ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
Revision: 1.77
Committed: Mon Jan 20 13:41:18 2020 UTC (4 years, 4 months ago) by root
Branch: MAIN
Changes since 1.76: +44 -8 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =head1 LIBECB - e-C-Builtins
2
3 =head2 ABOUT LIBECB
4
5 Libecb is currently a simple header file that doesn't require any
6 configuration to use or include in your project.
7
8 It's part of the e-suite of libraries, other members of which include
9 libev and libeio.
10
11 Its homepage can be found here:
12
13 http://software.schmorp.de/pkg/libecb
14
15 It mainly provides a number of wrappers around GCC built-ins, together
16 with replacement functions for other compilers. In addition to this,
17 it provides a number of other lowlevel C utilities, such as endianness
18 detection, byte swapping or bit rotations.
19
20 Or in other words, things that should be built into any standard C system,
21 but aren't, implemented as efficient as possible with GCC, and still
22 correct with other compilers.
23
24 More might come.
25
26 =head2 ABOUT THE HEADER
27
28 At the moment, all you have to do is copy F<ecb.h> somewhere where your
29 compiler can find it and include it:
30
31 #include <ecb.h>
32
33 The header should work fine for both C and C++ compilation, and gives you
34 all of F<inttypes.h> in addition to the ECB symbols.
35
36 There are currently no object files to link to - future versions might
37 come with an (optional) object code library to link against, to reduce
38 code size or gain access to additional features.
39
40 It also currently includes everything from F<inttypes.h>.
41
42 =head2 ABOUT THIS MANUAL / CONVENTIONS
43
44 This manual mainly describes each (public) function available after
45 including the F<ecb.h> header. The header might define other symbols than
46 these, but these are not part of the public API, and not supported in any
47 way.
48
49 When the manual mentions a "function" then this could be defined either as
50 as inline function, a macro, or an external symbol.
51
52 When functions use a concrete standard type, such as C<int> or
53 C<uint32_t>, then the corresponding function works only with that type. If
54 only a generic name is used (C<expr>, C<cond>, C<value> and so on), then
55 the corresponding function relies on C to implement the correct types, and
56 is usually implemented as a macro. Specifically, a "bool" in this manual
57 refers to any kind of boolean value, not a specific type.
58
59 =head2 TYPES / TYPE SUPPORT
60
61 ecb.h makes sure that the following types are defined (in the expected way):
62
63 int8_t uint8_
64 int16_t uint16_t
65 int32_t uint32_
66 int64_t uint64_t
67 int_fast8_t uint_fast8_t
68 int_fast16_t uint_fast16_t
69 int_fast32_t uint_fast32_t
70 int_fast64_t uint_fast64_t
71 intptr_t uintptr_t
72
73 The macro C<ECB_PTRSIZE> is defined to the size of a pointer on this
74 platform (currently C<4> or C<8>) and can be used in preprocessor
75 expressions.
76
77 For C<ptrdiff_t> and C<size_t> use C<stddef.h>/C<cstddef>.
78
79 =head2 LANGUAGE/ENVIRONMENT/COMPILER VERSIONS
80
81 All the following symbols expand to an expression that can be tested in
82 preprocessor instructions as well as treated as a boolean (use C<!!> to
83 ensure it's either C<0> or C<1> if you need that).
84
85 =over 4
86
87 =item ECB_C
88
89 True if the implementation defines the C<__STDC__> macro to a true value,
90 while not claiming to be C++.
91
92 =item ECB_C99
93
94 True if the implementation claims to be compliant to C99 (ISO/IEC
95 9899:1999) or any later version, while not claiming to be C++.
96
97 Note that later versions (ECB_C11) remove core features again (for
98 example, variable length arrays).
99
100 =item ECB_C11, ECB_C17
101
102 True if the implementation claims to be compliant to C11/C17 (ISO/IEC
103 9899:2011, :20187) or any later version, while not claiming to be C++.
104
105 =item ECB_CPP
106
107 True if the implementation defines the C<__cplusplus__> macro to a true
108 value, which is typically true for C++ compilers.
109
110 =item ECB_CPP11, ECB_CPP14, ECB_CPP17
111
112 True if the implementation claims to be compliant to C++11/C++14/C++17
113 (ISO/IEC 14882:2011, :2014, :2017) or any later version.
114
115 =item ECB_GCC_VERSION (major, minor)
116
117 Expands to a true value (suitable for testing in by the preprocessor)
118 if the compiler used is GNU C and the version is the given version, or
119 higher.
120
121 This macro tries to return false on compilers that claim to be GCC
122 compatible but aren't.
123
124 =item ECB_EXTERN_C
125
126 Expands to C<extern "C"> in C++, and a simple C<extern> in C.
127
128 This can be used to declare a single external C function:
129
130 ECB_EXTERN_C int printf (const char *format, ...);
131
132 =item ECB_EXTERN_C_BEG / ECB_EXTERN_C_END
133
134 These two macros can be used to wrap multiple C<extern "C"> definitions -
135 they expand to nothing in C.
