ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
Revision: 1.75
Committed: Sat Dec 28 08:01:05 2019 UTC (4 years, 5 months ago) by root
Branch: MAIN
Changes since 1.74: +6 -0 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_t int16_t uint16_t
64 int32_t uint32_t int64_t uint64_t
65 intptr_t uintptr_t
66
67 The macro C<ECB_PTRSIZE> is defined to the size of a pointer on this
68 platform (currently C<4> or C<8>) and can be used in preprocessor
69 expressions.
70
71 For C<ptrdiff_t> and C<size_t> use C<stddef.h>/C<cstddef>.
72
73 =head2 LANGUAGE/ENVIRONMENT/COMPILER VERSIONS
74
75 All the following symbols expand to an expression that can be tested in
76 preprocessor instructions as well as treated as a boolean (use C<!!> to
77 ensure it's either C<0> or C<1> if you need that).
78
79 =over 4
80
81 =item ECB_C
82
83 True if the implementation defines the C<__STDC__> macro to a true value,
84 while not claiming to be C++.
85
86 =item ECB_C99
87
88 True if the implementation claims to be compliant to C99 (ISO/IEC
89 9899:1999) or any later version, while not claiming to be C++.
90
91 Note that later versions (ECB_C11) remove core features again (for
92 example, variable length arrays).
93
94 =item ECB_C11, ECB_C17
95
96 True if the implementation claims to be compliant to C11/C17 (ISO/IEC
97 9899:2011, :20187) or any later version, while not claiming to be C++.
98
99 =item ECB_CPP
100
101 True if the implementation defines the C<__cplusplus__> macro to a true
102 value, which is typically true for C++ compilers.
103
104 =item ECB_CPP11, ECB_CPP14, ECB_CPP17
105
106 True if the implementation claims to be compliant to C++11/C++14/C++17
107 (ISO/IEC 14882:2011, :2014, :2017) or any later version.
108
109 =item ECB_GCC_VERSION (major, minor)
110
111 Expands to a true value (suitable for testing in by the preprocessor)
112 if the compiler used is GNU C and the version is the given version, or
113 higher.
114
115 This macro tries to return false on compilers that claim to be GCC
116 compatible but aren't.
117
118 =item ECB_EXTERN_C
119
120 Expands to C<extern "C"> in C++, and a simple C<extern> in C.
121
122 This can be used to declare a single external C function:
123
124 ECB_EXTERN_C int printf (const char *format, ...);
125
126 =item ECB_EXTERN_C_BEG / ECB_EXTERN_C_END
127
128 These two macros can be used to wrap multiple C<extern "C"> definitions -
129 they expand to nothing in C.
130
131 They are most useful in header files:
132
133 ECB_EXTERN_C_BEG
134
135 int mycfun1 (int x);
136 int mycfun2 (int x);
137
138 ECB_EXTERN_C_END
139
140 =item ECB_STDFP
141
142 If this evaluates to a true value (suitable for testing in by the
143 preprocessor), then C<float> and C<double> use IEEE 754 single/binary32
144 and double/binary64 representations internally I<and> the endianness of
145 both types match the endianness of C<uint32_t> and C<uint64_t>.
146
147 This means you can just copy the bits of a C<float> (or C<double>) to an
148 C<uint32_t> (or C<uint64_t>) and get the raw IEEE 754 bit representation
149 without having to think about format or endianness.
150
151 This is true for basically all modern platforms, although F<ecb.h> might
152 not be able to deduce this correctly everywhere and might err on the safe
153 side.
154
155 =item ECB_AMD64, ECB_AMD64_X32
156
157 These two macros are defined to C<1> on the x86_64/amd64 ABI and the X32
158 ABI, respectively, and undefined elsewhere.
159
160 The designers of the new X32 ABI for some inexplicable reason decided to
161 make it look exactly like amd64, even though it's completely incompatible
162 to that ABI, breaking about every piece of software that assumed that
163 C<__x86_64> stands for, well, the x86-64 ABI, making these macros
164 necessary.
