ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
Revision: 1.48
Committed: Wed May 30 16:56:06 2012 UTC (12 years ago) by root
Branch: MAIN
Changes since 1.47: +0 -6 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 ptrdiff_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 =head2 LANGUAGE/COMPILER VERSIONS
72
73 All the following symbols expand to an expression that can be tested in
74 preprocessor instructions as well as treated as a boolean (use C<!!> to
75 ensure it's either C<0> or C<1> if you need that).
76
77 =over 4
78
79 =item ECB_C
80
81 True if the implementation defines the C<__STDC__> macro to a true value,
82 which is typically true for both C and C++ compilers.
83
84 =item ECB_C99
85
86 True if the implementation claims to be compliant to C99 (ISO/IEC
87 9899:1999) or any later version.
88
89 Note that later versions (ECB_C11) remove core features again (for
90 example, variable length arrays).
91
92 =item ECB_C11
93
94 True if the implementation claims to be compliant to C11 (ISO/IEC
95 9899:2011) or any later version.
96
97 =item ECB_CPP
98
99 True if the implementation defines the C<__cplusplus__> macro to a true
100 value, which is typically true for C++ compilers.
101
102 =item ECB_CPP11
103
104 True if the implementation claims to be compliant to ISO/IEC 14882:2011
105 (C++11) or any later version.
106
107 =item ECB_GCC_VERSION(major,minor)
108
109 Expands to a true value (suitable for testing in by the preprocessor)
110 if the compiler used is GNU C and the version is the given version, or
111 higher.
112
113 This macro tries to return false on compilers that claim to be GCC
114 compatible but aren't.
115
116 =back
117
118 =head2 GCC ATTRIBUTES
119
120 A major part of libecb deals with GCC attributes. These are additional
121 attributes that you can assign to functions, variables and sometimes even
122 types - much like C<const> or C<volatile> in C.
123
124 While GCC allows declarations to show up in many surprising places,
125 but not in many expected places, the safest way is to put attribute
126 declarations before the whole declaration:
127
128 ecb_const int mysqrt (int a);
129 ecb_unused int i;
130
131 For variables, it is often nicer to put the attribute after the name, and
132 avoid multiple declarations using commas:
133
134 int i ecb_unused;
135
136 =over 4
137
138 =item ecb_attribute ((attrs...))
139
140 A simple wrapper that expands to C<__attribute__((attrs))> on GCC, and to
141 nothing on other compilers, so the effect is that only GCC sees these.
142
143 Example: use the C<deprecated> attribute on a function.
144
145 ecb_attribute((__deprecated__)) void
146 do_not_use_me_anymore (void);
147
148 =item ecb_unused
149
150 Marks a function or a variable as "unused", which simply suppresses a
151 warning by GCC when it detects it as unused. This is useful when you e.g.
152 declare a variable but do not always use it:
153
154 {
155 int var ecb_unused;
156
157 #ifdef SOMECONDITION
158 var = ...;
159 return var;
160 #else
161 return 0;
162 #endif
163 }
164
165 =item ecb_inline
166
167 This is not actually an attribute, but you use it like one. It expands
168 either to C<static inline> or to just C<static>, if inline isn't
169 supported. It should be used to declare functions that should be inlined,
170 for code size or speed reasons.
171
172 Example: inline this function, it surely will reduce codesize.
173
174 ecb_inline int
175 negmul (int a, int b)
176 {
177 return - (a * b);
178 }
179
180 =item ecb_noinline
181
182 Prevent a function from being inlined - it might be optimised away, but
183 not inlined into other functions. This is useful if you know your function
184 is rarely called and large enough for inlining not to be helpful.
185
186 =item ecb_noreturn
187
188 Marks a function as "not returning, ever". Some typical functions that
189 don't return are C<exit> or C<abort> (which really works hard to not
190 return), and now you can make your own:
191
192 ecb_noreturn void
193 my_abort (const char *errline)
194 {
195 puts (errline);
196 abort ();
197 }
198
199 In this case, the compiler would probably be smart enough to deduce it on
200 its own, so this is mainly useful for declarations.