136
137 They are most useful in header files:
138
139 ECB_EXTERN_C_BEG
140
141 int mycfun1 (int x);
142 int mycfun2 (int x);
143
144 ECB_EXTERN_C_END
145
146 =item ECB_STDFP
147
148 If this evaluates to a true value (suitable for testing in by the
149 preprocessor), then C<float> and C<double> use IEEE 754 single/binary32
150 and double/binary64 representations internally I<and> the endianness of
151 both types match the endianness of C<uint32_t> and C<uint64_t>.
152
153 This means you can just copy the bits of a C<float> (or C<double>) to an
154 C<uint32_t> (or C<uint64_t>) and get the raw IEEE 754 bit representation
155 without having to think about format or endianness.
156
157 This is true for basically all modern platforms, although F<ecb.h> might
158 not be able to deduce this correctly everywhere and might err on the safe
159 side.
160
161 =item ECB_AMD64, ECB_AMD64_X32
162
163 These two macros are defined to C<1> on the x86_64/amd64 ABI and the X32
164 ABI, respectively, and undefined elsewhere.
165
166 The designers of the new X32 ABI for some inexplicable reason decided to
167 make it look exactly like amd64, even though it's completely incompatible
168 to that ABI, breaking about every piece of software that assumed that
169 C<__x86_64> stands for, well, the x86-64 ABI, making these macros
170 necessary.
171
172 =back
173
174 =head2 MACRO TRICKERY
175
176 =over 4
177
178 =item ECB_CONCAT (a, b)
179
180 Expands any macros in C<a> and C<b>, then concatenates the result to form
181 a single token. This is mainly useful to form identifiers from components,
182 e.g.:
183
184 #define S1 str
185 #define S2 cpy
186
187 ECB_CONCAT (S1, S2)(dst, src); // == strcpy (dst, src);
188
189 =item ECB_STRINGIFY (arg)
190
191 Expands any macros in C<arg> and returns the stringified version of
192 it. This is mainly useful to get the contents of a macro in string form,
193 e.g.:
194
195 #define SQL_LIMIT 100
196 sql_exec ("select * from table limit " ECB_STRINGIFY (SQL_LIMIT));
197
198 =item ECB_STRINGIFY_EXPR (expr)
199
200 Like C<ECB_STRINGIFY>, but additionally evaluates C<expr> to make sure it
201 is a valid expression. This is useful to catch typos or cases where the
202 macro isn't available:
203
204 #include <errno.h>
205
206 ECB_STRINGIFY (EDOM); // "33" (on my system at least)
207 ECB_STRINGIFY_EXPR (EDOM); // "33"
208
209 // now imagine we had a typo:
210
211 ECB_STRINGIFY (EDAM); // "EDAM"
212 ECB_STRINGIFY_EXPR (EDAM); // error: EDAM undefined
213
214 =back
215
216 =head2 ATTRIBUTES
217
218 A major part of libecb deals with additional attributes that can be
219 assigned to functions, variables and sometimes even types - much like
220 C<const> or C<volatile> in C. They are implemented using either GCC
221 attributes or other compiler/language specific features. Attributes
222 declarations must be put before the whole declaration:
223
224 ecb_const int mysqrt (int a);
225 ecb_unused int i;
226
227 =over 4
228
229 =item ecb_unused
230
231 Marks a function or a variable as "unused", which simply suppresses a
232 warning by GCC when it detects it as unused. This is useful when you e.g.
233 declare a variable but do not always use it:
234
235 {
236 ecb_unused int var;
237
238 #ifdef SOMECONDITION
239 var = ...;
240 return var;
241 #else
242 return 0;
243 #endif
244 }
245
246 =item ecb_deprecated
247
248 Similar to C<ecb_unused>, but marks a function, variable or type as
249 deprecated. This makes some compilers warn when the type is used.
250
251 =item ecb_deprecated_message (message)
252
253 Same as C<ecb_deprecated>, but if possible, the specified diagnostic is
254 used instead of a generic depreciation message when the object is being
255 used.
256
257 =item ecb_inline
258
259 Expands either to (a compiler-specific equivalent of) C<static inline> or
260 to just C<static>, if inline isn't supported. It should be used to declare
261 functions that should be inlined, for code size or speed reasons.
262
263 Example: inline this function, it surely will reduce codesize.
264
265 ecb_inline int
266 negmul (int a, int b)
267 {
268 return - (a * b);
269 }
270
271 =item ecb_noinline
272
273 Prevents a function from being inlined - it might be optimised away, but
274 not inlined into other functions. This is useful if you know your function
275 is rarely called and large enough for inlining not to be helpful.
276
277 =item ecb_noreturn
278
279 Marks a function as "not returning, ever". Some typical functions that
280 don't return are C<exit> or C<abort> (which really works hard to not
281 return), and now you can make your own:
282
283 ecb_noreturn void
284 my_abort (const char *errline)
285 {
286 puts (errline);
287 abort ();
288 }
289
290 In this case, the compiler would probably be smart enough to deduce it on
291 its own, so this is mainly useful for declarations.
