]>
Commit | Line | Data |
---|---|---|
e0e53b2f CC |
1 | /* |
2 | * Bitops Module | |
3 | * | |
4 | * Copyright (C) 2010 Corentin Chary <corentin.chary@gmail.com> | |
5 | * | |
6 | * Mostly inspired by (stolen from) linux/bitmap.h and linux/bitops.h | |
7 | * | |
8 | * This work is licensed under the terms of the GNU LGPL, version 2.1 or later. | |
9 | * See the COPYING.LIB file in the top-level directory. | |
10 | */ | |
11 | ||
12 | #ifndef BITOPS_H | |
13 | #define BITOPS_H | |
14 | ||
afeeead1 | 15 | |
fbeadf50 | 16 | #include "host-utils.h" |
9f02cfc8 | 17 | #include "atomic.h" |
e0e53b2f CC |
18 | |
19 | #define BITS_PER_BYTE CHAR_BIT | |
20 | #define BITS_PER_LONG (sizeof (unsigned long) * BITS_PER_BYTE) | |
21 | ||
4188e390 | 22 | #define BIT(nr) (1UL << (nr)) |
0df9142d | 23 | #define BIT_ULL(nr) (1ULL << (nr)) |
4188e390 CMC |
24 | #define BIT_MASK(nr) (1UL << ((nr) % BITS_PER_LONG)) |
25 | #define BIT_WORD(nr) ((nr) / BITS_PER_LONG) | |
26 | #define BITS_TO_LONGS(nr) DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long)) | |
e0e53b2f | 27 | |
ae2923b5 AF |
28 | #define MAKE_64BIT_MASK(shift, length) \ |
29 | (((~0ULL) >> (64 - (length))) << (shift)) | |
30 | ||
e0e53b2f CC |
31 | /** |
32 | * set_bit - Set a bit in memory | |
33 | * @nr: the bit to set | |
34 | * @addr: the address to start counting from | |
35 | */ | |
9c22687e | 36 | static inline void set_bit(long nr, unsigned long *addr) |
e0e53b2f | 37 | { |
4188e390 CMC |
38 | unsigned long mask = BIT_MASK(nr); |
39 | unsigned long *p = addr + BIT_WORD(nr); | |
e0e53b2f | 40 | |
4188e390 | 41 | *p |= mask; |
e0e53b2f CC |
42 | } |
43 | ||
9f02cfc8 SH |
44 | /** |
45 | * set_bit_atomic - Set a bit in memory atomically | |
46 | * @nr: the bit to set | |
47 | * @addr: the address to start counting from | |
48 | */ | |
49 | static inline void set_bit_atomic(long nr, unsigned long *addr) | |
50 | { | |
51 | unsigned long mask = BIT_MASK(nr); | |
52 | unsigned long *p = addr + BIT_WORD(nr); | |
53 | ||
d73415a3 | 54 | qatomic_or(p, mask); |
9f02cfc8 SH |
55 | } |
56 | ||
e0e53b2f CC |
57 | /** |
58 | * clear_bit - Clears a bit in memory | |
59 | * @nr: Bit to clear | |
60 | * @addr: Address to start counting from | |
61 | */ | |
9c22687e | 62 | static inline void clear_bit(long nr, unsigned long *addr) |
e0e53b2f | 63 | { |
4188e390 CMC |
64 | unsigned long mask = BIT_MASK(nr); |
65 | unsigned long *p = addr + BIT_WORD(nr); | |
e0e53b2f | 66 | |
4188e390 | 67 | *p &= ~mask; |
e0e53b2f CC |
68 | } |
69 | ||
f427d90b FR |
70 | /** |
71 | * clear_bit_atomic - Clears a bit in memory atomically | |
72 | * @nr: Bit to clear | |
73 | * @addr: Address to start counting from | |
74 | */ | |
75 | static inline void clear_bit_atomic(long nr, unsigned long *addr) | |
76 | { | |
77 | unsigned long mask = BIT_MASK(nr); | |
78 | unsigned long *p = addr + BIT_WORD(nr); | |
79 | ||
80 | return qatomic_and(p, ~mask); | |
81 | } | |
82 | ||
e0e53b2f CC |
83 | /** |
84 | * change_bit - Toggle a bit in memory | |
85 | * @nr: Bit to change | |
86 | * @addr: Address to start counting from | |
87 | */ | |
9c22687e | 88 | static inline void change_bit(long nr, unsigned long *addr) |
e0e53b2f | 89 | { |
4188e390 CMC |
90 | unsigned long mask = BIT_MASK(nr); |
91 | unsigned long *p = addr + BIT_WORD(nr); | |
e0e53b2f | 92 | |
4188e390 | 93 | *p ^= mask; |
e0e53b2f CC |
94 | } |
95 | ||
96 | /** | |
97 | * test_and_set_bit - Set a bit and return its old value | |
98 | * @nr: Bit to set | |
99 | * @addr: Address to count from | |
100 | */ | |
