3 | * For conditions of distribution and use, see copyright notice in zlib.h |
4 | * |
5 | * This interleaved implementation of a CRC makes use of pipelined multiple |
6 | * arithmetic-logic units, commonly found in modern CPU cores. It is due to |
7 | * Kadatch and Jenkins (2010). See doc/crc-doc.1.0.pdf in this distribution. |
8 | */ |
9 | |
10 | /* @(#) $Id$ */ |
11 | |
12 | /* |
13 | Note on the use of DYNAMIC_CRC_TABLE: there is no mutex or semaphore |
14 | protection on the static variables used to control the first-use generation |
15 | of the crc tables. Therefore, if you #define DYNAMIC_CRC_TABLE, you should |
16 | first call get_crc_table() to initialize the tables before allowing more than |
17 | one thread to use crc32(). |
18 | |
19 | MAKECRCH can be #defined to write out crc32.h. A main() routine is also |
20 | produced, so that this one source file can be compiled to an executable. |
21 | */ |
22 | |
23 | #ifdef MAKECRCH |
24 | # include <stdio.h> |
25 | # ifndef DYNAMIC_CRC_TABLE |
26 | # define DYNAMIC_CRC_TABLE |
27 | # endif /* !DYNAMIC_CRC_TABLE */ |
28 | #endif /* MAKECRCH */ |
29 | |
30 | #include "zutil.h" /* for Z_U4, Z_U8, z_crc_t, and FAR definitions */ |
31 | |
32 | /* |
33 | A CRC of a message is computed on N braids of words in the message, where |
34 | each word consists of W bytes (4 or 8). If N is 3, for example, then three |
35 | running sparse CRCs are calculated respectively on each braid, at these |
36 | indices in the array of words: 0, 3, 6, ..., 1, 4, 7, ..., and 2, 5, 8, ... |
37 | This is done starting at a word boundary, and continues until as many blocks |
38 | of N * W bytes as are available have been processed. The results are combined |
39 | into a single CRC at the end. For this code, N must be in the range 1..6 and |
40 | W must be 4 or 8. The upper limit on N can be increased if desired by adding |
41 | more #if blocks, extending the patterns apparent in the code. In addition, |
42 | crc32.h would need to be regenerated, if the maximum N value is increased. |
43 | |
44 | N and W are chosen empirically by benchmarking the execution time on a given |
45 | processor. The choices for N and W below were based on testing on Intel Kaby |
46 | Lake i7, AMD Ryzen 7, ARM Cortex-A57, Sparc64-VII, PowerPC POWER9, and MIPS64 |
47 | Octeon II processors. The Intel, AMD, and ARM processors were all fastest |
48 | with N=5, W=8. The Sparc, PowerPC, and MIPS64 were all fastest at N=5, W=4. |
49 | They were all tested with either gcc or clang, all using the -O3 optimization |
50 | level. Your mileage may vary. |
51 | */ |
52 | |
53 | /* Define N */ |
54 | #ifdef Z_TESTN |
55 | # define N Z_TESTN |
56 | #else |
57 | # define N 5 |
58 | #endif |
59 | #if N < 1 || N > 6 |
60 | # error N must be in 1..6 |
61 | #endif |
62 | |
63 | /* |
64 | z_crc_t must be at least 32 bits. z_word_t must be at least as long as |
65 | z_crc_t. It is assumed here that z_word_t is either 32 bits or 64 bits, and |
66 | that bytes are eight bits. |
67 | */ |
68 | |
69 | /* |
70 | Define W and the associated z_word_t type. If W is not defined, then a |
71 | braided calculation is not used, and the associated tables and code are not |
72 | compiled. |
73 | */ |
74 | #ifdef Z_TESTW |
75 | # if Z_TESTW-1 != -1 |
76 | # define W Z_TESTW |
77 | # endif |
78 | #else |
79 | # ifdef MAKECRCH |
80 | # define W 8 /* required for MAKECRCH */ |
81 | # else |
82 | # if defined(__x86_64__) || defined(__aarch64__) |
83 | # define W 8 |
84 | # else |
85 | # define W 4 |
86 | # endif |
87 | # endif |
88 | #endif |
89 | #ifdef W |
90 | # if W == 8 && defined(Z_U8) |
91 | typedef Z_U8 z_word_t; |
92 | # elif defined(Z_U4) |
93 | # undef W |
94 | # define W 4 |
95 | typedef Z_U4 z_word_t; |
96 | # else |
97 | # undef W |
98 | # endif |
99 | #endif |
100 | |
101 | /* Local functions. */ |
102 | local z_crc_t multmodp OF((z_crc_t a, z_crc_t b)); |
103 | local z_crc_t x2nmodp OF((z_off64_t n, unsigned k)); |
104 | |
105 | /* If available, use the ARM processor CRC32 instruction. */ |
106 | #if defined(__aarch64__) && defined(__ARM_FEATURE_CRC32) && W == 8 |
107 | # define ARMCRC32 |
108 | #endif |
109 | |
110 | #if defined(W) && (!defined(ARMCRC32) || defined(DYNAMIC_CRC_TABLE)) |
111 | /* |
112 | Swap the bytes in a z_word_t to convert between little and big endian. Any |
113 | self-respecting compiler will optimize this to a single machine byte-swap |
