summaryrefslogtreecommitdiffstats
path: root/src/common/tuklib_integer.h
blob: b58ef68d0905baedf4ed5fd549dd34ee788c23a3 (plain)
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
///////////////////////////////////////////////////////////////////////////////
//
/// \file       tuklib_integer.h
/// \brief      Various integer and bit operations
///
/// This file provides macros or functions to do some basic integer and bit
/// operations.
///
/// Native endian inline functions (XX = 16, 32, or 64):
///   - Unaligned native endian reads: readXXne(ptr)
///   - Unaligned native endian writes: writeXXne(ptr, num)
///   - Aligned native endian reads: aligned_readXXne(ptr)
///   - Aligned native endian writes: aligned_writeXXne(ptr, num)
///
/// Endianness-converting integer operations (these can be macros!)
/// (XX = 16, 32, or 64; Y = b or l):
///   - Byte swapping: bswapXX(num)
///   - Byte order conversions to/from native (byteswaps if Y isn't
///     the native endianness): convXXYe(num)
///   - Unaligned reads: readXXYe(ptr)
///   - Unaligned writes: writeXXYe(ptr, num)
///   - Aligned reads: aligned_readXXYe(ptr)
///   - Aligned writes: aligned_writeXXYe(ptr, num)
///
/// Since the above can macros, the arguments should have no side effects
/// because they may be evaluated more than once.
///
/// Bit scan operations for non-zero 32-bit integers (inline functions):
///   - Bit scan reverse (find highest non-zero bit): bsr32(num)
///   - Count leading zeros: clz32(num)
///   - Count trailing zeros: ctz32(num)
///   - Bit scan forward (simply an alias for ctz32()): bsf32(num)
///
/// The above bit scan operations return 0-31. If num is zero,
/// the result is undefined.
//
//  Authors:    Lasse Collin
//              Joachim Henke
//
//  This file has been put into the public domain.
//  You can do whatever you want with this file.
//
///////////////////////////////////////////////////////////////////////////////

#ifndef TUKLIB_INTEGER_H
#define TUKLIB_INTEGER_H

#include "tuklib_common.h"
#include <string.h>

// Newer Intel C compilers require immintrin.h for _bit_scan_reverse()
// and such functions.
#if defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1500)
#	include <immintrin.h>
#endif


///////////////////
// Byte swapping //
///////////////////

#if defined(HAVE___BUILTIN_BSWAPXX)
	// GCC >= 4.8 and Clang
#	define bswap16(n) __builtin_bswap16(n)
#	define bswap32(n) __builtin_bswap32(n)
#	define bswap64(n) __builtin_bswap64(n)

#elif defined(HAVE_BYTESWAP_H)
	// glibc, uClibc, dietlibc
#	include <byteswap.h>
#	ifdef HAVE_BSWAP_16
#		define bswap16(num) bswap_16(num)
#	endif
#	ifdef HAVE_BSWAP_32
#		define bswap32(num) bswap_32(num)
#	endif
#	ifdef HAVE_BSWAP_64
#		define bswap64(num) bswap_64(num)
#	endif

#elif defined(HAVE_SYS_ENDIAN_H)
	// *BSDs and Darwin
#	include <sys/endian.h>

#elif defined(HAVE_SYS_BYTEORDER_H)
	// Solaris
#	include <sys/byteorder.h>
#	ifdef BSWAP_16
#		define bswap16(num) BSWAP_16(num)
#	endif
#	ifdef BSWAP_32
#		define bswap32(num) BSWAP_32(num)
#	endif
#	ifdef BSWAP_64
#		define bswap64(num) BSWAP_64(num)
#	endif
#	ifdef BE_16
#		define conv16be(num) BE_16(num)
#	endif
#	ifdef BE_32
#		define conv32be(num) BE_32(num)
#	endif
#	ifdef BE_64
#		define conv64be(num) BE_64(num)
#	endif
#	ifdef LE_16
#		define conv16le(num) LE_16(num)
#	endif
#	ifdef LE_32
#		define conv32le(num) LE_32(num)
#	endif
#	ifdef LE_64
#		define conv64le(num) LE_64(num)
#	endif
#endif

#ifndef bswap16
#	define bswap16(n) (uint16_t)( \
		  (((n) & 0x00FFU) << 8) \
		| (((n) & 0xFF00U) >> 8) \
	)
#endif