9c22687e | 101 | static inline int test_and_set_bit(long nr, unsigned long *addr) |
e0e53b2f | 102 | { |
4188e390 CMC |
103 | unsigned long mask = BIT_MASK(nr); |
104 | unsigned long *p = addr + BIT_WORD(nr); | |
105 | unsigned long old = *p; | |
e0e53b2f | 106 | |
4188e390 CMC |
107 | *p = old | mask; |
108 | return (old & mask) != 0; | |
e0e53b2f CC |
109 | } |
110 | ||
111 | /** | |
112 | * test_and_clear_bit - Clear a bit and return its old value | |
113 | * @nr: Bit to clear | |
114 | * @addr: Address to count from | |
115 | */ | |
9c22687e | 116 | static inline int test_and_clear_bit(long nr, unsigned long *addr) |
e0e53b2f | 117 | { |
4188e390 CMC |
118 | unsigned long mask = BIT_MASK(nr); |
119 | unsigned long *p = addr + BIT_WORD(nr); | |
120 | unsigned long old = *p; | |
e0e53b2f | 121 | |
4188e390 CMC |
122 | *p = old & ~mask; |
123 | return (old & mask) != 0; | |
e0e53b2f CC |
124 | } |
125 | ||
126 | /** | |
127 | * test_and_change_bit - Change a bit and return its old value | |
128 | * @nr: Bit to change | |
129 | * @addr: Address to count from | |
130 | */ | |
9c22687e | 131 | static inline int test_and_change_bit(long nr, unsigned long *addr) |
e0e53b2f | 132 | { |
4188e390 CMC |
133 | unsigned long mask = BIT_MASK(nr); |
134 | unsigned long *p = addr + BIT_WORD(nr); | |
135 | unsigned long old = *p; | |
e0e53b2f | 136 | |
4188e390 CMC |
137 | *p = old ^ mask; |
138 | return (old & mask) != 0; | |
e0e53b2f CC |
139 | } |
140 | ||
141 | /** | |
142 | * test_bit - Determine whether a bit is set | |
143 | * @nr: bit number to test | |
144 | * @addr: Address to start counting from | |
145 | */ | |
9c22687e | 146 | static inline int test_bit(long nr, const unsigned long *addr) |
e0e53b2f | 147 | { |
4188e390 | 148 | return 1UL & (addr[BIT_WORD(nr)] >> (nr & (BITS_PER_LONG-1))); |
e0e53b2f CC |
149 | } |
150 | ||
151 | /** | |
152 | * find_last_bit - find the last set bit in a memory region | |
153 | * @addr: The address to start the search at | |
154 | * @size: The maximum size to search | |
155 | * | |
5c6ae58d PMD |
156 | * Returns the bit number of the last set bit, |
157 | * or @size if there is no set bit in the bitmap. | |
e0e53b2f CC |
158 | */ |
159 | unsigned long find_last_bit(const unsigned long *addr, | |
160 | unsigned long size); | |
161 | ||
162 | /** | |
163 | * find_next_bit - find the next set bit in a memory region | |
164 | * @addr: The address to base the search on | |
165 | * @offset: The bitnumber to start searching at | |
166 | * @size: The bitmap size in bits | |
5c6ae58d PMD |
167 | * |
168 | * Returns the bit number of the next set bit, | |
169 | * or @size if there are no further set bits in the bitmap. | |
e0e53b2f CC |
170 | */ |
171 | unsigned long find_next_bit(const unsigned long *addr, | |
4188e390 CMC |
172 | unsigned long size, |
173 | unsigned long offset); | |
e0e53b2f CC |
174 | |
175 | /** | |
176 | * find_next_zero_bit - find the next cleared bit in a memory region | |
177 | * @addr: The address to base the search on | |
178 | * @offset: The bitnumber to start searching at | |
179 | * @size: The bitmap size in bits | |
5c6ae58d PMD |
180 | * |
181 | * Returns the bit number of the next cleared bit, | |
182 | * or @size if there are no further clear bits in the bitmap. | |
e0e53b2f CC |
183 | */ |
184 | ||
185 | unsigned long find_next_zero_bit(const unsigned long *addr, | |
186 | unsigned long size, | |
187 | unsigned long offset); | |
188 | ||
189 | /** | |
190 | * find_first_bit - find the first set bit in a memory region | |
191 | * @addr: The address to start the search at | |
192 | * @size: The maximum size to search | |