114 | instruction, if one is available. This assumes that word_t is either 32 bits |
115 | or 64 bits. |
116 | */ |
117 | local z_word_t byte_swap(word) |
118 | z_word_t word; |
119 | { |
120 | # if W == 8 |
121 | return |
122 | (word & 0xff00000000000000) >> 56 | |
123 | (word & 0xff000000000000) >> 40 | |
124 | (word & 0xff0000000000) >> 24 | |
125 | (word & 0xff00000000) >> 8 | |
126 | (word & 0xff000000) << 8 | |
127 | (word & 0xff0000) << 24 | |
128 | (word & 0xff00) << 40 | |
129 | (word & 0xff) << 56; |
130 | # else /* W == 4 */ |
131 | return |
132 | (word & 0xff000000) >> 24 | |
133 | (word & 0xff0000) >> 8 | |
134 | (word & 0xff00) << 8 | |
135 | (word & 0xff) << 24; |
136 | # endif |
137 | } |
138 | #endif |
139 | |
140 | /* CRC polynomial. */ |
141 | #define POLY 0xedb88320 /* p(x) reflected, with x^32 implied */ |
142 | |
143 | #ifdef DYNAMIC_CRC_TABLE |
144 | |
145 | local z_crc_t FAR crc_table[256]; |
146 | local z_crc_t FAR x2n_table[32]; |
147 | local void make_crc_table OF((void)); |
148 | #ifdef W |
149 | local z_word_t FAR crc_big_table[256]; |
150 | local z_crc_t FAR crc_braid_table[W][256]; |
151 | local z_word_t FAR crc_braid_big_table[W][256]; |
152 | local void braid OF((z_crc_t [][256], z_word_t [][256], int, int)); |
153 | #endif |
154 | #ifdef MAKECRCH |
155 | local void write_table OF((FILE *, const z_crc_t FAR *, int)); |
156 | local void write_table32hi OF((FILE *, const z_word_t FAR *, int)); |
157 | local void write_table64 OF((FILE *, const z_word_t FAR *, int)); |
158 | #endif /* MAKECRCH */ |
159 | |
160 | /* |
161 | Define a once() function depending on the availability of atomics. If this is |
162 | compiled with DYNAMIC_CRC_TABLE defined, and if CRCs will be computed in |
163 | multiple threads, and if atomics are not available, then get_crc_table() must |
164 | be called to initialize the tables and must return before any threads are |
165 | allowed to compute or combine CRCs. |
166 | */ |
167 | |
168 | /* Definition of once functionality. */ |
169 | typedef struct once_s once_t; |
170 | local void once OF((once_t *, void (*)(void))); |
171 | |
172 | /* Check for the availability of atomics. */ |
173 | #if defined(__STDC__) && __STDC_VERSION__ >= 201112L && \ |
174 | !defined(__STDC_NO_ATOMICS__) |
175 | |
176 | #include <stdatomic.h> |
177 | |
178 | /* Structure for once(), which must be initialized with ONCE_INIT. */ |
179 | struct once_s { |
180 | atomic_flag begun; |
181 | atomic_int done; |
182 | }; |
183 | #define ONCE_INIT {ATOMIC_FLAG_INIT, 0} |
184 | |
185 | /* |
186 | Run the provided init() function exactly once, even if multiple threads |
187 | invoke once() at the same time. The state must be a once_t initialized with |
188 | ONCE_INIT. |
189 | */ |
190 | local void once(state, init) |
191 | once_t *state; |
192 | void (*init)(void); |
193 | { |
194 | if (!atomic_load(&state->done)) { |
195 | if (atomic_flag_test_and_set(&state->begun)) |
196 | while (!atomic_load(&state->done)) |
197 | ; |
198 | else { |
199 | init(); |
200 | atomic_store(&state->done, 1); |
201 | } |
202 | } |
203 | } |
204 | |
205 | #else /* no atomics */ |
206 | |
207 | /* Structure for once(), which must be initialized with ONCE_INIT. */ |
208 | struct once_s { |
209 | volatile int begun; |
210 | volatile int done; |
211 | }; |
212 | #define ONCE_INIT {0, 0} |
213 | |
214 | /* Test and set. Alas, not atomic, but tries to minimize the period of |
215 | vulnerability. */ |
216 | local int test_and_set OF((int volatile *)); |
217 | local int test_and_set(flag) |
218 | int volatile *flag; |
219 | { |
220 | int was; |
221 | |
222 | was = *flag; |
223 | *flag = 1; |
224 | return was; |
225 | } |
226 | |
227 | /* Run the provided init() function once. This is not thread-safe. */ |
228 | local void once(state, init) |
229 | once_t *state; |
230 | void (*init)(void); |
231 | { |
232 | if (!state->done) { |
233 | if (test_and_set(&state->begun)) |
234 | while (!state->done) |
235 | ; |
236 | else { |
237 | init(); |
238 | state->done = 1; |
239 | } |
240 | } |
241 | } |
242 | |
243 | #endif |
244 | |
245 | /* State for once(). */ |
246 | local once_t made = ONCE_INIT; |
247 | |
248 | /* |
249 | Generate tables for a byte-wise 32-bit CRC calculation on the polynomial: |
250 | 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. |
251 | |
252 | Polynomials over GF(2) are represented in binary, one bit per coefficient, |