165
166 =back
167
168 =head2 MACRO TRICKERY
169
170 =over 4
171
172 =item ECB_CONCAT (a, b)
173
174 Expands any macros in C<a> and C<b>, then concatenates the result to form
175 a single token. This is mainly useful to form identifiers from components,
176 e.g.:
177
178 #define S1 str
179 #define S2 cpy
180
181 ECB_CONCAT (S1, S2)(dst, src); // == strcpy (dst, src);
182
183 =item ECB_STRINGIFY (arg)
184
185 Expands any macros in C<arg> and returns the stringified version of
186 it. This is mainly useful to get the contents of a macro in string form,
187 e.g.:
188
189 #define SQL_LIMIT 100
190 sql_exec ("select * from table limit " ECB_STRINGIFY (SQL_LIMIT));
191
192 =item ECB_STRINGIFY_EXPR (expr)
193
194 Like C<ECB_STRINGIFY>, but additionally evaluates C<expr> to make sure it
195 is a valid expression. This is useful to catch typos or cases where the
196 macro isn't available:
197
198 #include <errno.h>
199
200 ECB_STRINGIFY (EDOM); // "33" (on my system at least)
201 ECB_STRINGIFY_EXPR (EDOM); // "33"
202
203 // now imagine we had a typo:
204
205 ECB_STRINGIFY (EDAM); // "EDAM"
206 ECB_STRINGIFY_EXPR (EDAM); // error: EDAM undefined
207
208 =back
209
210 =head2 ATTRIBUTES
211
212 A major part of libecb deals with additional attributes that can be
213 assigned to functions, variables and sometimes even types - much like
214 C<const> or C<volatile> in C. They are implemented using either GCC
215 attributes or other compiler/language specific features. Attributes
216 declarations must be put before the whole declaration:
217
218 ecb_const int mysqrt (int a);
219 ecb_unused int i;
220
221 =over 4
222
223 =item ecb_unused
224
225 Marks a function or a variable as "unused", which simply suppresses a
226 warning by GCC when it detects it as unused. This is useful when you e.g.
227 declare a variable but do not always use it:
228
229 {
230 ecb_unused int var;
231
232 #ifdef SOMECONDITION
233 var = ...;
234 return var;
235 #else
236 return 0;
237 #endif
238 }
239
240 =item ecb_deprecated
241
242 Similar to C<ecb_unused>, but marks a function, variable or type as
243 deprecated. This makes some compilers warn when the type is used.
244
245 =item ecb_deprecated_message (message)
246
247 Same as C<ecb_deprecated>, but if possible, the specified diagnostic is
248 used instead of a generic depreciation message when the object is being
249 used.
250
251 =item ecb_inline
252
253 Expands either to (a compiler-specific equivalent of) C<static inline> or
254 to just C<static>, if inline isn't supported. It should be used to declare
255 functions that should be inlined, for code size or speed reasons.
256
257 Example: inline this function, it surely will reduce codesize.
258
259 ecb_inline int
260 negmul (int a, int b)
261 {
262 return - (a * b);
263 }
264
265 =item ecb_noinline
266
267 Prevents a function from being inlined - it might be optimised away, but
268 not inlined into other functions. This is useful if you know your function
269 is rarely called and large enough for inlining not to be helpful.
270
271 =item ecb_noreturn
272
273 Marks a function as "not returning, ever". Some typical functions that
274 don't return are C<exit> or C<abort> (which really works hard to not
275 return), and now you can make your own:
276
277 ecb_noreturn void
278 my_abort (const char *errline)
279 {
280 puts (errline);
281 abort ();
282 }
283
284 In this case, the compiler would probably be smart enough to deduce it on
285 its own, so this is mainly useful for declarations.
286
287 =item ecb_restrict
288
289 Expands to the C<restrict> keyword or equivalent on compilers that support
290 them, and to nothing on others. Must be specified on a pointer type or
291 an array index to indicate that the memory doesn't alias with any other
292 restricted pointer in the same scope.
293
294 Example: multiply a vector, and allow the compiler to parallelise the
295 loop, because it knows it doesn't overwrite input values.
296
297 void
298 multiply (ecb_restrict float *src,
299 ecb_restrict float *dst,
300 int len, float factor)
301 {
302 int i;
303
304 for (i = 0; i < len; ++i)
305 dst [i] = src [i] * factor;
306 }
307
308 =item ecb_const
309
310 Declares that the function only depends on the values of its arguments,
311 much like a mathematical function. It specifically does not read or write
312 any memory any arguments might point to, global variables, or call any
313 non-const functions. It also must not have any side effects.
