ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
Revision: 1.76
Committed: Mon Jan 20 13:13:56 2020 UTC (4 years, 4 months ago) by root
Branch: MAIN
Changes since 1.75: +181 -3 lines
Log Message:
*** empty log message ***

File Contents

# User Rev Content
1 root 1.14 =head1 LIBECB - e-C-Builtins
2 root 1.3
3 root 1.14 =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 sf-exg 1.16 It's part of the e-suite of libraries, other members of which include
9 root 1.14 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 sf-exg 1.16 it provides a number of other lowlevel C utilities, such as endianness
18 root 1.14 detection, byte swapping or bit rotations.
19    
20 root 1.24 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 root 1.17
24 root 1.14 More might come.
25 root 1.3
26     =head2 ABOUT THE HEADER
27    
28 root 1.14 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 sf-exg 1.16 There are currently no object files to link to - future versions might
37 root 1.14 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 root 1.1
59 root 1.40 =head2 TYPES / TYPE SUPPORT
60    
61     ecb.h makes sure that the following types are defined (in the expected way):
62    
63 root 1.76 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 root 1.40
73     The macro C<ECB_PTRSIZE> is defined to the size of a pointer on this
74 root 1.45 platform (currently C<4> or C<8>) and can be used in preprocessor
75     expressions.
76 root 1.40
77 root 1.74 For C<ptrdiff_t> and C<size_t> use C<stddef.h>/C<cstddef>.
78 root 1.49
79 root 1.62 =head2 LANGUAGE/ENVIRONMENT/COMPILER VERSIONS
80 root 1.43
81 sf-exg 1.46 All the following symbols expand to an expression that can be tested in
82 root 1.44 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 root 1.43 =over 4
86    
87 root 1.44 =item ECB_C
88    
89     True if the implementation defines the C<__STDC__> macro to a true value,
90 root 1.55 while not claiming to be C++.
91 root 1.44
92 root 1.43 =item ECB_C99
93    
94 root 1.47 True if the implementation claims to be compliant to C99 (ISO/IEC
95 root 1.55 9899:1999) or any later version, while not claiming to be C++.
96 root 1.47
97     Note that later versions (ECB_C11) remove core features again (for
98     example, variable length arrays).
99 root 1.43
100 root 1.74 =item ECB_C11, ECB_C17
101 root 1.43
102 root 1.74 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 root 1.44
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 root 1.74 =item ECB_CPP11, ECB_CPP14, ECB_CPP17
111 root 1.44
112 root 1.74 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 root 1.43
115 root 1.57 =item ECB_GCC_VERSION (major, minor)
116 root 1.43
117     Expands to a true value (suitable for testing in by the preprocessor)
118 sf-exg 1.46 if the compiler used is GNU C and the version is the given version, or
119 root 1.43 higher.
120    
121     This macro tries to return false on compilers that claim to be GCC
122     compatible but aren't.
123    
124 root 1.50 =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 root 1.54 =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 root 1.43 =back
173    
174 root 1.62 =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 root 1.64 =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 root 1.62 =back
215    
216 sf-exg 1.60 =head2 ATTRIBUTES
217 root 1.1
218 sf-exg 1.60 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 root 1.20
224     ecb_const int mysqrt (int a);
225     ecb_unused int i;
226    
227 root 1.1 =over 4
228    
229 root 1.3 =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 root 1.15 {
236 sf-exg 1.61 ecb_unused int var;
237 root 1.3
238 root 1.15 #ifdef SOMECONDITION
239     var = ...;
240     return var;
241     #else
242     return 0;
243     #endif
244     }
245 root 1.3
246 root 1.56 =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 root 1.62 =item ecb_deprecated_message (message)
252    
253 root 1.67 Same as C<ecb_deprecated>, but if possible, the specified diagnostic is
254 root 1.62 used instead of a generic depreciation message when the object is being
255     used.