201
202 =item ecb_const
203
204 Declares that the function only depends on the values of its arguments,
205 much like a mathematical function. It specifically does not read or write
206 any memory any arguments might point to, global variables, or call any
207 non-const functions. It also must not have any side effects.
208
209 Such a function can be optimised much more aggressively by the compiler -
210 for example, multiple calls with the same arguments can be optimised into
211 a single call, which wouldn't be possible if the compiler would have to
212 expect any side effects.
213
214 It is best suited for functions in the sense of mathematical functions,
215 such as a function returning the square root of its input argument.
216
217 Not suited would be a function that calculates the hash of some memory
218 area you pass in, prints some messages or looks at a global variable to
219 decide on rounding.
220
221 See C<ecb_pure> for a slightly less restrictive class of functions.
222
223 =item ecb_pure
224
225 Similar to C<ecb_const>, declares a function that has no side
226 effects. Unlike C<ecb_const>, the function is allowed to examine global
227 variables and any other memory areas (such as the ones passed to it via
228 pointers).
229
230 While these functions cannot be optimised as aggressively as C<ecb_const>
231 functions, they can still be optimised away in many occasions, and the
232 compiler has more freedom in moving calls to them around.
233
234 Typical examples for such functions would be C<strlen> or C<memcmp>. A
235 function that calculates the MD5 sum of some input and updates some MD5
236 state passed as argument would I<NOT> be pure, however, as it would modify
237 some memory area that is not the return value.
238
239 =item ecb_hot
240
241 This declares a function as "hot" with regards to the cache - the function
242 is used so often, that it is very beneficial to keep it in the cache if
243 possible.
244
245 The compiler reacts by trying to place hot functions near to each other in
246 memory.
247
248 Whether a function is hot or not often depends on the whole program,
249 and less on the function itself. C<ecb_cold> is likely more useful in
250 practise.
251
252 =item ecb_cold
253
254 The opposite of C<ecb_hot> - declares a function as "cold" with regards to
255 the cache, or in other words, this function is not called often, or not at
256 speed-critical times, and keeping it in the cache might be a waste of said
257 cache.
258
259 In addition to placing cold functions together (or at least away from hot
260 functions), this knowledge can be used in other ways, for example, the
261 function will be optimised for size, as opposed to speed, and codepaths
262 leading to calls to those functions can automatically be marked as if
263 C<ecb_expect_false> had been used to reach them.
264
265 Good examples for such functions would be error reporting functions, or
266 functions only called in exceptional or rare cases.
267
268 =item ecb_artificial
269
270 Declares the function as "artificial", in this case meaning that this
271 function is not really mean to be a function, but more like an accessor
272 - many methods in C++ classes are mere accessor functions, and having a
273 crash reported in such a method, or single-stepping through them, is not
274 usually so helpful, especially when it's inlined to just a few instructions.
275
276 Marking them as artificial will instruct the debugger about just this,
277 leading to happier debugging and thus happier lives.
278
279 Example: in some kind of smart-pointer class, mark the pointer accessor as
280 artificial, so that the whole class acts more like a pointer and less like
281 some C++ abstraction monster.
282
283 template<typename T>
284 struct my_smart_ptr
285 {
286 T *value;
287
288 ecb_artificial
289 operator T *()
290 {
291 return value;
292 }
293 };
294
295 =back
296
297 =head2 OPTIMISATION HINTS
298
299 =over 4
300
301 =item bool ecb_is_constant(expr)
302
303 Returns true iff the expression can be deduced to be a compile-time
304 constant, and false otherwise.