314
315 Such a function can be optimised much more aggressively by the compiler -
316 for example, multiple calls with the same arguments can be optimised into
317 a single call, which wouldn't be possible if the compiler would have to
318 expect any side effects.
319
320 It is best suited for functions in the sense of mathematical functions,
321 such as a function returning the square root of its input argument.
322
323 Not suited would be a function that calculates the hash of some memory
324 area you pass in, prints some messages or looks at a global variable to
325 decide on rounding.
326
327 See C<ecb_pure> for a slightly less restrictive class of functions.
328
329 =item ecb_pure
330
331 Similar to C<ecb_const>, declares a function that has no side
332 effects. Unlike C<ecb_const>, the function is allowed to examine global
333 variables and any other memory areas (such as the ones passed to it via
334 pointers).
335
336 While these functions cannot be optimised as aggressively as C<ecb_const>
337 functions, they can still be optimised away in many occasions, and the
338 compiler has more freedom in moving calls to them around.
339
340 Typical examples for such functions would be C<strlen> or C<memcmp>. A
341 function that calculates the MD5 sum of some input and updates some MD5
342 state passed as argument would I<NOT> be pure, however, as it would modify
343 some memory area that is not the return value.
344
345 =item ecb_hot
346
347 This declares a function as "hot" with regards to the cache - the function
348 is used so often, that it is very beneficial to keep it in the cache if
349 possible.
350
351 The compiler reacts by trying to place hot functions near to each other in
352 memory.
353
354 Whether a function is hot or not often depends on the whole program,
355 and less on the function itself. C<ecb_cold> is likely more useful in
356 practise.
357
358 =item ecb_cold
359
360 The opposite of C<ecb_hot> - declares a function as "cold" with regards to
361 the cache, or in other words, this function is not called often, or not at
362 speed-critical times, and keeping it in the cache might be a waste of said
363 cache.
364
365 In addition to placing cold functions together (or at least away from hot
366 functions), this knowledge can be used in other ways, for example, the
367 function will be optimised for size, as opposed to speed, and codepaths
368 leading to calls to those functions can automatically be marked as if
369 C<ecb_expect_false> had been used to reach them.
370
371 Good examples for such functions would be error reporting functions, or
372 functions only called in exceptional or rare cases.
373
374 =item ecb_artificial
375
376 Declares the function as "artificial", in this case meaning that this
377 function is not really meant to be a function, but more like an accessor
378 - many methods in C++ classes are mere accessor functions, and having a
379 crash reported in such a method, or single-stepping through them, is not
380 usually so helpful, especially when it's inlined to just a few instructions.
381
382 Marking them as artificial will instruct the debugger about just this,
383 leading to happier debugging and thus happier lives.
384
385 Example: in some kind of smart-pointer class, mark the pointer accessor as
386 artificial, so that the whole class acts more like a pointer and less like
387 some C++ abstraction monster.
388
389 template<typename T>
390 struct my_smart_ptr
391 {
392 T *value;
393
394 ecb_artificial
395 operator T *()
396 {
397 return value;
398 }
399 };
400
401 =back
402
403 =head2 OPTIMISATION HINTS
404
405 =over 4
406
407 =item ECB_OPTIMIZE_SIZE
408
409 Is C<1> when the compiler optimizes for size, C<0> otherwise. This symbol
410 can also be defined before including F<ecb.h>, in which case it will be
411 unchanged.
412
413 =item bool ecb_is_constant (expr)
414
415 Returns true iff the expression can be deduced to be a compile-time
416 constant, and false otherwise.