292
293 =item ecb_restrict
294
295 Expands to the C<restrict> keyword or equivalent on compilers that support
296 them, and to nothing on others. Must be specified on a pointer type or
297 an array index to indicate that the memory doesn't alias with any other
298 restricted pointer in the same scope.
299
300 Example: multiply a vector, and allow the compiler to parallelise the
301 loop, because it knows it doesn't overwrite input values.
302
303 void
304 multiply (ecb_restrict float *src,
305 ecb_restrict float *dst,
306 int len, float factor)
307 {
308 int i;
309
310 for (i = 0; i < len; ++i)
311 dst [i] = src [i] * factor;
312 }
313
314 =item ecb_const
315
316 Declares that the function only depends on the values of its arguments,
317 much like a mathematical function. It specifically does not read or write
318 any memory any arguments might point to, global variables, or call any
319 non-const functions. It also must not have any side effects.
320
321 Such a function can be optimised much more aggressively by the compiler -
322 for example, multiple calls with the same arguments can be optimised into
323 a single call, which wouldn't be possible if the compiler would have to
324 expect any side effects.
325
326 It is best suited for functions in the sense of mathematical functions,
327 such as a function returning the square root of its input argument.
328
329 Not suited would be a function that calculates the hash of some memory
330 area you pass in, prints some messages or looks at a global variable to
331 decide on rounding.
332
333 See C<ecb_pure> for a slightly less restrictive class of functions.
334
335 =item ecb_pure
336
337 Similar to C<ecb_const>, declares a function that has no side
338 effects. Unlike C<ecb_const>, the function is allowed to examine global
339 variables and any other memory areas (such as the ones passed to it via
340 pointers).
341
342 While these functions cannot be optimised as aggressively as C<ecb_const>
343 functions, they can still be optimised away in many occasions, and the
344 compiler has more freedom in moving calls to them around.
345
346 Typical examples for such functions would be C<strlen> or C<memcmp>. A
347 function that calculates the MD5 sum of some input and updates some MD5
348 state passed as argument would I<NOT> be pure, however, as it would modify
349 some memory area that is not the return value.
350
351 =item ecb_hot
352
353 This declares a function as "hot" with regards to the cache - the function
354 is used so often, that it is very beneficial to keep it in the cache if
355 possible.
356
357 The compiler reacts by trying to place hot functions near to each other in
358 memory.
359
360 Whether a function is hot or not often depends on the whole program,
361 and less on the function itself. C<ecb_cold> is likely more useful in
362 practise.
363
364 =item ecb_cold
365
366 The opposite of C<ecb_hot> - declares a function as "cold" with regards to
367 the cache, or in other words, this function is not called often, or not at
368 speed-critical times, and keeping it in the cache might be a waste of said
369 cache.
370
371 In addition to placing cold functions together (or at least away from hot
372 functions), this knowledge can be used in other ways, for example, the
373 function will be optimised for size, as opposed to speed, and codepaths
374 leading to calls to those functions can automatically be marked as if
375 C<ecb_expect_false> had been used to reach them.
376
377 Good examples for such functions would be error reporting functions, or
378 functions only called in exceptional or rare cases.
379
380 =item ecb_artificial
381
382 Declares the function as "artificial", in this case meaning that this
383 function is not really meant to be a function, but more like an accessor
384 - many methods in C++ classes are mere accessor functions, and having a
385 crash reported in such a method, or single-stepping through them, is not
386 usually so helpful, especially when it's inlined to just a few instructions.
387
388 Marking them as artificial will instruct the debugger about just this,
389 leading to happier debugging and thus happier lives.
390
391 Example: in some kind of smart-pointer class, mark the pointer accessor as
392 artificial, so that the whole class acts more like a pointer and less like
393 some C++ abstraction monster.
394
395 template<typename T>
396 struct my_smart_ptr
397 {
398 T *value;
399
400 ecb_artificial
401 operator T *()
402 {
403 return value;
404 }
405 };
406
407 =back
408
409 =head2 OPTIMISATION HINTS
410
411 =over 4
412
413 =item ECB_OPTIMIZE_SIZE
414
415 Is C<1> when the compiler optimizes for size, C<0> otherwise. This symbol
416 can also be defined before including F<ecb.h>, in which case it will be
417 unchanged.
418
419 =item bool ecb_is_constant (expr)
420
421 Returns true iff the expression can be deduced to be a compile-time
422 constant, and false otherwise.