#ifndef bswap32
#	define bswap32(n) (uint32_t)( \
		  (((n) & UINT32_C(0x000000FF)) << 24) \
		| (((n) & UINT32_C(0x0000FF00)) << 8) \
		| (((n) & UINT32_C(0x00FF0000)) >> 8) \
		| (((n) & UINT32_C(0xFF000000)) >> 24) \
	)
#endif

#ifndef bswap64
#	define bswap64(n) (uint64_t)( \
		  (((n) & UINT64_C(0x00000000000000FF)) << 56) \
		| (((n) & UINT64_C(0x000000000000FF00)) << 40) \
		| (((n) & UINT64_C(0x0000000000FF0000)) << 24) \
		| (((n) & UINT64_C(0x00000000FF000000)) << 8) \
		| (((n) & UINT64_C(0x000000FF00000000)) >> 8) \
		| (((n) & UINT64_C(0x0000FF0000000000)) >> 24) \
		| (((n) & UINT64_C(0x00FF000000000000)) >> 40) \
		| (((n) & UINT64_C(0xFF00000000000000)) >> 56) \
	)
#endif

// Define conversion macros using the basic byte swapping macros.
#ifdef WORDS_BIGENDIAN
#	ifndef conv16be
#		define conv16be(num) ((uint16_t)(num))
#	endif
#	ifndef conv32be
#		define conv32be(num) ((uint32_t)(num))
#	endif
#	ifndef conv64be
#		define conv64be(num) ((uint64_t)(num))
#	endif
#	ifndef conv16le
#		define conv16le(num) bswap16(num)
#	endif
#	ifndef conv32le
#		define conv32le(num) bswap32(num)
#	endif
#	ifndef conv64le
#		define conv64le(num) bswap64(num)
#	endif
#else
#	ifndef conv16be
#		define conv16be(num) bswap16(num)
#	endif
#	ifndef conv32be
#		define conv32be(num) bswap32(num)
#	endif
#	ifndef conv64be
#		define conv64be(num) bswap64(num)
#	endif
#	ifndef conv16le
#		define conv16le(num) ((uint16_t)(num))
#	endif
#	ifndef conv32le
#		define conv32le(num) ((uint32_t)(num))
#	endif
#	ifndef conv64le
#		define conv64le(num) ((uint64_t)(num))
#	endif
#endif


////////////////////////////////
// Unaligned reads and writes //
////////////////////////////////

// The traditional way of casting e.g. *(const uint16_t *)uint8_pointer
// is bad even if the uint8_pointer is properly aligned because this kind
// of casts break strict aliasing rules and result in undefined behavior.
// With unaligned pointers it's even worse: compilers may emit vector
// instructions that require aligned pointers even if non-vector
// instructions work with unaligned pointers.
//
// Using memcpy() is the standard compliant way to do unaligned access.
// Many modern compilers inline it so there is no function call overhead.
// For those compilers that don't handle the memcpy() method well, the
// old casting method (that violates strict aliasing) can be requested at
// build time. A third method, casting to a packed struct, would also be
// an option but isn't provided to keep things simpler (it's already a mess).
// Hopefully this is flexible enough in practice.

static inline uint16_t
read16ne(const uint8_t *buf)
{
#if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \
		&& defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING)
	return *(const uint16_t *)buf;
#else
	uint16_t num;
	memcpy(&num, buf, sizeof(num));
	return num;
#endif
}


static inline uint32_t
read32ne(const uint8_t *buf)
{
#if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \
		&& defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING)
	return *(const uint32_t *)buf;
#else
	uint32_t num;
	memcpy(&num, buf, sizeof(num));
	return num;
#endif
}


static inline uint64_t
read64ne(const uint8_t *buf)
{
#if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \
		&& defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING)
	return *(const uint64_t *)buf;
#else
	uint64_t num;
	memcpy(&num, buf, sizeof(num));
	return num;
#endif
}


static inline void
write16ne(uint8_t *buf, uint16_t num)
{
#if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \
		&& defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING)
	*(uint16_t *)buf = num;
#else
	memcpy(buf, &num, sizeof(num));
#endif
	return;
}


static inline void
write32ne(uint8_t *buf, uint32_t num)
{
#if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \
		&& defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING)
	*(uint32_t *)buf = num;
#else
	memcpy(buf, &num, sizeof(num));
#endif
	return;
}


static inline void
write64ne(uint8_t *buf, uint64_t num)
{
#if defined(TUKLIB_FAST_UNALIGNED_ACCESS) \
		&& defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING)
	*(uint64_t *)buf = num;
#else
	memcpy(buf, &num, sizeof(num));
#endif
	return;
}