417
418 For example, when you have a C<rndm16> function that returns a 16 bit
419 random number, and you have a function that maps this to a range from
420 0..n-1, then you could use this inline function in a header file:
421
422 ecb_inline uint32_t
423 rndm (uint32_t n)
424 {
425 return (n * (uint32_t)rndm16 ()) >> 16;
426 }
427
428 However, for powers of two, you could use a normal mask, but that is only
429 worth it if, at compile time, you can detect this case. This is the case
430 when the passed number is a constant and also a power of two (C<n & (n -
431 1) == 0>):
432
433 ecb_inline uint32_t
434 rndm (uint32_t n)
435 {
436 return is_constant (n) && !(n & (n - 1))
437 ? rndm16 () & (num - 1)
438 : (n * (uint32_t)rndm16 ()) >> 16;
439 }
440
441 =item ecb_expect (expr, value)
442
443 Evaluates C<expr> and returns it. In addition, it tells the compiler that
444 the C<expr> evaluates to C<value> a lot, which can be used for static
445 branch optimisations.
446
447 Usually, you want to use the more intuitive C<ecb_expect_true> and
448 C<ecb_expect_false> functions instead.
449
450 =item bool ecb_expect_true (cond)
451
452 =item bool ecb_expect_false (cond)
453
454 These two functions expect a expression that is true or false and return
455 C<1> or C<0>, respectively, so when used in the condition of an C<if> or
456 other conditional statement, it will not change the program:
457
458 /* these two do the same thing */
459 if (some_condition) ...;
460 if (ecb_expect_true (some_condition)) ...;
461
462 However, by using C<ecb_expect_true>, you tell the compiler that the
463 condition is likely to be true (and for C<ecb_expect_false>, that it is
464 unlikely to be true).
465
466 For example, when you check for a null pointer and expect this to be a
467 rare, exceptional, case, then use C<ecb_expect_false>:
468
469 void my_free (void *ptr)
470 {
471 if (ecb_expect_false (ptr == 0))
472 return;
473 }
474
475 Consequent use of these functions to mark away exceptional cases or to
476 tell the compiler what the hot path through a function is can increase
477 performance considerably.
478
479 You might know these functions under the name C<likely> and C<unlikely>
480 - while these are common aliases, we find that the expect name is easier
481 to understand when quickly skimming code. If you wish, you can use
482 C<ecb_likely> instead of C<ecb_expect_true> and C<ecb_unlikely> instead of
483 C<ecb_expect_false> - these are simply aliases.
484
485 A very good example is in a function that reserves more space for some
486 memory block (for example, inside an implementation of a string stream) -
487 each time something is added, you have to check for a buffer overrun, but
488 you expect that most checks will turn out to be false:
489
490 /* make sure we have "size" extra room in our buffer */
491 ecb_inline void
492 reserve (int size)
493 {
494 if (ecb_expect_false (current + size > end))
495 real_reserve_method (size); /* presumably noinline */
496 }
497
498 =item ecb_assume (cond)
499
500 Tries to tell the compiler that some condition is true, even if it's not
501 obvious. This is not a function, but a statement: it cannot be used in
502 another expression.
503
504 This can be used to teach the compiler about invariants or other
505 conditions that might improve code generation, but which are impossible to
506 deduce form the code itself.
507
508 For example, the example reservation function from the C<ecb_expect_false>
509 description could be written thus (only C<ecb_assume> was added):
510
511 ecb_inline void
512 reserve (int size)
513 {
514 if (ecb_expect_false (current + size > end))
515 real_reserve_method (size); /* presumably noinline */
516
517 ecb_assume (current + size <= end);
518 }
519
520 If you then call this function twice, like this:
521
522 reserve (10);
523 reserve (1);
524
525 Then the compiler I<might> be able to optimise out the second call
526 completely, as it knows that C<< current + 1 > end >> is false and the
527 call will never be executed.
528
529 =item ecb_unreachable ()
530
531 This function does nothing itself, except tell the compiler that it will
532 never be executed. Apart from suppressing a warning in some cases, this
533 function can be used to implement C<ecb_assume> or similar functionality.
534
535 =item ecb_prefetch (addr, rw, locality)
536
537 Tells the compiler to try to prefetch memory at the given C<addr>ess
538 for either reading (C<rw> = 0) or writing (C<rw> = 1). A C<locality> of
539 C<0> means that there will only be one access later, C<3> means that
540 the data will likely be accessed very often, and values in between mean
541 something... in between. The memory pointed to by the address does not
542 need to be accessible (it could be a null pointer for example), but C<rw>
543 and C<locality> must be compile-time constants.