423
424 For example, when you have a C<rndm16> function that returns a 16 bit
425 random number, and you have a function that maps this to a range from
426 0..n-1, then you could use this inline function in a header file:
427
428 ecb_inline uint32_t
429 rndm (uint32_t n)
430 {
431 return (n * (uint32_t)rndm16 ()) >> 16;
432 }
433
434 However, for powers of two, you could use a normal mask, but that is only
435 worth it if, at compile time, you can detect this case. This is the case
436 when the passed number is a constant and also a power of two (C<n & (n -
437 1) == 0>):
438
439 ecb_inline uint32_t
440 rndm (uint32_t n)
441 {
442 return is_constant (n) && !(n & (n - 1))
443 ? rndm16 () & (num - 1)
444 : (n * (uint32_t)rndm16 ()) >> 16;
445 }
446
447 =item ecb_expect (expr, value)
448
449 Evaluates C<expr> and returns it. In addition, it tells the compiler that
450 the C<expr> evaluates to C<value> a lot, which can be used for static
451 branch optimisations.
452
453 Usually, you want to use the more intuitive C<ecb_expect_true> and
454 C<ecb_expect_false> functions instead.
455
456 =item bool ecb_expect_true (cond)
457
458 =item bool ecb_expect_false (cond)
459
460 These two functions expect a expression that is true or false and return
461 C<1> or C<0>, respectively, so when used in the condition of an C<if> or
462 other conditional statement, it will not change the program:
463
464 /* these two do the same thing */
465 if (some_condition) ...;
466 if (ecb_expect_true (some_condition)) ...;
467
468 However, by using C<ecb_expect_true>, you tell the compiler that the
469 condition is likely to be true (and for C<ecb_expect_false>, that it is
470 unlikely to be true).
471
472 For example, when you check for a null pointer and expect this to be a
473 rare, exceptional, case, then use C<ecb_expect_false>:
474
475 void my_free (void *ptr)
476 {
477 if (ecb_expect_false (ptr == 0))
478 return;
479 }
480
481 Consequent use of these functions to mark away exceptional cases or to
482 tell the compiler what the hot path through a function is can increase
483 performance considerably.
484
485 You might know these functions under the name C<likely> and C<unlikely>
486 - while these are common aliases, we find that the expect name is easier
487 to understand when quickly skimming code. If you wish, you can use
488 C<ecb_likely> instead of C<ecb_expect_true> and C<ecb_unlikely> instead of
489 C<ecb_expect_false> - these are simply aliases.
490
491 A very good example is in a function that reserves more space for some
492 memory block (for example, inside an implementation of a string stream) -
493 each time something is added, you have to check for a buffer overrun, but
494 you expect that most checks will turn out to be false:
495
496 /* make sure we have "size" extra room in our buffer */
497 ecb_inline void
498 reserve (int size)
499 {
500 if (ecb_expect_false (current + size > end))
501 real_reserve_method (size); /* presumably noinline */
502 }
503
504 =item ecb_assume (cond)
505
506 Tries to tell the compiler that some condition is true, even if it's not
507 obvious. This is not a function, but a statement: it cannot be used in
508 another expression.
509
510 This can be used to teach the compiler about invariants or other
511 conditions that might improve code generation, but which are impossible to
512 deduce form the code itself.
513
514 For example, the example reservation function from the C<ecb_expect_false>
515 description could be written thus (only C<ecb_assume> was added):
516
517 ecb_inline void
518 reserve (int size)
519 {
520 if (ecb_expect_false (current + size > end))
521 real_reserve_method (size); /* presumably noinline */
522
523 ecb_assume (current + size <= end);
524 }
525
526 If you then call this function twice, like this:
527
528 reserve (10);
529 reserve (1);
530
531 Then the compiler I<might> be able to optimise out the second call
532 completely, as it knows that C<< current + 1 > end >> is false and the
533 call will never be executed.
534
535 =item ecb_unreachable ()
536
537 This function does nothing itself, except tell the compiler that it will
538 never be executed. Apart from suppressing a warning in some cases, this
539 function can be used to implement C<ecb_assume> or similar functionality.
540
541 =item ecb_prefetch (addr, rw, locality)
542
543 Tells the compiler to try to prefetch memory at the given C<addr>ess
544 for either reading (C<rw> = 0) or writing (C<rw> = 1). A C<locality> of
545 C<0> means that there will only be one access later, C<3> means that
546 the data will likely be accessed very often, and values in between mean
547 something... in between. The memory pointed to by the address does not
548 need to be accessible (it could be a null pointer for example), but C<rw>
549 and C<locality> must be compile-time constants.
550
551 This is a statement, not a function: you cannot use it as part of an
552 expression.
553
554 An obvious way to use this is to prefetch some data far away, in a big
555 array you loop over. This prefetches memory some 128 array elements later,
556 in the hope that it will be ready when the CPU arrives at that location.
557
558 int sum = 0;
559
560 for (i = 0; i < N; ++i)
561 {
562 sum += arr [i]
563 ecb_prefetch (arr + i + 128, 0, 0);
564 }
565
566 It's hard to predict how far to prefetch, and most CPUs that can prefetch
567 are often good enough to predict this kind of behaviour themselves. It
568 gets more interesting with linked lists, especially when you do some fair
569 processing on each list element:
570
571 for (node *n = start; n; n = n->next)
572 {
573 ecb_prefetch (n->next, 0, 0);
574 ... do medium amount of work with *n
575 }
576
577 After processing the node, (part of) the next node might already be in
578 cache.