256    
257 root 1.31 =item ecb_inline
258 root 1.29
259 root 1.73 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 root 1.29
263     Example: inline this function, it surely will reduce codesize.
264    
265 root 1.31 ecb_inline int
266 root 1.29 negmul (int a, int b)
267     {
268     return - (a * b);
269     }
270    
271 root 1.2 =item ecb_noinline
272    
273 sf-exg 1.66 Prevents a function from being inlined - it might be optimised away, but
274 root 1.3 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 root 1.2 =item ecb_noreturn
278    
279 root 1.17 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 sf-exg 1.19 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 root 1.17
293 root 1.53 =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 sf-exg 1.61 multiply (ecb_restrict float *src,
305     ecb_restrict float *dst,
306 root 1.53 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 root 1.2 =item ecb_const
315    
316 sf-exg 1.19 Declares that the function only depends on the values of its arguments,
317 root 1.17 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 sf-exg 1.19 such as a function returning the square root of its input argument.
328 root 1.17
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 root 1.2 =item ecb_pure
336    
337 root 1.17 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 root 1.2 =item ecb_hot
352    
353 root 1.17 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 sf-exg 1.19 Whether a function is hot or not often depends on the whole program,
361 root 1.17 and less on the function itself. C<ecb_cold> is likely more useful in
362     practise.
363    
364 root 1.2 =item ecb_cold
365    
366 root 1.17 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 root 1.27 C<ecb_expect_false> had been used to reach them.
376 root 1.17
377     Good examples for such functions would be error reporting functions, or
378     functions only called in exceptional or rare cases.
379    
380 root 1.2 =item ecb_artificial
381    
382 root 1.17 Declares the function as "artificial", in this case meaning that this
383 root 1.52 function is not really meant to be a function, but more like an accessor
384 root 1.17 - 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 root 1.2 =back
408 root 1.1
409     =head2 OPTIMISATION HINTS
410    
411     =over 4
412    
413 root 1.75 =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 root 1.58 =item bool ecb_is_constant (expr)
420 root 1.1
421 root 1.3 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 root 1.5 0..n-1, then you could use this inline function in a header file:
427 root 1.3
428     ecb_inline uint32_t
429     rndm (uint32_t n)
430     {
431 root 1.6 return (n * (uint32_t)rndm16 ()) >> 16;
432 root 1.3 }
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 root 1.6 : (n * (uint32_t)rndm16 ()) >> 16;
445 root 1.3 }
446    
447 root 1.62 =item ecb_expect (expr, value)
448 root 1.1
449 root 1.7 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 root 1.1
453 root 1.27 Usually, you want to use the more intuitive C<ecb_expect_true> and
454     C<ecb_expect_false> functions instead.
455 root 1.1
456 root 1.27 =item bool ecb_expect_true (cond)
457 root 1.1
458 root 1.27 =item bool ecb_expect_false (cond)
459 root 1.1
460 root 1.7 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 root 1.27 if (ecb_expect_true (some_condition)) ...;
467 root 1.7
468 root 1.27 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 root 1.7
472 root 1.9 For example, when you check for a null pointer and expect this to be a
473 root 1.27 rare, exceptional, case, then use C<ecb_expect_false>:
474 root 1.7
475     void my_free (void *ptr)
476     {
477 root 1.27 if (ecb_expect_false (ptr == 0))
478 root 1.7 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 root 1.27 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 root 1.7 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 root 1.9 each time something is added, you have to check for a buffer overrun, but
494 root 1.7 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 root 1.27 if (ecb_expect_false (current + size > end))
501 root 1.7 real_reserve_method (size); /* presumably noinline */
502     }
503    
504 root 1.62 =item ecb_assume (cond)
505 root 1.7
506 sf-exg 1.66 Tries to tell the compiler that some condition is true, even if it's not
507 root 1.65 obvious. This is not a function, but a statement: it cannot be used in
508     another expression.