193 | * | |
5c6ae58d PMD |
194 | * Returns the bit number of the first set bit, |
195 | * or @size if there is no set bit in the bitmap. | |
e0e53b2f CC |
196 | */ |
197 | static inline unsigned long find_first_bit(const unsigned long *addr, | |
198 | unsigned long size) | |
199 | { | |
739b7a90 AJ |
200 | unsigned long result, tmp; |
201 | ||
202 | for (result = 0; result < size; result += BITS_PER_LONG) { | |
203 | tmp = *addr++; | |
204 | if (tmp) { | |
205 | result += ctzl(tmp); | |
206 | return result < size ? result : size; | |
207 | } | |
208 | } | |
209 | /* Not found */ | |
210 | return size; | |
e0e53b2f CC |
211 | } |
212 | ||
213 | /** | |
214 | * find_first_zero_bit - find the first cleared bit in a memory region | |
215 | * @addr: The address to start the search at | |
216 | * @size: The maximum size to search | |
217 | * | |
5c6ae58d PMD |
218 | * Returns the bit number of the first cleared bit, |
219 | * or @size if there is no clear bit in the bitmap. | |
e0e53b2f CC |
220 | */ |
221 | static inline unsigned long find_first_zero_bit(const unsigned long *addr, | |
222 | unsigned long size) | |
223 | { | |
224 | return find_next_zero_bit(addr, size, 0); | |
225 | } | |
226 | ||
6aa25b4a SW |
227 | /** |
228 | * rol8 - rotate an 8-bit value left | |
229 | * @word: value to rotate | |
230 | * @shift: bits to roll | |
231 | */ | |
232 | static inline uint8_t rol8(uint8_t word, unsigned int shift) | |
233 | { | |
8841c815 | 234 | return (word << (shift & 7)) | (word >> (-shift & 7)); |
6aa25b4a SW |
235 | } |
236 | ||
237 | /** | |
238 | * ror8 - rotate an 8-bit value right | |
239 | * @word: value to rotate | |
240 | * @shift: bits to roll | |
241 | */ | |
242 | static inline uint8_t ror8(uint8_t word, unsigned int shift) | |
243 | { | |
8841c815 | 244 | return (word >> (shift & 7)) | (word << (-shift & 7)); |
6aa25b4a SW |
245 | } |
246 | ||
247 | /** | |
248 | * rol16 - rotate a 16-bit value left | |
249 | * @word: value to rotate | |
250 | * @shift: bits to roll | |
251 | */ | |
252 | static inline uint16_t rol16(uint16_t word, unsigned int shift) | |
253 | { | |
8841c815 | 254 | return (word << (shift & 15)) | (word >> (-shift & 15)); |
6aa25b4a SW |
255 | } |
256 | ||
257 | /** | |
258 | * ror16 - rotate a 16-bit value right | |
259 | * @word: value to rotate | |
260 | * @shift: bits to roll | |
261 | */ | |
262 | static inline uint16_t ror16(uint16_t word, unsigned int shift) | |
263 | { | |
8841c815 | 264 | return (word >> (shift & 15)) | (word << (-shift & 15)); |
6aa25b4a SW |
265 | } |
266 | ||
267 | /** | |
268 | * rol32 - rotate a 32-bit value left | |
269 | * @word: value to rotate | |
270 | * @shift: bits to roll | |
271 | */ | |
272 | static inline uint32_t rol32(uint32_t word, unsigned int shift) | |
273 | { | |
8841c815 | 274 | return (word << (shift & 31)) | (word >> (-shift & 31)); |
6aa25b4a SW |
275 | } |
276 | ||
277 | /** | |
278 | * ror32 - rotate a 32-bit value right | |
279 | * @word: value to rotate | |
280 | * @shift: bits to roll | |
281 | */ | |
282 | static inline uint32_t ror32(uint32_t word, unsigned int shift) | |
283 | { | |
8841c815 | 284 | return (word >> (shift & 31)) | (word << (-shift & 31)); |
6aa25b4a SW |
285 | } |
286 | ||
287 | /** | |
288 | * rol64 - rotate a 64-bit value left | |
289 | * @word: value to rotate | |
290 | * @shift: bits to roll | |
291 | */ | |
292 | static inline uint64_t rol64(uint64_t word, unsigned int shift) | |
293 | { | |
8841c815 | 294 | return (word << (shift & 63)) | (word >> (-shift & 63)); |
6aa25b4a SW |
295 | } |
296 | ||