579
580 =back
581
582 =head2 BIT FIDDLING / BIT WIZARDRY
583
584 =over 4
585
586 =item bool ecb_big_endian ()
587
588 =item bool ecb_little_endian ()
589
590 These two functions return true if the byte order is big endian
591 (most-significant byte first) or little endian (least-significant byte
592 first) respectively.
593
594 On systems that are neither, their return values are unspecified.
595
596 =item int ecb_ctz32 (uint32_t x)
597
598 =item int ecb_ctz64 (uint64_t x)
599
600 =item int ecb_ctz (T x) [C++]
601
602 Returns the index of the least significant bit set in C<x> (or
603 equivalently the number of bits set to 0 before the least significant bit
604 set), starting from 0. If C<x> is 0 the result is undefined.
605
606 For smaller types than C<uint32_t> you can safely use C<ecb_ctz32>.
607
608 The overloaded C++ C<ecb_ctz> function supports C<uint8_t>, C<uint16_t>,
609 C<uint32_t> and C<uint64_t> types.
610
611 For example:
612
613 ecb_ctz32 (3) = 0
614 ecb_ctz32 (6) = 1
615
616 =item bool ecb_is_pot32 (uint32_t x)
617
618 =item bool ecb_is_pot64 (uint32_t x)
619
620 =item bool ecb_is_pot (T x) [C++]
621
622 Returns true iff C<x> is a power of two or C<x == 0>.
623
624 For smaller types than C<uint32_t> you can safely use C<ecb_is_pot32>.
625
626 The overloaded C++ C<ecb_is_pot> function supports C<uint8_t>, C<uint16_t>,
627 C<uint32_t> and C<uint64_t> types.
628
629 =item int ecb_ld32 (uint32_t x)
630
631 =item int ecb_ld64 (uint64_t x)
632
633 =item int ecb_ld64 (T x) [C++]
634
635 Returns the index of the most significant bit set in C<x>, or the number
636 of digits the number requires in binary (so that C<< 2**ld <= x <
637 2**(ld+1) >>). If C<x> is 0 the result is undefined. A common use case is
638 to compute the integer binary logarithm, i.e. C<floor (log2 (n))>, for
639 example to see how many bits a certain number requires to be encoded.
640
641 This function is similar to the "count leading zero bits" function, except
642 that that one returns how many zero bits are "in front" of the number (in
643 the given data type), while C<ecb_ld> returns how many bits the number
644 itself requires.
645
646 For smaller types than C<uint32_t> you can safely use C<ecb_ld32>.
647
648 The overloaded C++ C<ecb_ld> function supports C<uint8_t>, C<uint16_t>,
649 C<uint32_t> and C<uint64_t> types.
650
651 =item int ecb_popcount32 (uint32_t x)
652
653 =item int ecb_popcount64 (uint64_t x)
654
655 =item int ecb_popcount (T x) [C++]
656
657 Returns the number of bits set to 1 in C<x>.
658
659 For smaller types than C<uint32_t> you can safely use C<ecb_popcount32>.
660
661 The overloaded C++ C<ecb_popcount> function supports C<uint8_t>, C<uint16_t>,
662 C<uint32_t> and C<uint64_t> types.
663
664 For example:
665
666 ecb_popcount32 (7) = 3
667 ecb_popcount32 (255) = 8
668
669 =item uint8_t ecb_bitrev8 (uint8_t x)
670
671 =item uint16_t ecb_bitrev16 (uint16_t x)
672
673 =item uint32_t ecb_bitrev32 (uint32_t x)
674
675 =item T ecb_bitrev (T x) [C++]
676
677 Reverses the bits in x, i.e. the MSB becomes the LSB, MSB-1 becomes LSB+1
678 and so on.
679
680 The overloaded C++ C<ecb_bitrev> function supports C<uint8_t>, C<uint16_t> and C<uint32_t> types.
681
682 Example:
683
684 ecb_bitrev8 (0xa7) = 0xea
685 ecb_bitrev32 (0xffcc4411) = 0x882233ff
686
687 =item T ecb_bitrev (T x) [C++]
688
689 Overloaded C++ bitrev function.
690
691 C<T> must be one of C<uint8_t>, C<uint16_t> or C<uint32_t>.
692
693 =item uint32_t ecb_bswap16 (uint32_t x)
694
695 =item uint32_t ecb_bswap32 (uint32_t x)
696
697 =item uint64_t ecb_bswap64 (uint64_t x)