253 | with the lowest powers in the most significant bit. Then adding polynomials |
254 | is just exclusive-or, and multiplying a polynomial by x is a right shift by |
255 | one. If we call the above polynomial p, and represent a byte as the |
256 | polynomial q, also with the lowest power in the most significant bit (so the |
257 | byte 0xb1 is the polynomial x^7+x^3+x^2+1), then the CRC is (q*x^32) mod p, |
258 | where a mod b means the remainder after dividing a by b. |
259 | |
260 | This calculation is done using the shift-register method of multiplying and |
261 | taking the remainder. The register is initialized to zero, and for each |
262 | incoming bit, x^32 is added mod p to the register if the bit is a one (where |
263 | x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by x |
264 | (which is shifting right by one and adding x^32 mod p if the bit shifted out |
265 | is a one). We start with the highest power (least significant bit) of q and |
266 | repeat for all eight bits of q. |
267 | |
268 | The table is simply the CRC of all possible eight bit values. This is all the |
269 | information needed to generate CRCs on data a byte at a time for all |
270 | combinations of CRC register values and incoming bytes. |
271 | */ |
272 | |
273 | local void make_crc_table() |
274 | { |
275 | unsigned i, j, n; |
276 | z_crc_t p; |
277 | |
278 | /* initialize the CRC of bytes tables */ |
279 | for (i = 0; i < 256; i++) { |
280 | p = i; |
281 | for (j = 0; j < 8; j++) |
282 | p = p & 1 ? (p >> 1) ^ POLY : p >> 1; |
283 | crc_table[i] = p; |
284 | #ifdef W |
285 | crc_big_table[i] = byte_swap(p); |
286 | #endif |
287 | } |
288 | |
289 | /* initialize the x^2^n mod p(x) table */ |
290 | p = (z_crc_t)1 << 30; /* x^1 */ |
291 | x2n_table[0] = p; |
292 | for (n = 1; n < 32; n++) |
293 | x2n_table[n] = p = multmodp(p, p); |
294 | |
295 | #ifdef W |
296 | /* initialize the braiding tables -- needs x2n_table[] */ |
297 | braid(crc_braid_table, crc_braid_big_table, N, W); |
298 | #endif |
299 | |
300 | #ifdef MAKECRCH |
301 | { |
302 | /* |
303 | The crc32.h header file contains tables for both 32-bit and 64-bit |
304 | z_word_t's, and so requires a 64-bit type be available. In that case, |
305 | z_word_t must be defined to be 64-bits. This code then also generates |
306 | and writes out the tables for the case that z_word_t is 32 bits. |
307 | */ |
308 | #if !defined(W) || W != 8 |
309 | # error Need a 64-bit integer type in order to generate crc32.h. |
310 | #endif |
311 | FILE *out; |
312 | int k, n; |
313 | z_crc_t ltl[8][256]; |
314 | z_word_t big[8][256]; |
315 | |
316 | out = fopen("crc32.h" , "w" ); |
317 | if (out == NULL) return; |
318 | |
319 | /* write out little-endian CRC table to crc32.h */ |
320 | fprintf(out, |
321 | "/* crc32.h -- tables for rapid CRC calculation\n" |
322 | " * Generated automatically by crc32.c\n */\n" |
323 | "\n" |
324 | "local const z_crc_t FAR crc_table[] = {\n" |
325 | " " ); |
326 | write_table(out, crc_table, 256); |
327 | fprintf(out, |
328 | "};\n" ); |
329 | |
330 | /* write out big-endian CRC table for 64-bit z_word_t to crc32.h */ |
331 | fprintf(out, |
332 | "\n" |
333 | "#ifdef W\n" |
334 | "\n" |
335 | "#if W == 8\n" |
336 | "\n" |
337 | "local const z_word_t FAR crc_big_table[] = {\n" |
338 | " " ); |
339 | write_table64(out, crc_big_table, 256); |
340 | fprintf(out, |
341 | "};\n" ); |
342 | |
343 | /* write out big-endian CRC table for 32-bit z_word_t to crc32.h */ |
344 | fprintf(out, |
345 | "\n" |
346 | "#else /* W == 4 */\n" |
347 | "\n" |
348 | "local const z_word_t FAR crc_big_table[] = {\n" |
349 | " " ); |
350 | write_table32hi(out, crc_big_table, 256); |
351 | fprintf(out, |
352 | "};\n" |
353 | "\n" |
354 | "#endif\n" ); |
355 | |
356 | /* write out braid tables for each value of N */ |
357 | for (n = 1; n <= 6; n++) { |
358 | fprintf(out, |
359 | "\n" |
360 | "#if N == %d\n" , n); |
361 | |
362 | /* compute braid tables for this N and 64-bit word_t */ |
363 | braid(ltl, big, n, 8); |
364 | |
365 | /* write out braid tables for 64-bit z_word_t to crc32.h */ |
366 | fprintf(out, |
367 | "\n" |
368 | "#if W == 8\n" |
369 | "\n" |
370 | "local const z_crc_t FAR crc_braid_table[][256] = {\n" ); |
371 | for (k = 0; k < 8; k++) { |
372 | fprintf(out, " {" ); |
373 | write_table(out, ltl[k], 256); |
374 | fprintf(out, "}%s" , k < 7 ? ",\n" : "" ); |
375 | } |
376 | fprintf(out, |