297 | /** | |
298 | * ror64 - rotate a 64-bit value right | |
299 | * @word: value to rotate | |
300 | * @shift: bits to roll | |
301 | */ | |
302 | static inline uint64_t ror64(uint64_t word, unsigned int shift) | |
303 | { | |
8841c815 | 304 | return (word >> (shift & 63)) | (word << (-shift & 63)); |
6aa25b4a SW |
305 | } |
306 | ||
dbcf6f93 PM |
307 | /** |
308 | * hswap32 - swap 16-bit halfwords within a 32-bit value | |
309 | * @h: value to swap | |
310 | */ | |
311 | static inline uint32_t hswap32(uint32_t h) | |
312 | { | |
313 | return rol32(h, 16); | |
314 | } | |
315 | ||
316 | /** | |
317 | * hswap64 - swap 16-bit halfwords within a 64-bit value | |
318 | * @h: value to swap | |
319 | */ | |
320 | static inline uint64_t hswap64(uint64_t h) | |
321 | { | |
322 | uint64_t m = 0x0000ffff0000ffffull; | |
323 | h = rol64(h, 32); | |
324 | return ((h & m) << 16) | ((h >> 16) & m); | |
325 | } | |
326 | ||
327 | /** | |
328 | * wswap64 - swap 32-bit words within a 64-bit value | |
329 | * @h: value to swap | |
330 | */ | |
331 | static inline uint64_t wswap64(uint64_t h) | |
332 | { | |
333 | return rol64(h, 32); | |
334 | } | |
335 | ||
84988cf9 PM |
336 | /** |
337 | * extract32: | |
338 | * @value: the value to extract the bit field from | |
339 | * @start: the lowest bit in the bit field (numbered from 0) | |
340 | * @length: the length of the bit field | |
341 | * | |
342 | * Extract from the 32 bit input @value the bit field specified by the | |
343 | * @start and @length parameters, and return it. The bit field must | |
344 | * lie entirely within the 32 bit word. It is valid to request that | |
345 | * all 32 bits are returned (ie @length 32 and @start 0). | |
346 | * | |
347 | * Returns: the value of the bit field extracted from the input value. | |
348 | */ | |
349 | static inline uint32_t extract32(uint32_t value, int start, int length) | |
350 | { | |
351 | assert(start >= 0 && length > 0 && length <= 32 - start); | |
352 | return (value >> start) & (~0U >> (32 - length)); | |
353 | } | |
354 | ||
ed04c8b1 YS |
355 | /** |
356 | * extract8: | |
357 | * @value: the value to extract the bit field from | |
358 | * @start: the lowest bit in the bit field (numbered from 0) | |
359 | * @length: the length of the bit field | |
360 | * | |
361 | * Extract from the 8 bit input @value the bit field specified by the | |
362 | * @start and @length parameters, and return it. The bit field must | |
363 | * lie entirely within the 8 bit word. It is valid to request that | |
364 | * all 8 bits are returned (ie @length 8 and @start 0). | |
365 | * | |
366 | * Returns: the value of the bit field extracted from the input value. | |
367 | */ | |
368 | static inline uint8_t extract8(uint8_t value, int start, int length) | |
369 | { | |
370 | assert(start >= 0 && length > 0 && length <= 8 - start); | |
371 | return extract32(value, start, length); | |
372 | } | |
373 | ||
374 | /** | |
375 | * extract16: | |
376 | * @value: the value to extract the bit field from | |
377 | * @start: the lowest bit in the bit field (numbered from 0) | |
378 | * @length: the length of the bit field | |
379 | * | |
380 | * Extract from the 16 bit input @value the bit field specified by the | |
381 | * @start and @length parameters, and return it. The bit field must | |
382 | * lie entirely within the 16 bit word. It is valid to request that | |
383 | * all 16 bits are returned (ie @length 16 and @start 0). | |
384 | * | |
385 | * Returns: the value of the bit field extracted from the input value. | |
386 | */ | |