698
699 These functions return the value of the 16-bit (32-bit, 64-bit) value
700 C<x> after reversing the order of bytes (0x11223344 becomes 0x44332211 in
701 C<ecb_bswap32>).
702
703 The overloaded C++ C<ecb_bswap> function supports C<uint8_t>, C<uint16_t>,
704 C<uint32_t> and C<uint64_t> types.
705
706 =item uint8_t ecb_rotl8 (uint8_t x, unsigned int count)
707
708 =item uint16_t ecb_rotl16 (uint16_t x, unsigned int count)
709
710 =item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
711
712 =item uint64_t ecb_rotl64 (uint64_t x, unsigned int count)
713
714 =item uint8_t ecb_rotr8 (uint8_t x, unsigned int count)
715
716 =item uint16_t ecb_rotr16 (uint16_t x, unsigned int count)
717
718 =item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
719
720 =item uint64_t ecb_rotr64 (uint64_t x, unsigned int count)
721
722 These two families of functions return the value of C<x> after rotating
723 all the bits by C<count> positions to the right (C<ecb_rotr>) or left
724 (C<ecb_rotl>).
725
726 Current GCC versions understand these functions and usually compile them
727 to "optimal" code (e.g. a single C<rol> or a combination of C<shld> on
728 x86).
729
730 =item T ecb_rotl (T x, unsigned int count) [C++]
731
732 =item T ecb_rotr (T x, unsigned int count) [C++]
733
734 Overloaded C++ rotl/rotr functions.
735
736 C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>.
737
738 =back
739
740 =head2 HOST ENDIANNESS CONVERSION
741
742 =over 4
743
744 =item uint_fast16_t ecb_be_u16_to_host (uint_fast16_t v)
745
746 =item uint_fast32_t ecb_be_u32_to_host (uint_fast32_t v)
747
748 =item uint_fast64_t ecb_be_u64_to_host (uint_fast64_t v)
749
750 =item uint_fast16_t ecb_le_u16_to_host (uint_fast16_t v)
751
752 =item uint_fast32_t ecb_le_u32_to_host (uint_fast32_t v)
753
754 =item uint_fast64_t ecb_le_u64_to_host (uint_fast64_t v)
755
756 Convert an unsigned 16, 32 or 64 bit value from big or little endian to host byte order.
757
758 The naming convention is C<ecb_>(C<be>|C<le>)C<_u>C<16|32|64>C<_to_host>,
759 where be and le stand for big endian and little endian, respectively.
760
761 =item uint_fast16_t ecb_host_to_be_u16 (uint_fast16_t v)
762
763 =item uint_fast32_t ecb_host_to_be_u32 (uint_fast32_t v)
764
765 =item uint_fast64_t ecb_host_to_be_u64 (uint_fast64_t v)
766
767 =item uint_fast16_t ecb_host_to_le_u16 (uint_fast16_t v)
768
769 =item uint_fast32_t ecb_host_to_le_u32 (uint_fast32_t v)
770
771 =item uint_fast64_t ecb_host_to_le_u64 (uint_fast64_t v)
772
773 Like above, but converts I<from> host byte order to the specified
774 endianness.
775
776 =back
777
778 In C++ the following additional template functions are supported:
779
780 =over 4
781
782 =item T ecb_be_to_host (T v)
783
784 =item T ecb_le_to_host (T v)
785
786 =item T ecb_host_to_be (T v)
787
788 =item T ecb_host_to_le (T v)
789
790 These functions work like their C counterparts, above, but use templates,
791 which make them useful in generic code.
792
793 C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>
794 (so unlike their C counterparts, there is a version for C<uint8_t>, which
795 again can be useful in generic code).
796
797 =head2 UNALIGNED LOAD/STORE
798
799 These function load or store unaligned multi-byte values.
800
801 =over 4
802
803 =item uint_fast16_t ecb_peek_u16_u (const void *ptr)
804
805 =item uint_fast32_t ecb_peek_u32_u (const void *ptr)
806
807 =item uint_fast64_t ecb_peek_u64_u (const void *ptr)
808
809 These functions load an unaligned, unsigned 16, 32 or 64 bit value from
810 memory.
811
812 =item uint_fast16_t ecb_peek_be_u16_u (const void *ptr)
813
814 =item uint_fast32_t ecb_peek_be_u32_u (const void *ptr)
815
816 =item uint_fast64_t ecb_peek_be_u64_u (const void *ptr)
817
818 =item uint_fast16_t ecb_peek_le_u16_u (const void *ptr)
819
820 =item uint_fast32_t ecb_peek_le_u32_u (const void *ptr)
821
822 =item uint_fast64_t ecb_peek_le_u64_u (const void *ptr)
823
824 Like above, but additionally convert from big endian (C<be>) or little
825 endian (C<le>) byte order to host byte order while doing so.
826
827 =item ecb_poke_u16_u (void *ptr, uint16_t v)
828
829 =item ecb_poke_u32_u (void *ptr, uint32_t v)
830
831 =item ecb_poke_u64_u (void *ptr, uint64_t v)
832
833 These functions store an unaligned, unsigned 16, 32 or 64 bit value to
834 memory.