377 | "};\n" |
378 | "\n" |
379 | "local const z_word_t FAR crc_braid_big_table[][256] = {\n" ); |
380 | for (k = 0; k < 8; k++) { |
381 | fprintf(out, " {" ); |
382 | write_table64(out, big[k], 256); |
383 | fprintf(out, "}%s" , k < 7 ? ",\n" : "" ); |
384 | } |
385 | fprintf(out, |
386 | "};\n" ); |
387 | |
388 | /* compute braid tables for this N and 32-bit word_t */ |
389 | braid(ltl, big, n, 4); |
390 | |
391 | /* write out braid tables for 32-bit z_word_t to crc32.h */ |
392 | fprintf(out, |
393 | "\n" |
394 | "#else /* W == 4 */\n" |
395 | "\n" |
396 | "local const z_crc_t FAR crc_braid_table[][256] = {\n" ); |
397 | for (k = 0; k < 4; k++) { |
398 | fprintf(out, " {" ); |
399 | write_table(out, ltl[k], 256); |
400 | fprintf(out, "}%s" , k < 3 ? ",\n" : "" ); |
401 | } |
402 | fprintf(out, |
403 | "};\n" |
404 | "\n" |
405 | "local const z_word_t FAR crc_braid_big_table[][256] = {\n" ); |
406 | for (k = 0; k < 4; k++) { |
407 | fprintf(out, " {" ); |
408 | write_table32hi(out, big[k], 256); |
409 | fprintf(out, "}%s" , k < 3 ? ",\n" : "" ); |
410 | } |
411 | fprintf(out, |
412 | "};\n" |
413 | "\n" |
414 | "#endif\n" |
415 | "\n" |
416 | "#endif\n" ); |
417 | } |
418 | fprintf(out, |
419 | "\n" |
420 | "#endif\n" ); |
421 | |
422 | /* write out zeros operator table to crc32.h */ |
423 | fprintf(out, |
424 | "\n" |
425 | "local const z_crc_t FAR x2n_table[] = {\n" |
426 | " " ); |
427 | write_table(out, x2n_table, 32); |
428 | fprintf(out, |
429 | "};\n" ); |
430 | fclose(out); |
431 | } |
432 | #endif /* MAKECRCH */ |
433 | } |
434 | |
435 | #ifdef MAKECRCH |
436 | |
437 | /* |
438 | Write the 32-bit values in table[0..k-1] to out, five per line in |
439 | hexadecimal separated by commas. |
440 | */ |
441 | local void write_table(out, table, k) |
442 | FILE *out; |
443 | const z_crc_t FAR *table; |
444 | int k; |
445 | { |
446 | int n; |
447 | |
448 | for (n = 0; n < k; n++) |
449 | fprintf(out, "%s0x%08lx%s" , n == 0 || n % 5 ? "" : " " , |
450 | (unsigned long)(table[n]), |
451 | n == k - 1 ? "" : (n % 5 == 4 ? ",\n" : ", " )); |
452 | } |
453 | |
454 | /* |
455 | Write the high 32-bits of each value in table[0..k-1] to out, five per line |
456 | in hexadecimal separated by commas. |
457 | */ |
458 | local void write_table32hi(out, table, k) |
459 | FILE *out; |
460 | const z_word_t FAR *table; |
461 | int k; |
462 | { |
463 | int n; |
464 | |
465 | for (n = 0; n < k; n++) |
466 | fprintf(out, "%s0x%08lx%s" , n == 0 || n % 5 ? "" : " " , |
467 | (unsigned long)(table[n] >> 32), |
468 | n == k - 1 ? "" : (n % 5 == 4 ? ",\n" : ", " )); |
469 | } |
470 | |
471 | /* |
472 | Write the 64-bit values in table[0..k-1] to out, three per line in |
473 | hexadecimal separated by commas. This assumes that if there is a 64-bit |
474 | type, then there is also a long long integer type, and it is at least 64 |
475 | bits. If not, then the type cast and format string can be adjusted |
476 | accordingly. |
477 | */ |
478 | local void write_table64(out, table, k) |
479 | FILE *out; |
480 | const z_word_t FAR *table; |
481 | int k; |
482 | { |
483 | int n; |
484 | |
485 | for (n = 0; n < k; n++) |
486 | fprintf(out, "%s0x%016llx%s" , n == 0 || n % 3 ? "" : " " , |
487 | (unsigned long long)(table[n]), |
488 | n == k - 1 ? "" : (n % 3 == 2 ? ",\n" : ", " )); |
489 | } |
490 | |
491 | /* Actually do the deed. */ |
492 | int main() |
493 | { |
494 | make_crc_table(); |
495 | return 0; |
496 | } |
497 | |
498 | #endif /* MAKECRCH */ |
499 | |
500 | #ifdef W |
501 | /* |
502 | Generate the little and big-endian braid tables for the given n and z_word_t |
503 | size w. Each array must have room for w blocks of 256 elements. |
504 | */ |
505 | local void braid(ltl, big, n, w) |
506 | z_crc_t ltl[][256]; |
507 | z_word_t big[][256]; |
508 | int n; |
509 | int w; |
510 | { |
511 | int k; |
512 | z_crc_t i, p, q; |
513 | for (k = 0; k < w; k++) { |
514 | p = x2nmodp((n * w + 3 - k) << 3, 0); |
515 | ltl[k][0] = 0; |
516 | big[w - 1 - k][0] = 0; |
517 | for (i = 1; i < 256; i++) { |
518 | ltl[k][i] = q = multmodp(i << 24, p); |
519 | big[w - 1 - k][i] = byte_swap(q); |
520 | } |
521 | } |
522 | } |
523 | #endif |
524 | |
525 | #else /* !DYNAMIC_CRC_TABLE */ |
526 | /* ======================================================================== |
527 | * Tables for byte-wise and braided CRC-32 calculations, and a table of powers |