305
306 For example, when you have a C<rndm16> function that returns a 16 bit
307 random number, and you have a function that maps this to a range from
308 0..n-1, then you could use this inline function in a header file:
309
310 ecb_inline uint32_t
311 rndm (uint32_t n)
312 {
313 return (n * (uint32_t)rndm16 ()) >> 16;
314 }
315
316 However, for powers of two, you could use a normal mask, but that is only
317 worth it if, at compile time, you can detect this case. This is the case
318 when the passed number is a constant and also a power of two (C<n & (n -
319 1) == 0>):
320
321 ecb_inline uint32_t
322 rndm (uint32_t n)
323 {
324 return is_constant (n) && !(n & (n - 1))
325 ? rndm16 () & (num - 1)
326 : (n * (uint32_t)rndm16 ()) >> 16;
327 }
328
329 =item bool ecb_expect (expr, value)
330
331 Evaluates C<expr> and returns it. In addition, it tells the compiler that
332 the C<expr> evaluates to C<value> a lot, which can be used for static
333 branch optimisations.
334
335 Usually, you want to use the more intuitive C<ecb_expect_true> and
336 C<ecb_expect_false> functions instead.
337
338 =item bool ecb_expect_true (cond)
339
340 =item bool ecb_expect_false (cond)
341
342 These two functions expect a expression that is true or false and return
343 C<1> or C<0>, respectively, so when used in the condition of an C<if> or
344 other conditional statement, it will not change the program:
345
346 /* these two do the same thing */
347 if (some_condition) ...;
348 if (ecb_expect_true (some_condition)) ...;
349
350 However, by using C<ecb_expect_true>, you tell the compiler that the
351 condition is likely to be true (and for C<ecb_expect_false>, that it is
352 unlikely to be true).
353
354 For example, when you check for a null pointer and expect this to be a
355 rare, exceptional, case, then use C<ecb_expect_false>:
356
357 void my_free (void *ptr)
358 {
359 if (ecb_expect_false (ptr == 0))
360 return;
361 }
362
363 Consequent use of these functions to mark away exceptional cases or to
364 tell the compiler what the hot path through a function is can increase
365 performance considerably.
366
367 You might know these functions under the name C<likely> and C<unlikely>
368 - while these are common aliases, we find that the expect name is easier
369 to understand when quickly skimming code. If you wish, you can use
370 C<ecb_likely> instead of C<ecb_expect_true> and C<ecb_unlikely> instead of
371 C<ecb_expect_false> - these are simply aliases.
372
373 A very good example is in a function that reserves more space for some
374 memory block (for example, inside an implementation of a string stream) -
375 each time something is added, you have to check for a buffer overrun, but
376 you expect that most checks will turn out to be false:
377
378 /* make sure we have "size" extra room in our buffer */
379 ecb_inline void
380 reserve (int size)
381 {
382 if (ecb_expect_false (current + size > end))
383 real_reserve_method (size); /* presumably noinline */
384 }
385
386 =item bool ecb_assume (cond)
387
388 Try to tell the compiler that some condition is true, even if it's not
389 obvious.
390
391 This can be used to teach the compiler about invariants or other
392 conditions that might improve code generation, but which are impossible to
393 deduce form the code itself.
394
395 For example, the example reservation function from the C<ecb_expect_false>
396 description could be written thus (only C<ecb_assume> was added):
397
398 ecb_inline void
399 reserve (int size)
400 {
401 if (ecb_expect_false (current + size > end))
402 real_reserve_method (size); /* presumably noinline */
403
404 ecb_assume (current + size <= end);
405 }
406
407 If you then call this function twice, like this:
408
409 reserve (10);
410 reserve (1);
411
412 Then the compiler I<might> be able to optimise out the second call
413 completely, as it knows that C<< current + 1 > end >> is false and the
414 call will never be executed.
415
416 =item bool ecb_unreachable ()
417
418 This function does nothing itself, except tell the compiler that it will
419 never be executed. Apart from suppressing a warning in some cases, this
420 function can be used to implement C<ecb_assume> or similar functions.
421
422 =item bool ecb_prefetch (addr, rw, locality)
423
424 Tells the compiler to try to prefetch memory at the given C<addr>ess
425 for either reading (C<rw> = 0) or writing (C<rw> = 1). A C<locality> of
426 C<0> means that there will only be one access later, C<3> means that
427 the data will likely be accessed very often, and values in between mean
428 something... in between. The memory pointed to by the address does not
429 need to be accessible (it could be a null pointer for example), but C<rw>
430 and C<locality> must be compile-time constants.