544
545 This is a statement, not a function: you cannot use it as part of an
546 expression.
547
548 An obvious way to use this is to prefetch some data far away, in a big
549 array you loop over. This prefetches memory some 128 array elements later,
550 in the hope that it will be ready when the CPU arrives at that location.
551
552 int sum = 0;
553
554 for (i = 0; i < N; ++i)
555 {
556 sum += arr [i]
557 ecb_prefetch (arr + i + 128, 0, 0);
558 }
559
560 It's hard to predict how far to prefetch, and most CPUs that can prefetch
561 are often good enough to predict this kind of behaviour themselves. It
562 gets more interesting with linked lists, especially when you do some fair
563 processing on each list element:
564
565 for (node *n = start; n; n = n->next)
566 {
567 ecb_prefetch (n->next, 0, 0);
568 ... do medium amount of work with *n
569 }
570
571 After processing the node, (part of) the next node might already be in
572 cache.
573
574 =back
575
576 =head2 BIT FIDDLING / BIT WIZARDRY
577
578 =over 4
579
580 =item bool ecb_big_endian ()
581
582 =item bool ecb_little_endian ()
583
584 These two functions return true if the byte order is big endian
585 (most-significant byte first) or little endian (least-significant byte
586 first) respectively.
587
588 On systems that are neither, their return values are unspecified.
589
590 =item int ecb_ctz32 (uint32_t x)
591
592 =item int ecb_ctz64 (uint64_t x)
593
594 Returns the index of the least significant bit set in C<x> (or
595 equivalently the number of bits set to 0 before the least significant bit
596 set), starting from 0. If C<x> is 0 the result is undefined.
597
598 For smaller types than C<uint32_t> you can safely use C<ecb_ctz32>.
599
600 For example:
601
602 ecb_ctz32 (3) = 0
603 ecb_ctz32 (6) = 1
604
605 =item bool ecb_is_pot32 (uint32_t x)
606
607 =item bool ecb_is_pot64 (uint32_t x)
608
609 Returns true iff C<x> is a power of two or C<x == 0>.
610
611 For smaller types than C<uint32_t> you can safely use C<ecb_is_pot32>.
612
613 =item int ecb_ld32 (uint32_t x)
614
615 =item int ecb_ld64 (uint64_t x)
616
617 Returns the index of the most significant bit set in C<x>, or the number
618 of digits the number requires in binary (so that C<< 2**ld <= x <
619 2**(ld+1) >>). If C<x> is 0 the result is undefined. A common use case is
620 to compute the integer binary logarithm, i.e. C<floor (log2 (n))>, for
621 example to see how many bits a certain number requires to be encoded.
622
623 This function is similar to the "count leading zero bits" function, except
624 that that one returns how many zero bits are "in front" of the number (in
625 the given data type), while C<ecb_ld> returns how many bits the number
626 itself requires.
627
628 For smaller types than C<uint32_t> you can safely use C<ecb_ld32>.
629
630 =item int ecb_popcount32 (uint32_t x)
631
632 =item int ecb_popcount64 (uint64_t x)
633
634 Returns the number of bits set to 1 in C<x>.
635
636 For smaller types than C<uint32_t> you can safely use C<ecb_popcount32>.
637
638 For example:
639
640 ecb_popcount32 (7) = 3
641 ecb_popcount32 (255) = 8
642
643 =item uint8_t ecb_bitrev8 (uint8_t x)
644
645 =item uint16_t ecb_bitrev16 (uint16_t x)
646
647 =item uint32_t ecb_bitrev32 (uint32_t x)
648
649 Reverses the bits in x, i.e. the MSB becomes the LSB, MSB-1 becomes LSB+1
650 and so on.