387 | static inline uint16_t extract16(uint16_t value, int start, int length) | |
388 | { | |
389 | assert(start >= 0 && length > 0 && length <= 16 - start); | |
390 | return extract32(value, start, length); | |
391 | } | |
392 | ||
84988cf9 PM |
393 | /** |
394 | * extract64: | |
395 | * @value: the value to extract the bit field from | |
396 | * @start: the lowest bit in the bit field (numbered from 0) | |
397 | * @length: the length of the bit field | |
398 | * | |
399 | * Extract from the 64 bit input @value the bit field specified by the | |
400 | * @start and @length parameters, and return it. The bit field must | |
401 | * lie entirely within the 64 bit word. It is valid to request that | |
402 | * all 64 bits are returned (ie @length 64 and @start 0). | |
403 | * | |
404 | * Returns: the value of the bit field extracted from the input value. | |
405 | */ | |
406 | static inline uint64_t extract64(uint64_t value, int start, int length) | |
407 | { | |
408 | assert(start >= 0 && length > 0 && length <= 64 - start); | |
409 | return (value >> start) & (~0ULL >> (64 - length)); | |
410 | } | |
411 | ||
2dc6bebd PM |
412 | /** |
413 | * sextract32: | |
414 | * @value: the value to extract the bit field from | |
415 | * @start: the lowest bit in the bit field (numbered from 0) | |
416 | * @length: the length of the bit field | |
417 | * | |
418 | * Extract from the 32 bit input @value the bit field specified by the | |
419 | * @start and @length parameters, and return it, sign extended to | |
420 | * an int32_t (ie with the most significant bit of the field propagated | |
421 | * to all the upper bits of the return value). The bit field must lie | |
422 | * entirely within the 32 bit word. It is valid to request that | |
423 | * all 32 bits are returned (ie @length 32 and @start 0). | |
424 | * | |
425 | * Returns: the sign extended value of the bit field extracted from the | |
426 | * input value. | |
427 | */ | |
428 | static inline int32_t sextract32(uint32_t value, int start, int length) | |
429 | { | |
430 | assert(start >= 0 && length > 0 && length <= 32 - start); | |
431 | /* Note that this implementation relies on right shift of signed | |
432 | * integers being an arithmetic shift. | |
433 | */ | |
434 | return ((int32_t)(value << (32 - length - start))) >> (32 - length); | |
435 | } | |
436 | ||
437 | /** | |
438 | * sextract64: | |
439 | * @value: the value to extract the bit field from | |
440 | * @start: the lowest bit in the bit field (numbered from 0) | |
441 | * @length: the length of the bit field | |
442 | * | |
443 | * Extract from the 64 bit input @value the bit field specified by the | |
444 | * @start and @length parameters, and return it, sign extended to | |
445 | * an int64_t (ie with the most significant bit of the field propagated | |
446 | * to all the upper bits of the return value). The bit field must lie | |
447 | * entirely within the 64 bit word. It is valid to request that | |
448 | * all 64 bits are returned (ie @length 64 and @start 0). | |
449 | * | |
450 | * Returns: the sign extended value of the bit field extracted from the | |
451 | * input value. | |
452 | */ | |
4f995052 | 453 | static inline int64_t sextract64(uint64_t value, int start, int length) |
2dc6bebd PM |
454 | { |
455 | assert(start >= 0 && length > 0 && length <= 64 - start); | |
456 | /* Note that this implementation relies on right shift of signed | |
457 | * integers being an arithmetic shift. | |
458 | */ | |
459 | return ((int64_t)(value << (64 - length - start))) >> (64 - length); | |
460 | } | |
461 | ||
84988cf9 PM |
462 | /** |