528 | * of x for combining CRC-32s, all made by make_crc_table(). |
529 | */ |
530 | #include "crc32.h" |
531 | #endif /* DYNAMIC_CRC_TABLE */ |
532 | |
533 | /* ======================================================================== |
534 | * Routines used for CRC calculation. Some are also required for the table |
535 | * generation above. |
536 | */ |
537 | |
538 | /* |
539 | Return a(x) multiplied by b(x) modulo p(x), where p(x) is the CRC polynomial, |
540 | reflected. For speed, this requires that a not be zero. |
541 | */ |
542 | local z_crc_t multmodp(a, b) |
543 | z_crc_t a; |
544 | z_crc_t b; |
545 | { |
546 | z_crc_t m, p; |
547 | |
548 | m = (z_crc_t)1 << 31; |
549 | p = 0; |
550 | for (;;) { |
551 | if (a & m) { |
552 | p ^= b; |
553 | if ((a & (m - 1)) == 0) |
554 | break; |
555 | } |
556 | m >>= 1; |
557 | b = b & 1 ? (b >> 1) ^ POLY : b >> 1; |
558 | } |
559 | return p; |
560 | } |
561 | |
562 | /* |
563 | Return x^(n * 2^k) modulo p(x). Requires that x2n_table[] has been |
564 | initialized. |
565 | */ |
566 | local z_crc_t x2nmodp(n, k) |
567 | z_off64_t n; |
568 | unsigned k; |
569 | { |
570 | z_crc_t p; |
571 | |
572 | p = (z_crc_t)1 << 31; /* x^0 == 1 */ |
573 | while (n) { |
574 | if (n & 1) |
575 | p = multmodp(x2n_table[k & 31], p); |
576 | n >>= 1; |
577 | k++; |
578 | } |
579 | return p; |
580 | } |
581 | |
582 | /* ========================================================================= |
583 | * This function can be used by asm versions of crc32(), and to force the |
584 | * generation of the CRC tables in a threaded application. |
585 | */ |
586 | const z_crc_t FAR * ZEXPORT get_crc_table() |
587 | { |
588 | #ifdef DYNAMIC_CRC_TABLE |
589 | once(&made, make_crc_table); |
590 | #endif /* DYNAMIC_CRC_TABLE */ |
591 | return (const z_crc_t FAR *)crc_table; |
592 | } |
593 | |
594 | /* ========================================================================= |
595 | * Use ARM machine instructions if available. This will compute the CRC about |
596 | * ten times faster than the braided calculation. This code does not check for |
597 | * the presence of the CRC instruction at run time. __ARM_FEATURE_CRC32 will |
598 | * only be defined if the compilation specifies an ARM processor architecture |
599 | * that has the instructions. For example, compiling with -march=armv8.1-a or |
600 | * -march=armv8-a+crc, or -march=native if the compile machine has the crc32 |
601 | * instructions. |
602 | */ |
603 | #ifdef ARMCRC32 |
604 | |
605 | /* |
606 | Constants empirically determined to maximize speed. These values are from |
607 | measurements on a Cortex-A57. Your mileage may vary. |
608 | */ |
609 | #define Z_BATCH 3990 /* number of words in a batch */ |
610 | #define Z_BATCH_ZEROS 0xa10d3d0c /* computed from Z_BATCH = 3990 */ |
611 | #define Z_BATCH_MIN 800 /* fewest words in a final batch */ |
612 | |
613 | unsigned long ZEXPORT crc32_z(crc, buf, len) |
614 | unsigned long crc; |
615 | const unsigned char FAR *buf; |
616 | z_size_t len; |
617 | { |
618 | z_crc_t val; |
619 | z_word_t crc1, crc2; |
620 | const z_word_t *word; |
621 | z_word_t val0, val1, val2; |
622 | z_size_t last, last2, i; |
623 | z_size_t num; |
624 | |
625 | /* Return initial CRC, if requested. */ |
626 | if (buf == Z_NULL) return 0; |
627 | |
628 | #ifdef DYNAMIC_CRC_TABLE |
629 | once(&made, make_crc_table); |
630 | #endif /* DYNAMIC_CRC_TABLE */ |
631 | |
632 | /* Pre-condition the CRC */ |
633 | crc ^= 0xffffffff; |
634 | |
635 | /* Compute the CRC up to a word boundary. */ |
636 | while (len && ((z_size_t)buf & 7) != 0) { |
637 | len--; |
638 | val = *buf++; |
639 | __asm__ volatile("crc32b %w0, %w0, %w1" : "+r" (crc) : "r" (val)); |
640 | } |
641 | |
642 | /* Prepare to compute the CRC on full 64-bit words word[0..num-1]. */ |
643 | word = (z_word_t const *)buf; |
644 | num = len >> 3; |
645 | len &= 7; |
646 | |
647 | /* Do three interleaved CRCs to realize the throughput of one crc32x |
648 | instruction per cycle. Each CRC is calcuated on Z_BATCH words. The three |
649 | CRCs are combined into a single CRC after each set of batches. */ |
650 | while (num >= 3 * Z_BATCH) { |
651 | crc1 = 0; |
652 | crc2 = 0; |
653 | for (i = 0; i < Z_BATCH; i++) { |
654 | val0 = word[i]; |
655 | val1 = word[i + Z_BATCH]; |
656 | val2 = word[i + 2 * Z_BATCH]; |
657 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc) : "r" (val0)); |