651
652 Example:
653
654 ecb_bitrev8 (0xa7) = 0xea
655 ecb_bitrev32 (0xffcc4411) = 0x882233ff
656
657 =item uint32_t ecb_bswap16 (uint32_t x)
658
659 =item uint32_t ecb_bswap32 (uint32_t x)
660
661 =item uint64_t ecb_bswap64 (uint64_t x)
662
663 These functions return the value of the 16-bit (32-bit, 64-bit) value
664 C<x> after reversing the order of bytes (0x11223344 becomes 0x44332211 in
665 C<ecb_bswap32>).
666
667 =item uint8_t ecb_rotl8 (uint8_t x, unsigned int count)
668
669 =item uint16_t ecb_rotl16 (uint16_t x, unsigned int count)
670
671 =item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
672
673 =item uint64_t ecb_rotl64 (uint64_t x, unsigned int count)
674
675 =item uint8_t ecb_rotr8 (uint8_t x, unsigned int count)
676
677 =item uint16_t ecb_rotr16 (uint16_t x, unsigned int count)
678
679 =item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
680
681 =item uint64_t ecb_rotr64 (uint64_t x, unsigned int count)
682
683 These two families of functions return the value of C<x> after rotating
684 all the bits by C<count> positions to the right (C<ecb_rotr>) or left
685 (C<ecb_rotl>).
686
687 Current GCC versions understand these functions and usually compile them
688 to "optimal" code (e.g. a single C<rol> or a combination of C<shld> on
689 x86).
690
691 =back
692
693 =head2 FLOATING POINT FIDDLING
694
695 =over 4
696
697 =item ECB_INFINITY [-UECB_NO_LIBM]
698
699 Evaluates to positive infinity if supported by the platform, otherwise to
700 a truly huge number.
701
702 =item ECB_NAN [-UECB_NO_LIBM]
703
704 Evaluates to a quiet NAN if supported by the platform, otherwise to
705 C<ECB_INFINITY>.
706
707 =item float ecb_ldexpf (float x, int exp) [-UECB_NO_LIBM]
708
709 Same as C<ldexpf>, but always available.
710
711 =item uint32_t ecb_float_to_binary16 (float x) [-UECB_NO_LIBM]
712
713 =item uint32_t ecb_float_to_binary32 (float x) [-UECB_NO_LIBM]
714
715 =item uint64_t ecb_double_to_binary64 (double x) [-UECB_NO_LIBM]
716
717 These functions each take an argument in the native C<float> or C<double>
718 type and return the IEEE 754 bit representation of it (binary16/half,
719 binary32/single or binary64/double precision).
720
721 The bit representation is just as IEEE 754 defines it, i.e. the sign bit
722 will be the most significant bit, followed by exponent and mantissa.
723
724 This function should work even when the native floating point format isn't
725 IEEE compliant, of course at a speed and code size penalty, and of course
726 also within reasonable limits (it tries to convert NaNs, infinities and
727 denormals, but will likely convert negative zero to positive zero).
728
729 On all modern platforms (where C<ECB_STDFP> is true), the compiler should
730 be able to optimise away this function completely.
731
732 These functions can be helpful when serialising floats to the network - you
733 can serialise the return value like a normal uint16_t/uint32_t/uint64_t.
734
735 Another use for these functions is to manipulate floating point values
736 directly.
737
738 Silly example: toggle the sign bit of a float.
739
740 /* On gcc-4.7 on amd64, */
741 /* this results in a single add instruction to toggle the bit, and 4 extra */
742 /* instructions to move the float value to an integer register and back. */
743
744 x = ecb_binary32_to_float (ecb_float_to_binary32 (x) ^ 0x80000000U)
745
746 =item float ecb_binary16_to_float (uint16_t x) [-UECB_NO_LIBM]
747
748 =item float ecb_binary32_to_float (uint32_t x) [-UECB_NO_LIBM]
749
750 =item double ecb_binary64_to_double (uint64_t x) [-UECB_NO_LIBM]
751
752 The reverse operation of the previous function - takes the bit
753 representation of an IEEE binary16, binary32 or binary64 number (half,
754 single or double precision) and converts it to the native C<float> or
755 C<double> format.
756
757 This function should work even when the native floating point format isn't
758 IEEE compliant, of course at a speed and code size penalty, and of course
759 also within reasonable limits (it tries to convert normals and denormals,
760 and might be lucky for infinities, and with extraordinary luck, also for
761 negative zero).
762
763 On all modern platforms (where C<ECB_STDFP> is true), the compiler should
764 be able to optimise away this function completely.