static inline uint16_t
read16be(const uint8_t *buf)
{
#if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
	uint16_t num = read16ne(buf);
	return conv16be(num);
#else
	uint16_t num = ((uint16_t)buf[0] << 8) | (uint16_t)buf[1];
	return num;
#endif
}


static inline uint16_t
read16le(const uint8_t *buf)
{
#if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
	uint16_t num = read16ne(buf);
	return conv16le(num);
#else
	uint16_t num = ((uint16_t)buf[0]) | ((uint16_t)buf[1] << 8);
	return num;
#endif
}


static inline uint32_t
read32be(const uint8_t *buf)
{
#if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
	uint32_t num = read32ne(buf);
	return conv32be(num);
#else
	uint32_t num = (uint32_t)buf[0] << 24;
	num |= (uint32_t)buf[1] << 16;
	num |= (uint32_t)buf[2] << 8;
	num |= (uint32_t)buf[3];
	return num;
#endif
}


static inline uint32_t
read32le(const uint8_t *buf)
{
#if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
	uint32_t num = read32ne(buf);
	return conv32le(num);
#else
	uint32_t num = (uint32_t)buf[0];
	num |= (uint32_t)buf[1] << 8;
	num |= (uint32_t)buf[2] << 16;
	num |= (uint32_t)buf[3] << 24;
	return num;
#endif
}


static inline uint64_t
read64be(const uint8_t *buf)
{
#if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
	uint64_t num = read64ne(buf);
	return conv64be(num);
#else
	uint64_t num = (uint64_t)buf[0] << 56;
	num |= (uint64_t)buf[1] << 48;
	num |= (uint64_t)buf[2] << 40;
	num |= (uint64_t)buf[3] << 32;
	num |= (uint64_t)buf[4] << 24;
	num |= (uint64_t)buf[5] << 16;
	num |= (uint64_t)buf[6] << 8;
	num |= (uint64_t)buf[7];
	return num;
#endif
}


static inline uint64_t
read64le(const uint8_t *buf)
{
#if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
	uint64_t num = read64ne(buf);
	return conv64le(num);
#else
	uint64_t num = (uint64_t)buf[0];
	num |= (uint64_t)buf[1] << 8;
	num |= (uint64_t)buf[2] << 16;
	num |= (uint64_t)buf[3] << 24;
	num |= (uint64_t)buf[4] << 32;
	num |= (uint64_t)buf[5] << 40;
	num |= (uint64_t)buf[6] << 48;
	num |= (uint64_t)buf[7] << 56;
	return num;
#endif
}


// NOTE: Possible byte swapping must be done in a macro to allow the compiler
// to optimize byte swapping of constants when using glibc's or *BSD's
// byte swapping macros. The actual write is done in an inline function
// to make type checking of the buf pointer possible.
#if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
#	define write16be(buf, num) write16ne(buf, conv16be(num))
#	define write32be(buf, num) write32ne(buf, conv32be(num))
#	define write64be(buf, num) write64ne(buf, conv64be(num))
#endif

#if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS)
#	define write16le(buf, num) write16ne(buf, conv16le(num))
#	define write32le(buf, num) write32ne(buf, conv32le(num))
#	define write64le(buf, num) write64ne(buf, conv64le(num))
#endif


#ifndef write16be
static inline void
write16be(uint8_t *buf, uint16_t num)
{
	buf[0] = (uint8_t)(num >> 8);
	buf[1] = (uint8_t)num;
	return;
}
#endif


#ifndef write16le
static inline void
write16le(uint8_t *buf, uint16_t num)
{
	buf[0] = (uint8_t)num;
	buf[1] = (uint8_t)(num >> 8);
	return;
}
#endif


#ifndef write32be
static inline void
write32be(uint8_t *buf, uint32_t num)
{
	buf[0] = (uint8_t)(num >> 24);
	buf[1] = (uint8_t)(num >> 16);
	buf[2] = (uint8_t)(num >> 8);
	buf[3] = (uint8_t)num;
	return;
}
#endif


#ifndef write32le
static inline void
write32le(uint8_t *buf, uint32_t num)
{
	buf[0] = (uint8_t)num;
	buf[1] = (uint8_t)(num >> 8);
	buf[2] = (uint8_t)(num >> 16);
	buf[3] = (uint8_t)(num >> 24);
	return;
}
#endif