658 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc1) : "r" (val1)); |
659 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc2) : "r" (val2)); |
660 | } |
661 | word += 3 * Z_BATCH; |
662 | num -= 3 * Z_BATCH; |
663 | crc = multmodp(Z_BATCH_ZEROS, crc) ^ crc1; |
664 | crc = multmodp(Z_BATCH_ZEROS, crc) ^ crc2; |
665 | } |
666 | |
667 | /* Do one last smaller batch with the remaining words, if there are enough |
668 | to pay for the combination of CRCs. */ |
669 | last = num / 3; |
670 | if (last >= Z_BATCH_MIN) { |
671 | last2 = last << 1; |
672 | crc1 = 0; |
673 | crc2 = 0; |
674 | for (i = 0; i < last; i++) { |
675 | val0 = word[i]; |
676 | val1 = word[i + last]; |
677 | val2 = word[i + last2]; |
678 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc) : "r" (val0)); |
679 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc1) : "r" (val1)); |
680 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc2) : "r" (val2)); |
681 | } |
682 | word += 3 * last; |
683 | num -= 3 * last; |
684 | val = x2nmodp(last, 6); |
685 | crc = multmodp(val, crc) ^ crc1; |
686 | crc = multmodp(val, crc) ^ crc2; |
687 | } |
688 | |
689 | /* Compute the CRC on any remaining words. */ |
690 | for (i = 0; i < num; i++) { |
691 | val0 = word[i]; |
692 | __asm__ volatile("crc32x %w0, %w0, %x1" : "+r" (crc) : "r" (val0)); |
693 | } |
694 | word += num; |
695 | |
696 | /* Complete the CRC on any remaining bytes. */ |
697 | buf = (const unsigned char FAR *)word; |
698 | while (len) { |
699 | len--; |
700 | val = *buf++; |
701 | __asm__ volatile("crc32b %w0, %w0, %w1" : "+r" (crc) : "r" (val)); |
702 | } |
703 | |
704 | /* Return the CRC, post-conditioned. */ |
705 | return crc ^ 0xffffffff; |
706 | } |
707 | |
708 | #else |
709 | |
710 | #ifdef W |
711 | |
712 | /* |
713 | Return the CRC of the W bytes in the word_t data, taking the |
714 | least-significant byte of the word as the first byte of data, without any pre |
715 | or post conditioning. This is used to combine the CRCs of each braid. |
716 | */ |
717 | local z_crc_t crc_word(data) |
718 | z_word_t data; |
719 | { |
720 | int k; |
721 | for (k = 0; k < W; k++) |
722 | data = (data >> 8) ^ crc_table[data & 0xff]; |
723 | return (z_crc_t)data; |
724 | } |
725 | |
726 | local z_word_t crc_word_big(data) |
727 | z_word_t data; |
728 | { |
729 | int k; |
730 | for (k = 0; k < W; k++) |
731 | data = (data << 8) ^ |
732 | crc_big_table[(data >> ((W - 1) << 3)) & 0xff]; |
733 | return data; |
734 | } |
735 | |
736 | #endif |
737 | |
738 | /* ========================================================================= */ |
739 | unsigned long ZEXPORT crc32_z(crc, buf, len) |
740 | unsigned long crc; |
741 | const unsigned char FAR *buf; |
742 | z_size_t len; |
743 | { |
744 | /* Return initial CRC, if requested. */ |
745 | if (buf == Z_NULL) return 0; |
746 | |
747 | #ifdef DYNAMIC_CRC_TABLE |
748 | once(&made, make_crc_table); |
749 | #endif /* DYNAMIC_CRC_TABLE */ |
750 | |
751 | /* Pre-condition the CRC */ |
752 | crc ^= 0xffffffff; |
753 | |
754 | #ifdef W |
755 | |
756 | /* If provided enough bytes, do a braided CRC calculation. */ |
757 | if (len >= N * W + W - 1) { |
758 | z_size_t blks; |
759 | z_word_t const *words; |
760 | unsigned endian; |
761 | int k; |
762 | |
763 | /* Compute the CRC up to a z_word_t boundary. */ |
764 | while (len && ((z_size_t)buf & (W - 1)) != 0) { |
765 | len--; |
766 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
767 | } |
768 | |
769 | /* Compute the CRC on as many N z_word_t blocks as are available. */ |
770 | blks = len / (N * W); |
771 | len -= blks * N * W; |
772 | words = (z_word_t const *)buf; |
773 | |
774 | /* Do endian check at execution time instead of compile time, since ARM |
775 | processors can change the endianess at execution time. If the |
776 | compiler knows what the endianess will be, it can optimize out the |
777 | check and the unused branch. */ |
778 | endian = 1; |
779 | if (*(unsigned char *)&endian) { |
780 | /* Little endian. */ |
781 | |
782 | z_crc_t crc0; |
783 | z_word_t word0; |
784 | #if N > 1 |
785 | z_crc_t crc1; |
786 | z_word_t word1; |
787 | #if N > 2 |
788 | z_crc_t crc2; |
789 | z_word_t word2; |
790 | #if N > 3 |
791 | z_crc_t crc3; |
792 | z_word_t word3; |
793 | #if N > 4 |
794 | z_crc_t crc4; |
795 | z_word_t word4; |
796 | #if N > 5 |
797 | z_crc_t crc5; |
798 | z_word_t word5; |
799 | #endif |
800 | #endif |
801 | #endif |
802 | #endif |