835
836 =item ecb_poke_be_u16_u (void *ptr, uint_fast16_t v)
837
838 =item ecb_poke_be_u32_u (void *ptr, uint_fast32_t v)
839
840 =item ecb_poke_be_u64_u (void *ptr, uint_fast64_t v)
841
842 =item ecb_poke_le_u16_u (void *ptr, uint_fast16_t v)
843
844 =item ecb_poke_le_u32_u (void *ptr, uint_fast32_t v)
845
846 =item ecb_poke_le_u64_u (void *ptr, uint_fast64_t v)
847
848 Like above, but additionally convert from host byte order to big endian
849 (C<be>) or little endian (C<le>) byte order while doing so.
850
851 =back
852
853 In C++ the following additional template functions are supported:
854
855 =over 4
856
857 =item T ecb_peek (const void *ptr)
858
859 =item T ecb_peek_be (const void *ptr)
860
861 =item T ecb_peek_le (const void *ptr)
862
863 =item T ecb_peek_u (const void *ptr)
864
865 =item T ecb_peek_be_u (const void *ptr)
866
867 =item T ecb_peek_le_u (const void *ptr)
868
869 Similarly to their C counterparts, these functions load an unsigned 8, 16,
870 32 or 64 bit value from memory, with optional conversion from big/little
871 endian.
872
873 Since the type cannot be deduced, it has top be specified explicitly, e.g.
874
875 uint_fast16_t v = ecb_peek<uint16_t> (ptr);
876
877 C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>.
878
879 Unlike their C counterparts, these functions support 8 bit quantities
880 (C<uint8_t>) and also have an aligned version (without the C<_u> prefix),
881 all of which hopefully makes them more useful in generic code.
882
883 =item ecb_poke (void *ptr, T v)
884
885 =item ecb_poke_be (void *ptr, T v)
886
887 =item ecb_poke_le (void *ptr, T v)
888
889 =item ecb_poke_u (void *ptr, T v)
890
891 =item ecb_poke_be_u (void *ptr, T v)
892
893 =item ecb_poke_le_u (void *ptr, T v)
894
895 Again, similarly to their C counterparts, these functions store an
896 unsigned 8, 16, 32 or z64 bit value to memory, with optional conversion to
897 big/little endian.
898
899 C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>.
900
901 Unlike their C counterparts, these functions support 8 bit quantities
902 (C<uint8_t>) and also have an aligned version (without the C<_u> prefix),
903 all of which hopefully makes them more useful in generic code.
904
905 =back
906
907 =head2 FLOATING POINT FIDDLING
908
909 =over 4
910
911 =item ECB_INFINITY [-UECB_NO_LIBM]
912
913 Evaluates to positive infinity if supported by the platform, otherwise to
914 a truly huge number.
915
916 =item ECB_NAN [-UECB_NO_LIBM]
917
918 Evaluates to a quiet NAN if supported by the platform, otherwise to
919 C<ECB_INFINITY>.
920
921 =item float ecb_ldexpf (float x, int exp) [-UECB_NO_LIBM]
922
923 Same as C<ldexpf>, but always available.
924
925 =item uint32_t ecb_float_to_binary16 (float x) [-UECB_NO_LIBM]
926
927 =item uint32_t ecb_float_to_binary32 (float x) [-UECB_NO_LIBM]
928
929 =item uint64_t ecb_double_to_binary64 (double x) [-UECB_NO_LIBM]
930
931 These functions each take an argument in the native C<float> or C<double>
932 type and return the IEEE 754 bit representation of it (binary16/half,
933 binary32/single or binary64/double precision).
934
935 The bit representation is just as IEEE 754 defines it, i.e. the sign bit
936 will be the most significant bit, followed by exponent and mantissa.
937
938 This function should work even when the native floating point format isn't
939 IEEE compliant, of course at a speed and code size penalty, and of course
940 also within reasonable limits (it tries to convert NaNs, infinities and
941 denormals, but will likely convert negative zero to positive zero).
942
943 On all modern platforms (where C<ECB_STDFP> is true), the compiler should
944 be able to optimise away this function completely.
945
946 These functions can be helpful when serialising floats to the network - you
947 can serialise the return value like a normal uint16_t/uint32_t/uint64_t.
948
949 Another use for these functions is to manipulate floating point values
950 directly.
951
952 Silly example: toggle the sign bit of a float.
953
954 /* On gcc-4.7 on amd64, */
955 /* this results in a single add instruction to toggle the bit, and 4 extra */
956 /* instructions to move the float value to an integer register and back. */
957
958 x = ecb_binary32_to_float (ecb_float_to_binary32 (x) ^ 0x80000000U)
959
960 =item float ecb_binary16_to_float (uint16_t x) [-UECB_NO_LIBM]
961
962 =item float ecb_binary32_to_float (uint32_t x) [-UECB_NO_LIBM]
963
964 =item double ecb_binary64_to_double (uint64_t x) [-UECB_NO_LIBM]
965
966 The reverse operation of the previous function - takes the bit
967 representation of an IEEE binary16, binary32 or binary64 number (half,
968 single or double precision) and converts it to the native C<float> or
969 C<double> format.
970
971 This function should work even when the native floating point format isn't
972 IEEE compliant, of course at a speed and code size penalty, and of course
973 also within reasonable limits (it tries to convert normals and denormals,
974 and might be lucky for infinities, and with extraordinary luck, also for
975 negative zero).