431
432 An obvious way to use this is to prefetch some data far away, in a big
433 array you loop over. This prefetches memory some 128 array elements later,
434 in the hope that it will be ready when the CPU arrives at that location.
435
436 int sum = 0;
437
438 for (i = 0; i < N; ++i)
439 {
440 sum += arr [i]
441 ecb_prefetch (arr + i + 128, 0, 0);
442 }
443
444 It's hard to predict how far to prefetch, and most CPUs that can prefetch
445 are often good enough to predict this kind of behaviour themselves. It
446 gets more interesting with linked lists, especially when you do some fair
447 processing on each list element:
448
449 for (node *n = start; n; n = n->next)
450 {
451 ecb_prefetch (n->next, 0, 0);
452 ... do medium amount of work with *n
453 }
454
455 After processing the node, (part of) the next node might already be in
456 cache.
457
458 =back
459
460 =head2 BIT FIDDLING / BIT WIZARDRY
461
462 =over 4
463
464 =item bool ecb_big_endian ()
465
466 =item bool ecb_little_endian ()
467
468 These two functions return true if the byte order is big endian
469 (most-significant byte first) or little endian (least-significant byte
470 first) respectively.
471
472 On systems that are neither, their return values are unspecified.
473
474 =item int ecb_ctz32 (uint32_t x)
475
476 =item int ecb_ctz64 (uint64_t x)
477
478 Returns the index of the least significant bit set in C<x> (or
479 equivalently the number of bits set to 0 before the least significant bit
480 set), starting from 0. If C<x> is 0 the result is undefined.
481
482 For smaller types than C<uint32_t> you can safely use C<ecb_ctz32>.
483
484 For example:
485
486 ecb_ctz32 (3) = 0
487 ecb_ctz32 (6) = 1
488
489 =item bool ecb_is_pot32 (uint32_t x)
490
491 =item bool ecb_is_pot64 (uint32_t x)
492
493 Return true iff C<x> is a power of two or C<x == 0>.
494
495 For smaller types then C<uint32_t> you can safely use C<ecb_is_pot32>.
496
497 =item int ecb_ld32 (uint32_t x)
498
499 =item int ecb_ld64 (uint64_t x)
500
501 Returns the index of the most significant bit set in C<x>, or the number
502 of digits the number requires in binary (so that C<< 2**ld <= x <
503 2**(ld+1) >>). If C<x> is 0 the result is undefined. A common use case is
504 to compute the integer binary logarithm, i.e. C<floor (log2 (n))>, for
505 example to see how many bits a certain number requires to be encoded.
506
507 This function is similar to the "count leading zero bits" function, except
508 that that one returns how many zero bits are "in front" of the number (in
509 the given data type), while C<ecb_ld> returns how many bits the number
510 itself requires.
511
512 For smaller types than C<uint32_t> you can safely use C<ecb_ld32>.
513
514 =item int ecb_popcount32 (uint32_t x)
515
516 =item int ecb_popcount64 (uint64_t x)
517
518 Returns the number of bits set to 1 in C<x>.
519
520 For smaller types than C<uint32_t> you can safely use C<ecb_popcount32>.
521
522 For example:
523
524 ecb_popcount32 (7) = 3
525 ecb_popcount32 (255) = 8
526
527 =item uint8_t ecb_bitrev8 (uint8_t x)
528
529 =item uint16_t ecb_bitrev16 (uint16_t x)
530
531 =item uint32_t ecb_bitrev32 (uint32_t x)
532
533 Reverses the bits in x, i.e. the MSB becomes the LSB, MSB-1 becomes LSB+1
534 and so on.