803 | #endif |
804 | |
805 | /* Initialize the CRC for each braid. */ |
806 | crc0 = crc; |
807 | #if N > 1 |
808 | crc1 = 0; |
809 | #if N > 2 |
810 | crc2 = 0; |
811 | #if N > 3 |
812 | crc3 = 0; |
813 | #if N > 4 |
814 | crc4 = 0; |
815 | #if N > 5 |
816 | crc5 = 0; |
817 | #endif |
818 | #endif |
819 | #endif |
820 | #endif |
821 | #endif |
822 | |
823 | /* |
824 | Process the first blks-1 blocks, computing the CRCs on each braid |
825 | independently. |
826 | */ |
827 | while (--blks) { |
828 | /* Load the word for each braid into registers. */ |
829 | word0 = crc0 ^ words[0]; |
830 | #if N > 1 |
831 | word1 = crc1 ^ words[1]; |
832 | #if N > 2 |
833 | word2 = crc2 ^ words[2]; |
834 | #if N > 3 |
835 | word3 = crc3 ^ words[3]; |
836 | #if N > 4 |
837 | word4 = crc4 ^ words[4]; |
838 | #if N > 5 |
839 | word5 = crc5 ^ words[5]; |
840 | #endif |
841 | #endif |
842 | #endif |
843 | #endif |
844 | #endif |
845 | words += N; |
846 | |
847 | /* Compute and update the CRC for each word. The loop should |
848 | get unrolled. */ |
849 | crc0 = crc_braid_table[0][word0 & 0xff]; |
850 | #if N > 1 |
851 | crc1 = crc_braid_table[0][word1 & 0xff]; |
852 | #if N > 2 |
853 | crc2 = crc_braid_table[0][word2 & 0xff]; |
854 | #if N > 3 |
855 | crc3 = crc_braid_table[0][word3 & 0xff]; |
856 | #if N > 4 |
857 | crc4 = crc_braid_table[0][word4 & 0xff]; |
858 | #if N > 5 |
859 | crc5 = crc_braid_table[0][word5 & 0xff]; |
860 | #endif |
861 | #endif |
862 | #endif |
863 | #endif |
864 | #endif |
865 | for (k = 1; k < W; k++) { |
866 | crc0 ^= crc_braid_table[k][(word0 >> (k << 3)) & 0xff]; |
867 | #if N > 1 |
868 | crc1 ^= crc_braid_table[k][(word1 >> (k << 3)) & 0xff]; |
869 | #if N > 2 |
870 | crc2 ^= crc_braid_table[k][(word2 >> (k << 3)) & 0xff]; |
871 | #if N > 3 |
872 | crc3 ^= crc_braid_table[k][(word3 >> (k << 3)) & 0xff]; |
873 | #if N > 4 |
874 | crc4 ^= crc_braid_table[k][(word4 >> (k << 3)) & 0xff]; |
875 | #if N > 5 |
876 | crc5 ^= crc_braid_table[k][(word5 >> (k << 3)) & 0xff]; |
877 | #endif |
878 | #endif |
879 | #endif |
880 | #endif |
881 | #endif |
882 | } |
883 | } |
884 | |
885 | /* |
886 | Process the last block, combining the CRCs of the N braids at the |
887 | same time. |
888 | */ |
889 | crc = crc_word(crc0 ^ words[0]); |
890 | #if N > 1 |
891 | crc = crc_word(crc1 ^ words[1] ^ crc); |
892 | #if N > 2 |
893 | crc = crc_word(crc2 ^ words[2] ^ crc); |
894 | #if N > 3 |
895 | crc = crc_word(crc3 ^ words[3] ^ crc); |
896 | #if N > 4 |
897 | crc = crc_word(crc4 ^ words[4] ^ crc); |
898 | #if N > 5 |
899 | crc = crc_word(crc5 ^ words[5] ^ crc); |
900 | #endif |
901 | #endif |
902 | #endif |
903 | #endif |
904 | #endif |
905 | words += N; |
906 | } |
907 | else { |
908 | /* Big endian. */ |
909 | |
910 | z_word_t crc0, word0, comb; |
911 | #if N > 1 |
912 | z_word_t crc1, word1; |
913 | #if N > 2 |
914 | z_word_t crc2, word2; |
915 | #if N > 3 |
916 | z_word_t crc3, word3; |
917 | #if N > 4 |
918 | z_word_t crc4, word4; |
919 | #if N > 5 |
920 | z_word_t crc5, word5; |
921 | #endif |
922 | #endif |
923 | #endif |
924 | #endif |
925 | #endif |
926 | |
927 | /* Initialize the CRC for each braid. */ |
928 | crc0 = byte_swap(crc); |
929 | #if N > 1 |
930 | crc1 = 0; |
931 | #if N > 2 |
932 | crc2 = 0; |
933 | #if N > 3 |
934 | crc3 = 0; |
935 | #if N > 4 |
936 | crc4 = 0; |
937 | #if N > 5 |
938 | crc5 = 0; |
939 | #endif |
940 | #endif |
941 | #endif |
942 | #endif |
943 | #endif |
944 | |
945 | /* |
946 | Process the first blks-1 blocks, computing the CRCs on each braid |
947 | independently. |
948 | */ |
949 | while (--blks) { |
950 | /* Load the word for each braid into registers. */ |
951 | word0 = crc0 ^ words[0]; |
952 | #if N > 1 |
953 | word1 = crc1 ^ words[1]; |
954 | #if N > 2 |
955 | word2 = crc2 ^ words[2]; |
956 | #if N > 3 |
957 | word3 = crc3 ^ words[3]; |
958 | #if N > 4 |
959 | word4 = crc4 ^ words[4]; |
960 | #if N > 5 |
961 | word5 = crc5 ^ words[5]; |
962 | #endif |
963 | #endif |
964 | #endif |
965 | #endif |
966 | #endif |
967 | words += N; |
968 | |
969 | /* Compute and update the CRC for each word. The loop should |
970 | get unrolled. */ |
971 | crc0 = crc_braid_big_table[0][word0 & 0xff]; |
972 | #if N > 1 |
973 | crc1 = crc_braid_big_table[0][word1 & 0xff]; |
974 | #if N > 2 |
975 | crc2 = crc_braid_big_table[0][word2 & 0xff]; |
976 | #if N > 3 |
977 | crc3 = crc_braid_big_table[0][word3 & 0xff]; |