976
977 On all modern platforms (where C<ECB_STDFP> is true), the compiler should
978 be able to optimise away this function completely.
979
980 =item uint16_t ecb_binary32_to_binary16 (uint32_t x)
981
982 =item uint32_t ecb_binary16_to_binary32 (uint16_t x)
983
984 Convert a IEEE binary32/single precision to binary16/half format, and vice
985 versa, handling all details (round-to-nearest-even, subnormals, infinity
986 and NaNs) correctly.
987
988 These are functions are available under C<-DECB_NO_LIBM>, since
989 they do not rely on the platform floating point format. The
990 C<ecb_float_to_binary16> and C<ecb_binary16_to_float> functions are
991 usually what you want.
992
993 =back
994
995 =head2 ARITHMETIC
996
997 =over 4
998
999 =item x = ecb_mod (m, n)
1000
1001 Returns C<m> modulo C<n>, which is the same as the positive remainder
1002 of the division operation between C<m> and C<n>, using floored
1003 division. Unlike the C remainder operator C<%>, this function ensures that
1004 the return value is always positive and that the two numbers I<m> and
1005 I<m' = m + i * n> result in the same value modulo I<n> - in other words,
1006 C<ecb_mod> implements the mathematical modulo operation, which is missing
1007 in the language.
1008
1009 C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
1010 negatable, that is, both C<m> and C<-m> must be representable in its
1011 type (this typically excludes the minimum signed integer value, the same
1012 limitation as for C</> and C<%> in C).
1013
1014 Current GCC versions compile this into an efficient branchless sequence on
1015 almost all CPUs.
1016
1017 For example, when you want to rotate forward through the members of an
1018 array for increasing C<m> (which might be negative), then you should use
1019 C<ecb_mod>, as the C<%> operator might give either negative results, or
1020 change direction for negative values:
1021
1022 for (m = -100; m <= 100; ++m)
1023 int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
1024
1025 =item x = ecb_div_rd (val, div)
1026
1027 =item x = ecb_div_ru (val, div)
1028
1029 Returns C<val> divided by C<div> rounded down or up, respectively.
1030 C<val> and C<div> must have integer types and C<div> must be strictly
1031 positive. Note that these functions are implemented with macros in C
1032 and with function templates in C++.
1033
1034 =back
1035
1036 =head2 UTILITY
1037
1038 =over 4
1039
1040 =item element_count = ecb_array_length (name)
1041
1042 Returns the number of elements in the array C<name>. For example:
1043
1044 int primes[] = { 2, 3, 5, 7, 11 };
1045 int sum = 0;
1046
1047 for (i = 0; i < ecb_array_length (primes); i++)
1048 sum += primes [i];
1049
1050 =back
1051
1052 =head2 SYMBOLS GOVERNING COMPILATION OF ECB.H ITSELF
1053
1054 These symbols need to be defined before including F<ecb.h> the first time.
1055
1056 =over 4
1057
1058 =item ECB_NO_THREADS
1059
1060 If F<ecb.h> is never used from multiple threads, then this symbol can
1061 be defined, in which case memory fences (and similar constructs) are
1062 completely removed, leading to more efficient code and fewer dependencies.
1063
1064 Setting this symbol to a true value implies C<ECB_NO_SMP>.
1065
1066 =item ECB_NO_SMP
1067
1068 The weaker version of C<ECB_NO_THREADS> - if F<ecb.h> is used from
1069 multiple threads, but never concurrently (e.g. if the system the program
1070 runs on has only a single CPU with a single core, no hyperthreading and so
1071 on), then this symbol can be defined, leading to more efficient code and
1072 fewer dependencies.
1073
1074 =item ECB_NO_LIBM
1075
1076 When defined to C<1>, do not export any functions that might introduce
1077 dependencies on the math library (usually called F<-lm>) - these are
1078 marked with [-UECB_NO_LIBM].
1079
1080 =back
1081
1082 =head1 UNDOCUMENTED FUNCTIONALITY
1083
1084 F<ecb.h> is full of undocumented functionality as well, some of which is
1085 intended to be internal-use only, some of which we forgot to document, and
1086 some of which we hide because we are not sure we will keep the interface
1087 stable.
1088
1089 While you are welcome to rummage around and use whatever you find useful
1090 (we can't stop you), keep in mind that we will change undocumented
1091 functionality in incompatible ways without thinking twice, while we are
1092 considerably more conservative with documented things.
1093
1094 =head1 AUTHORS
1095
1096 C<libecb> is designed and maintained by:
1097
1098 Emanuele Giaquinta <e.giaquinta@glauco.it>
1099 Marc Alexander Lehmann <schmorp@schmorp.de>
1100
1101