765
766 =item uint16_t ecb_binary32_to_binary16 (uint32_t x)
767
768 =item uint32_t ecb_binary16_to_binary32 (uint16_t x)
769
770 Convert a IEEE binary32/single precision to binary16/half format, and vice
771 versa, handling all details (round-to-nearest-even, subnormals, infinity
772 and NaNs) correctly.
773
774 These are functions are available under C<-DECB_NO_LIBM>, since
775 they do not rely on the platform floating point format. The
776 C<ecb_float_to_binary16> and C<ecb_binary16_to_float> functions are
777 usually what you want.
778
779 =back
780
781 =head2 ARITHMETIC
782
783 =over 4
784
785 =item x = ecb_mod (m, n)
786
787 Returns C<m> modulo C<n>, which is the same as the positive remainder
788 of the division operation between C<m> and C<n>, using floored
789 division. Unlike the C remainder operator C<%>, this function ensures that
790 the return value is always positive and that the two numbers I<m> and
791 I<m' = m + i * n> result in the same value modulo I<n> - in other words,
792 C<ecb_mod> implements the mathematical modulo operation, which is missing
793 in the language.
794
795 C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
796 negatable, that is, both C<m> and C<-m> must be representable in its
797 type (this typically excludes the minimum signed integer value, the same
798 limitation as for C</> and C<%> in C).
799
800 Current GCC versions compile this into an efficient branchless sequence on
801 almost all CPUs.
802
803 For example, when you want to rotate forward through the members of an
804 array for increasing C<m> (which might be negative), then you should use
805 C<ecb_mod>, as the C<%> operator might give either negative results, or
806 change direction for negative values:
807
808 for (m = -100; m <= 100; ++m)
809 int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
810
811 =item x = ecb_div_rd (val, div)
812
813 =item x = ecb_div_ru (val, div)
814
815 Returns C<val> divided by C<div> rounded down or up, respectively.
816 C<val> and C<div> must have integer types and C<div> must be strictly
817 positive. Note that these functions are implemented with macros in C
818 and with function templates in C++.
819
820 =back
821
822 =head2 UTILITY
823
824 =over 4
825
826 =item element_count = ecb_array_length (name)
827
828 Returns the number of elements in the array C<name>. For example:
829
830 int primes[] = { 2, 3, 5, 7, 11 };
831 int sum = 0;
832
833 for (i = 0; i < ecb_array_length (primes); i++)
834 sum += primes [i];
835
836 =back
837
838 =head2 SYMBOLS GOVERNING COMPILATION OF ECB.H ITSELF
839
840 These symbols need to be defined before including F<ecb.h> the first time.
841
842 =over 4
843
844 =item ECB_NO_THREADS
845
846 If F<ecb.h> is never used from multiple threads, then this symbol can
847 be defined, in which case memory fences (and similar constructs) are
848 completely removed, leading to more efficient code and fewer dependencies.
849
850 Setting this symbol to a true value implies C<ECB_NO_SMP>.
851
852 =item ECB_NO_SMP
853
854 The weaker version of C<ECB_NO_THREADS> - if F<ecb.h> is used from
855 multiple threads, but never concurrently (e.g. if the system the program
856 runs on has only a single CPU with a single core, no hyperthreading and so
857 on), then this symbol can be defined, leading to more efficient code and
858 fewer dependencies.
859
860 =item ECB_NO_LIBM
861
862 When defined to C<1>, do not export any functions that might introduce
863 dependencies on the math library (usually called F<-lm>) - these are
864 marked with [-UECB_NO_LIBM].
865
866 =back
867
868 =head1 UNDOCUMENTED FUNCTIONALITY
869
870 F<ecb.h> is full of undocumented functionality as well, some of which is
871 intended to be internal-use only, some of which we forgot to document, and
872 some of which we hide because we are not sure we will keep the interface
873 stable.
874
875 While you are welcome to rummage around and use whatever you find useful
876 (we can't stop you), keep in mind that we will change undocumented
877 functionality in incompatible ways without thinking twice, while we are
878 considerably more conservative with documented things.
879
880 =head1 AUTHORS
881
882 C<libecb> is designed and maintained by:
883
884 Emanuele Giaquinta <e.giaquinta@glauco.it>
885 Marc Alexander Lehmann <schmorp@schmorp.de>
886
887