509 root 1.7
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 root 1.27 For example, the example reservation function from the C<ecb_expect_false>
515 root 1.7 description could be written thus (only C<ecb_assume> was added):
516    
517     ecb_inline void
518     reserve (int size)
519     {
520 root 1.27 if (ecb_expect_false (current + size > end))
521 root 1.7 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 root 1.62 =item ecb_unreachable ()
536 root 1.7
537     This function does nothing itself, except tell the compiler that it will
538 root 1.9 never be executed. Apart from suppressing a warning in some cases, this
539 root 1.65 function can be used to implement C<ecb_assume> or similar functionality.
540 root 1.7
541 root 1.62 =item ecb_prefetch (addr, rw, locality)
542 root 1.7
543     Tells the compiler to try to prefetch memory at the given C<addr>ess
544 root 1.10 for either reading (C<rw> = 0) or writing (C<rw> = 1). A C<locality> of
545 root 1.7 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 root 1.65 This is a statement, not a function: you cannot use it as part of an
552     expression.
553    
554 root 1.7 An obvious way to use this is to prefetch some data far away, in a big
555 root 1.9 array you loop over. This prefetches memory some 128 array elements later,
556 root 1.7 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 root 1.1
580 root 1.2 =back
581 root 1.1
582 root 1.36 =head2 BIT FIDDLING / BIT WIZARDRY
583 root 1.1
584 root 1.4 =over 4
585    
586 root 1.3 =item bool ecb_big_endian ()
587    
588     =item bool ecb_little_endian ()
589    
590 sf-exg 1.11 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 root 1.24 On systems that are neither, their return values are unspecified.
595    
596 root 1.3 =item int ecb_ctz32 (uint32_t x)
597    
598 root 1.35 =item int ecb_ctz64 (uint64_t x)
599    
600 sf-exg 1.11 Returns the index of the least significant bit set in C<x> (or
601 root 1.24 equivalently the number of bits set to 0 before the least significant bit
602 root 1.35 set), starting from 0. If C<x> is 0 the result is undefined.
603    
604 root 1.36 For smaller types than C<uint32_t> you can safely use C<ecb_ctz32>.
605    
606 root 1.35 For example:
607 sf-exg 1.11
608 root 1.15 ecb_ctz32 (3) = 0
609     ecb_ctz32 (6) = 1
610 sf-exg 1.11
611 root 1.41 =item bool ecb_is_pot32 (uint32_t x)
612    
613     =item bool ecb_is_pot64 (uint32_t x)
614    
615 sf-exg 1.66 Returns true iff C<x> is a power of two or C<x == 0>.
616 root 1.41
617 sf-exg 1.66 For smaller types than C<uint32_t> you can safely use C<ecb_is_pot32>.
618 root 1.41
619 root 1.35 =item int ecb_ld32 (uint32_t x)
620    
621     =item int ecb_ld64 (uint64_t x)
622    
623     Returns the index of the most significant bit set in C<x>, or the number
624     of digits the number requires in binary (so that C<< 2**ld <= x <
625     2**(ld+1) >>). If C<x> is 0 the result is undefined. A common use case is
626     to compute the integer binary logarithm, i.e. C<floor (log2 (n))>, for
627     example to see how many bits a certain number requires to be encoded.
628    
629     This function is similar to the "count leading zero bits" function, except
630     that that one returns how many zero bits are "in front" of the number (in
631     the given data type), while C<ecb_ld> returns how many bits the number
632     itself requires.
633    
634 root 1.36 For smaller types than C<uint32_t> you can safely use C<ecb_ld32>.
635    
636 root 1.3 =item int ecb_popcount32 (uint32_t x)
637    
638 root 1.35 =item int ecb_popcount64 (uint64_t x)
639    
640 root 1.36 Returns the number of bits set to 1 in C<x>.
641    
642     For smaller types than C<uint32_t> you can safely use C<ecb_popcount32>.
643    
644     For example:
645 sf-exg 1.11
646 root 1.15 ecb_popcount32 (7) = 3
647     ecb_popcount32 (255) = 8
648 sf-exg 1.11
649 root 1.39 =item uint8_t ecb_bitrev8 (uint8_t x)
650    
651     =item uint16_t ecb_bitrev16 (uint16_t x)
652    
653     =item uint32_t ecb_bitrev32 (uint32_t x)
654    
655     Reverses the bits in x, i.e. the MSB becomes the LSB, MSB-1 becomes LSB+1
656     and so on.