//////////////////////////////
// Aligned reads and writes //
//////////////////////////////

// Separate functions for aligned reads and writes are provided since on
// strict-align archs aligned access is much faster than unaligned access.
//
// Just like in the unaligned case, memcpy() is needed to avoid
// strict aliasing violations. However, on archs that don't support
// unaligned access the compiler cannot know that the pointers given
// to memcpy() are aligned which results in slow code. As of C11 there is
// no standard way to tell the compiler that we know that the address is
// aligned but some compilers have language extensions to do that. With
// such language extensions the memcpy() method gives excellent results.
//
// What to do on a strict-align system when no known language extentensions
// are available? Falling back to byte-by-byte access would be safe but ruin
// optimizations that have been made specifically with aligned access in mind.
// As a compromise, aligned reads will fall back to non-compliant type punning
// but aligned writes will be byte-by-byte, that is, fast reads are preferred
// over fast writes. This obviously isn't great but hopefully it's a working
// compromise for now.
//
// __builtin_assume_aligned is support by GCC >= 4.7 and clang >= 3.6.
#ifdef HAVE___BUILTIN_ASSUME_ALIGNED
#	define tuklib_memcpy_aligned(dest, src, size) \
		memcpy(dest, __builtin_assume_aligned(src, size), size)
#else
#	define tuklib_memcpy_aligned(dest, src, size) \
		memcpy(dest, src, size)
#	ifndef TUKLIB_FAST_UNALIGNED_ACCESS
#		define TUKLIB_USE_UNSAFE_ALIGNED_READS 1
#	endif
#endif


static inline uint16_t
aligned_read16ne(const uint8_t *buf)
{
#if defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) \
		|| defined(TUKLIB_USE_UNSAFE_ALIGNED_READS)
	return *(const uint16_t *)buf;
#else
	uint16_t num;
	tuklib_memcpy_aligned(&num, buf, sizeof(num));
	return num;
#endif
}


static inline uint32_t
aligned_read32ne(const uint8_t *buf)
{
#if defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) \
		|| defined(TUKLIB_USE_UNSAFE_ALIGNED_READS)
	return *(const uint32_t *)buf;
#else
	uint32_t num;
	tuklib_memcpy_aligned(&num, buf, sizeof(num));
	return num;
#endif
}


static inline uint64_t
aligned_read64ne(const uint8_t *buf)
{
#if defined(TUKLIB_USE_UNSAFE_TYPE_PUNNING) \
		|| defined(TUKLIB_USE_UNSAFE_ALIGNED_READS)
	return *(const uint64_t *)buf;
#else
	uint64_t num;
	tuklib_memcpy_aligned(&num, buf, sizeof(num));
	return num;
#endif
}


static inline void
aligned_write16ne(uint8_t *buf, uint16_t num)
{
#ifdef TUKLIB_USE_UNSAFE_TYPE_PUNNING
	*(uint16_t *)buf = num;
#else
	tuklib_memcpy_aligned(buf, &num, sizeof(num));
#endif
	return;
}


static inline void
aligned_write32ne(uint8_t *buf, uint32_t num)
{
#ifdef TUKLIB_USE_UNSAFE_TYPE_PUNNING
	*(uint32_t *)buf = num;
#else
	tuklib_memcpy_aligned(buf, &num, sizeof(num));
#endif
	return;
}


static inline void
aligned_write64ne(uint8_t *buf, uint64_t num)
{
#ifdef TUKLIB_USE_UNSAFE_TYPE_PUNNING
	*(uint64_t *)buf = num;
#else
	tuklib_memcpy_aligned(buf, &num, sizeof(num));
#endif
	return;
}


static inline uint16_t
aligned_read16be(const uint8_t *buf)
{
	uint16_t num = aligned_read16ne(buf);
	return conv16be(num);
}


static inline uint16_t
aligned_read16le(const uint8_t *buf)
{
	uint16_t num = aligned_read16ne(buf);
	return conv16le(num);
}


static inline uint32_t
aligned_read32be(const uint8_t *buf)
{
	uint32_t num = aligned_read32ne(buf);
	return conv32be(num);
}


static inline uint32_t
aligned_read32le(const uint8_t *buf)
{
	uint32_t num = aligned_read32ne(buf);
	return conv32le(num);
}


static inline uint64_t
aligned_read64be(const uint8_t *buf)
{
	uint64_t num = aligned_read64ne(buf);
	return conv64be(num);
}


static inline uint64_t
aligned_read64le(const uint8_t *buf)
{
	uint64_t num = aligned_read64ne(buf);
	return conv64le(num);
}