463 | * deposit32: | |
464 | * @value: initial value to insert bit field into | |
465 | * @start: the lowest bit in the bit field (numbered from 0) | |
466 | * @length: the length of the bit field | |
467 | * @fieldval: the value to insert into the bit field | |
468 | * | |
469 | * Deposit @fieldval into the 32 bit @value at the bit field specified | |
470 | * by the @start and @length parameters, and return the modified | |
471 | * @value. Bits of @value outside the bit field are not modified. | |
472 | * Bits of @fieldval above the least significant @length bits are | |
473 | * ignored. The bit field must lie entirely within the 32 bit word. | |
ab411770 SW |
474 | * It is valid to request that all 32 bits are modified (ie @length |
475 | * 32 and @start 0). | |
84988cf9 PM |
476 | * |
477 | * Returns: the modified @value. | |
478 | */ | |
479 | static inline uint32_t deposit32(uint32_t value, int start, int length, | |
480 | uint32_t fieldval) | |
481 | { | |
482 | uint32_t mask; | |
483 | assert(start >= 0 && length > 0 && length <= 32 - start); | |
484 | mask = (~0U >> (32 - length)) << start; | |
485 | return (value & ~mask) | ((fieldval << start) & mask); | |
486 | } | |
487 | ||
488 | /** | |
ab411770 | 489 | * deposit64: |
84988cf9 PM |
490 | * @value: initial value to insert bit field into |
491 | * @start: the lowest bit in the bit field (numbered from 0) | |
492 | * @length: the length of the bit field | |
493 | * @fieldval: the value to insert into the bit field | |
494 | * | |
495 | * Deposit @fieldval into the 64 bit @value at the bit field specified | |
496 | * by the @start and @length parameters, and return the modified | |
497 | * @value. Bits of @value outside the bit field are not modified. | |
498 | * Bits of @fieldval above the least significant @length bits are | |
ab411770 | 499 | * ignored. The bit field must lie entirely within the 64 bit word. |
84988cf9 PM |
500 | * It is valid to request that all 64 bits are modified (ie @length |
501 | * 64 and @start 0). | |
502 | * | |
503 | * Returns: the modified @value. | |
504 | */ | |
505 | static inline uint64_t deposit64(uint64_t value, int start, int length, | |
506 | uint64_t fieldval) | |
507 | { | |
508 | uint64_t mask; | |
509 | assert(start >= 0 && length > 0 && length <= 64 - start); | |
510 | mask = (~0ULL >> (64 - length)) << start; | |
511 | return (value & ~mask) | ((fieldval << start) & mask); | |
512 | } | |
513 | ||
b355438d PM |
514 | /** |
515 | * half_shuffle32: | |
7d41d764 PM |
516 | * @x: 32-bit value (of which only the bottom 16 bits are of interest) |
517 | * | |
518 | * Given an input value:: | |
519 | * | |
520 | * xxxx xxxx xxxx xxxx ABCD EFGH IJKL MNOP | |
b355438d | 521 | * |
b355438d | 522 | * return the value where the bottom 16 bits are spread out into |
7d41d764 PM |
523 | * the odd bits in the word, and the even bits are zeroed:: |
524 | * | |
525 | * 0A0B 0C0D 0E0F 0G0H 0I0J 0K0L 0M0N 0O0P | |
b355438d PM |
526 | * |
527 | * Any bits set in the top half of the input are ignored. | |
528 | * | |
529 | * Returns: the shuffled bits. | |
530 | */ | |
531 | static inline uint32_t half_shuffle32(uint32_t x) | |
532 | { | |
533 | /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". | |
534 | * It ignores any bits set in the top half of the input. | |
535 | */ | |
536 | x = ((x & 0xFF00) << 8) | (x & 0x00FF); | |
537 | x = ((x << 4) | x) & 0x0F0F0F0F; | |
538 | x = ((x << 2) | x) & 0x33333333; | |
539 | x = ((x << 1) | x) & 0x55555555; | |
540 | return x; | |
541 | } | |
542 | ||
543 | /** | |
544 | * half_shuffle64: | |