978 | #if N > 4 |
979 | crc4 = crc_braid_big_table[0][word4 & 0xff]; |
980 | #if N > 5 |
981 | crc5 = crc_braid_big_table[0][word5 & 0xff]; |
982 | #endif |
983 | #endif |
984 | #endif |
985 | #endif |
986 | #endif |
987 | for (k = 1; k < W; k++) { |
988 | crc0 ^= crc_braid_big_table[k][(word0 >> (k << 3)) & 0xff]; |
989 | #if N > 1 |
990 | crc1 ^= crc_braid_big_table[k][(word1 >> (k << 3)) & 0xff]; |
991 | #if N > 2 |
992 | crc2 ^= crc_braid_big_table[k][(word2 >> (k << 3)) & 0xff]; |
993 | #if N > 3 |
994 | crc3 ^= crc_braid_big_table[k][(word3 >> (k << 3)) & 0xff]; |
995 | #if N > 4 |
996 | crc4 ^= crc_braid_big_table[k][(word4 >> (k << 3)) & 0xff]; |
997 | #if N > 5 |
998 | crc5 ^= crc_braid_big_table[k][(word5 >> (k << 3)) & 0xff]; |
999 | #endif |
1000 | #endif |
1001 | #endif |
1002 | #endif |
1003 | #endif |
1004 | } |
1005 | } |
1006 | |
1007 | /* |
1008 | Process the last block, combining the CRCs of the N braids at the |
1009 | same time. |
1010 | */ |
1011 | comb = crc_word_big(crc0 ^ words[0]); |
1012 | #if N > 1 |
1013 | comb = crc_word_big(crc1 ^ words[1] ^ comb); |
1014 | #if N > 2 |
1015 | comb = crc_word_big(crc2 ^ words[2] ^ comb); |
1016 | #if N > 3 |
1017 | comb = crc_word_big(crc3 ^ words[3] ^ comb); |
1018 | #if N > 4 |
1019 | comb = crc_word_big(crc4 ^ words[4] ^ comb); |
1020 | #if N > 5 |
1021 | comb = crc_word_big(crc5 ^ words[5] ^ comb); |
1022 | #endif |
1023 | #endif |
1024 | #endif |
1025 | #endif |
1026 | #endif |
1027 | words += N; |
1028 | crc = byte_swap(comb); |
1029 | } |
1030 | |
1031 | /* |
1032 | Update the pointer to the remaining bytes to process. |
1033 | */ |
1034 | buf = (unsigned char const *)words; |
1035 | } |
1036 | |
1037 | #endif /* W */ |
1038 | |
1039 | /* Complete the computation of the CRC on any remaining bytes. */ |
1040 | while (len >= 8) { |
1041 | len -= 8; |
1042 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1043 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1044 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1045 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1046 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1047 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1048 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1049 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1050 | } |
1051 | while (len) { |
1052 | len--; |
1053 | crc = (crc >> 8) ^ crc_table[(crc ^ *buf++) & 0xff]; |
1054 | } |
1055 | |
1056 | /* Return the CRC, post-conditioned. */ |
1057 | return crc ^ 0xffffffff; |
1058 | } |
1059 | |
1060 | #endif |
1061 | |
1062 | /* ========================================================================= */ |
1063 | unsigned long ZEXPORT crc32(crc, buf, len) |
1064 | unsigned long crc; |
1065 | const unsigned char FAR *buf; |
1066 | uInt len; |
1067 | { |
1068 | return crc32_z(crc, buf, len); |
1069 | } |
1070 | |
1071 | /* ========================================================================= */ |
1072 | uLong ZEXPORT crc32_combine64(crc1, crc2, len2) |
1073 | uLong crc1; |
1074 | uLong crc2; |
1075 | z_off64_t len2; |
1076 | { |
1077 | #ifdef DYNAMIC_CRC_TABLE |
1078 | once(&made, make_crc_table); |
1079 | #endif /* DYNAMIC_CRC_TABLE */ |
1080 | return multmodp(x2nmodp(len2, 3), crc1) ^ crc2; |
1081 | } |
1082 | |
1083 | /* ========================================================================= */ |
1084 | uLong ZEXPORT crc32_combine(crc1, crc2, len2) |
1085 | uLong crc1; |
1086 | uLong crc2; |
1087 | z_off_t len2; |
1088 | { |
1089 | return crc32_combine64(crc1, crc2, len2); |
1090 | } |
1091 | |
1092 | /* ========================================================================= */ |
1093 | uLong ZEXPORT crc32_combine_gen64(len2) |
1094 | z_off64_t len2; |
1095 | { |
1096 | #ifdef DYNAMIC_CRC_TABLE |
1097 | once(&made, make_crc_table); |
1098 | #endif /* DYNAMIC_CRC_TABLE */ |
1099 | return x2nmodp(len2, 3); |
1100 | } |
1101 | |
1102 | /* ========================================================================= */ |
1103 | uLong ZEXPORT crc32_combine_gen(len2) |
1104 | z_off_t len2; |
1105 | { |
1106 | return crc32_combine_gen64(len2); |
1107 | } |
1108 | |
1109 | /* ========================================================================= */ |
1110 | uLong crc32_combine_op(crc1, crc2, op) |
1111 | uLong crc1; |
1112 | uLong crc2; |
1113 | uLong op; |
1114 | { |
1115 | return multmodp(op, crc1) ^ crc2; |
1116 | } |
1117 | |