// These need to be macros like in the unaligned case.
#define aligned_write16be(buf, num) aligned_write16ne((buf), conv16be(num))
#define aligned_write16le(buf, num) aligned_write16ne((buf), conv16le(num))
#define aligned_write32be(buf, num) aligned_write32ne((buf), conv32be(num))
#define aligned_write32le(buf, num) aligned_write32ne((buf), conv32le(num))
#define aligned_write64be(buf, num) aligned_write64ne((buf), conv64be(num))
#define aligned_write64le(buf, num) aligned_write64ne((buf), conv64le(num))


////////////////////
// Bit operations //
////////////////////

static inline uint32_t
bsr32(uint32_t n)
{
	// Check for ICC first, since it tends to define __GNUC__ too.
#if defined(__INTEL_COMPILER)
	return _bit_scan_reverse(n);

#elif TUKLIB_GNUC_REQ(3, 4) && UINT_MAX == UINT32_MAX
	// GCC >= 3.4 has __builtin_clz(), which gives good results on
	// multiple architectures. On x86, __builtin_clz() ^ 31U becomes
	// either plain BSR (so the XOR gets optimized away) or LZCNT and
	// XOR (if -march indicates that SSE4a instructions are supported).
	return (uint32_t)__builtin_clz(n) ^ 31U;

#elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
	uint32_t i;
	__asm__("bsrl %1, %0" : "=r" (i) : "rm" (n));
	return i;

#elif defined(_MSC_VER)
	unsigned long i;
	_BitScanReverse(&i, n);
	return i;

#else
	uint32_t i = 31;

	if ((n & 0xFFFF0000) == 0) {
		n <<= 16;
		i = 15;
	}

	if ((n & 0xFF000000) == 0) {
		n <<= 8;
		i -= 8;
	}

	if ((n & 0xF0000000) == 0) {
		n <<= 4;
		i -= 4;
	}

	if ((n & 0xC0000000) == 0) {
		n <<= 2;
		i -= 2;
	}

	if ((n & 0x80000000) == 0)
		--i;

	return i;
#endif
}


static inline uint32_t
clz32(uint32_t n)
{
#if defined(__INTEL_COMPILER)
	return _bit_scan_reverse(n) ^ 31U;

#elif TUKLIB_GNUC_REQ(3, 4) && UINT_MAX == UINT32_MAX
	return (uint32_t)__builtin_clz(n);

#elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
	uint32_t i;
	__asm__("bsrl %1, %0\n\t"
		"xorl $31, %0"
		: "=r" (i) : "rm" (n));
	return i;

#elif defined(_MSC_VER)
	unsigned long i;
	_BitScanReverse(&i, n);
	return i ^ 31U;

#else
	uint32_t i = 0;

	if ((n & 0xFFFF0000) == 0) {
		n <<= 16;
		i = 16;
	}

	if ((n & 0xFF000000) == 0) {
		n <<= 8;
		i += 8;
	}

	if ((n & 0xF0000000) == 0) {
		n <<= 4;
		i += 4;
	}

	if ((n & 0xC0000000) == 0) {
		n <<= 2;
		i += 2;
	}

	if ((n & 0x80000000) == 0)
		++i;

	return i;
#endif
}


static inline uint32_t
ctz32(uint32_t n)
{
#if defined(__INTEL_COMPILER)
	return _bit_scan_forward(n);

#elif TUKLIB_GNUC_REQ(3, 4) && UINT_MAX >= UINT32_MAX
	return (uint32_t)__builtin_ctz(n);

#elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
	uint32_t i;
	__asm__("bsfl %1, %0" : "=r" (i) : "rm" (n));
	return i;

#elif defined(_MSC_VER)
	unsigned long i;
	_BitScanForward(&i, n);
	return i;

#else
	uint32_t i = 0;

	if ((n & 0x0000FFFF) == 0) {
		n >>= 16;
		i = 16;
	}

	if ((n & 0x000000FF) == 0) {
		n >>= 8;
		i += 8;
	}

	if ((n & 0x0000000F) == 0) {
		n >>= 4;
		i += 4;
	}

	if ((n & 0x00000003) == 0) {
		n >>= 2;
		i += 2;
	}

	if ((n & 0x00000001) == 0)
		++i;

	return i;
#endif
}

#define bsf32 ctz32

#endif