7d41d764 PM |
545 | * @x: 64-bit value (of which only the bottom 32 bits are of interest) |
546 | * | |
547 | * Given an input value:: | |
548 | * | |
549 | * xxxx xxxx xxxx .... xxxx xxxx ABCD EFGH IJKL MNOP QRST UVWX YZab cdef | |
b355438d | 550 | * |
b355438d | 551 | * return the value where the bottom 32 bits are spread out into |
7d41d764 PM |
552 | * the odd bits in the word, and the even bits are zeroed:: |
553 | * | |
554 | * 0A0B 0C0D 0E0F 0G0H 0I0J 0K0L 0M0N .... 0U0V 0W0X 0Y0Z 0a0b 0c0d 0e0f | |
b355438d PM |
555 | * |
556 | * Any bits set in the top half of the input are ignored. | |
557 | * | |
558 | * Returns: the shuffled bits. | |
559 | */ | |
560 | static inline uint64_t half_shuffle64(uint64_t x) | |
561 | { | |
562 | /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". | |
563 | * It ignores any bits set in the top half of the input. | |
564 | */ | |
565 | x = ((x & 0xFFFF0000ULL) << 16) | (x & 0xFFFF); | |
566 | x = ((x << 8) | x) & 0x00FF00FF00FF00FFULL; | |
567 | x = ((x << 4) | x) & 0x0F0F0F0F0F0F0F0FULL; | |
568 | x = ((x << 2) | x) & 0x3333333333333333ULL; | |
569 | x = ((x << 1) | x) & 0x5555555555555555ULL; | |
570 | return x; | |
571 | } | |
572 | ||
573 | /** | |
574 | * half_unshuffle32: | |
7d41d764 PM |
575 | * @x: 32-bit value (of which only the odd bits are of interest) |
576 | * | |
577 | * Given an input value:: | |
578 | * | |
579 | * xAxB xCxD xExF xGxH xIxJ xKxL xMxN xOxP | |
b355438d | 580 | * |
b355438d | 581 | * return the value where all the odd bits are compressed down |
7d41d764 PM |
582 | * into the low half of the word, and the high half is zeroed:: |
583 | * | |
584 | * 0000 0000 0000 0000 ABCD EFGH IJKL MNOP | |
b355438d PM |
585 | * |
586 | * Any even bits set in the input are ignored. | |
587 | * | |
588 | * Returns: the unshuffled bits. | |
589 | */ | |
590 | static inline uint32_t half_unshuffle32(uint32_t x) | |
591 | { | |
592 | /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". | |
593 | * where it is called an inverse half shuffle. | |
594 | */ | |
595 | x &= 0x55555555; | |
596 | x = ((x >> 1) | x) & 0x33333333; | |
597 | x = ((x >> 2) | x) & 0x0F0F0F0F; | |
598 | x = ((x >> 4) | x) & 0x00FF00FF; | |
599 | x = ((x >> 8) | x) & 0x0000FFFF; | |
600 | return x; | |
601 | } | |
602 | ||
603 | /** | |
604 | * half_unshuffle64: | |
7d41d764 PM |
605 | * @x: 64-bit value (of which only the odd bits are of interest) |
606 | * | |
607 | * Given an input value:: | |
608 | * | |
609 | * xAxB xCxD xExF xGxH xIxJ xKxL xMxN .... xUxV xWxX xYxZ xaxb xcxd xexf | |
b355438d | 610 | * |
b355438d | 611 | * return the value where all the odd bits are compressed down |
7d41d764 PM |
612 | * into the low half of the word, and the high half is zeroed:: |
613 | * | |
614 | * 0000 0000 0000 .... 0000 0000 ABCD EFGH IJKL MNOP QRST UVWX YZab cdef | |
b355438d PM |
615 | * |
616 | * Any even bits set in the input are ignored. | |
617 | * | |
618 | * Returns: the unshuffled bits. | |
619 | */ | |
620 | static inline uint64_t half_unshuffle64(uint64_t x) | |
621 | { | |
622 | /* This algorithm is from _Hacker's Delight_ section 7-2 "Shuffling Bits". | |
623 | * where it is called an inverse half shuffle. | |
624 | */ | |
625 | x &= 0x5555555555555555ULL; | |
626 | x = ((x >> 1) | x) & 0x3333333333333333ULL; | |
627 | x = ((x >> 2) | x) & 0x0F0F0F0F0F0F0F0FULL; | |
628 | x = ((x >> 4) | x) & 0x00FF00FF00FF00FFULL; | |
629 | x = ((x >> 8) | x) & 0x0000FFFF0000FFFFULL; | |
630 | x = ((x >> 16) | x) & 0x00000000FFFFFFFFULL; | |
631 | return x; | |
632 | } | |
633 | ||
e0e53b2f | 634 | #endif |