535
536 Example:
537
538 ecb_bitrev8 (0xa7) = 0xea
539 ecb_bitrev32 (0xffcc4411) = 0x882233ff
540
541 =item uint32_t ecb_bswap16 (uint32_t x)
542
543 =item uint32_t ecb_bswap32 (uint32_t x)
544
545 =item uint64_t ecb_bswap64 (uint64_t x)
546
547 These functions return the value of the 16-bit (32-bit, 64-bit) value
548 C<x> after reversing the order of bytes (0x11223344 becomes 0x44332211 in
549 C<ecb_bswap32>).
550
551 =item uint8_t ecb_rotl8 (uint8_t x, unsigned int count)
552
553 =item uint16_t ecb_rotl16 (uint16_t x, unsigned int count)
554
555 =item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
556
557 =item uint64_t ecb_rotl64 (uint64_t x, unsigned int count)
558
559 =item uint8_t ecb_rotr8 (uint8_t x, unsigned int count)
560
561 =item uint16_t ecb_rotr16 (uint16_t x, unsigned int count)
562
563 =item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
564
565 =item uint64_t ecb_rotr64 (uint64_t x, unsigned int count)
566
567 These two families of functions return the value of C<x> after rotating
568 all the bits by C<count> positions to the right (C<ecb_rotr>) or left
569 (C<ecb_rotl>).
570
571 Current GCC versions understand these functions and usually compile them
572 to "optimal" code (e.g. a single C<rol> or a combination of C<shld> on
573 x86).
574
575 =back
576
577 =head2 ARITHMETIC
578
579 =over 4
580
581 =item x = ecb_mod (m, n)
582
583 Returns C<m> modulo C<n>, which is the same as the positive remainder
584 of the division operation between C<m> and C<n>, using floored
585 division. Unlike the C remainder operator C<%>, this function ensures that
586 the return value is always positive and that the two numbers I<m> and
587 I<m' = m + i * n> result in the same value modulo I<n> - in other words,
588 C<ecb_mod> implements the mathematical modulo operation, which is missing
589 in the language.
590
591 C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
592 negatable, that is, both C<m> and C<-m> must be representable in its
593 type (this typically excludes the minimum signed integer value, the same
594 limitation as for C</> and C<%> in C).
595
596 Current GCC versions compile this into an efficient branchless sequence on
597 almost all CPUs.
598
599 For example, when you want to rotate forward through the members of an
600 array for increasing C<m> (which might be negative), then you should use
601 C<ecb_mod>, as the C<%> operator might give either negative results, or
602 change direction for negative values:
603
604 for (m = -100; m <= 100; ++m)
605 int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
606
607 =item x = ecb_div_rd (val, div)
608
609 =item x = ecb_div_ru (val, div)
610
611 Returns C<val> divided by C<div> rounded down or up, respectively.
612 C<val> and C<div> must have integer types and C<div> must be strictly
613 positive. Note that these functions are implemented with macros in C
614 and with function templates in C++.
615
616 =back
617
618 =head2 UTILITY
619
620 =over 4
621
622 =item element_count = ecb_array_length (name)
623
624 Returns the number of elements in the array C<name>. For example:
625
626 int primes[] = { 2, 3, 5, 7, 11 };
627 int sum = 0;
628
629 for (i = 0; i < ecb_array_length (primes); i++)
630 sum += primes [i];
631
632 =back
633
634 =head2 SYMBOLS GOVERNING COMPILATION OF ECB.H ITSELF
635
636 These symbols need to be defined before including F<ecb.h> the first time.
637
638 =over 4
639
640 =item ECB_NO_THRADS
641
642 If F<ecb.h> is never used from multiple threads, then this symbol can
643 be defined, in which case memory fences (and similar constructs) are
644 completely removed, leading to more efficient code and fewer dependencies.
645
646 Setting this symbol to a true value implies C<ECB_NO_SMP>.
647
648 =item ECB_NO_SMP
649
650 The weaker version of C<ECB_NO_THREADS> - if F<ecb.h> is used from
651 multiple threads, but never concurrently (e.g. if the system the program
652 runs on has only a single CPU with a single core, no hyperthreading and so
653 on), then this symbol can be defined, leading to more efficient code and
654 fewer dependencies.
655
656 =back
657
658