657    
658     Example:
659    
660     ecb_bitrev8 (0xa7) = 0xea
661     ecb_bitrev32 (0xffcc4411) = 0x882233ff
662    
663 root 1.8 =item uint32_t ecb_bswap16 (uint32_t x)
664    
665 root 1.3 =item uint32_t ecb_bswap32 (uint32_t x)
666    
667 root 1.34 =item uint64_t ecb_bswap64 (uint64_t x)
668 sf-exg 1.13
669 root 1.34 These functions return the value of the 16-bit (32-bit, 64-bit) value
670     C<x> after reversing the order of bytes (0x11223344 becomes 0x44332211 in
671     C<ecb_bswap32>).
672    
673 root 1.76 =item T ecb_bswap (T x) [C++]
674    
675     For C++, an additional generic bswap function is provided. It supports
676     C<uint8_t>, C<uint16_t>, C<uint32_t> and C<uint64_t>.
677    
678 root 1.34 =item uint8_t ecb_rotl8 (uint8_t x, unsigned int count)
679    
680     =item uint16_t ecb_rotl16 (uint16_t x, unsigned int count)
681 root 1.3
682     =item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
683    
684 root 1.34 =item uint64_t ecb_rotl64 (uint64_t x, unsigned int count)
685    
686     =item uint8_t ecb_rotr8 (uint8_t x, unsigned int count)
687    
688     =item uint16_t ecb_rotr16 (uint16_t x, unsigned int count)
689    
690     =item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
691    
692 root 1.33 =item uint64_t ecb_rotr64 (uint64_t x, unsigned int count)
693    
694 root 1.34 These two families of functions return the value of C<x> after rotating
695     all the bits by C<count> positions to the right (C<ecb_rotr>) or left
696     (C<ecb_rotl>).
697 sf-exg 1.11
698 root 1.20 Current GCC versions understand these functions and usually compile them
699 root 1.34 to "optimal" code (e.g. a single C<rol> or a combination of C<shld> on
700     x86).
701 root 1.20
702 root 1.3 =back
703 root 1.1
704 root 1.76 =head2 HOST ENDIANNESS CONVERSION
705    
706     =over 4
707    
708     =item uint_fast16_t ecb_be_u16_to_host (uint_fast16_t v)
709    
710     =item uint_fast32_t ecb_be_u32_to_host (uint_fast32_t v)
711    
712     =item uint_fast64_t ecb_be_u64_to_host (uint_fast64_t v)
713    
714     =item uint_fast16_t ecb_le_u16_to_host (uint_fast16_t v)
715    
716     =item uint_fast32_t ecb_le_u32_to_host (uint_fast32_t v)
717    
718     =item uint_fast64_t ecb_le_u64_to_host (uint_fast64_t v)
719    
720     Convert an unsigned 16, 32 or 64 bit value from big or little endian to host byte order.
721    
722     The naming convention is C<ecb_>(C<be>|C<le>)C<_u>C<16|32|64>C<_to_host>,
723     where be and le stand for big endian and little endian, respectively.
724    
725     =item uint_fast16_t ecb_host_to_be_u16 (uint_fast16_t v)
726    
727     =item uint_fast32_t ecb_host_to_be_u32 (uint_fast32_t v)
728    
729     =item uint_fast64_t ecb_host_to_be_u64 (uint_fast64_t v)
730    
731     =item uint_fast16_t ecb_host_to_le_u16 (uint_fast16_t v)
732    
733     =item uint_fast32_t ecb_host_to_le_u32 (uint_fast32_t v)
734    
735     =item uint_fast64_t ecb_host_to_le_u64 (uint_fast64_t v)
736    
737     Like above, but converts I<from> host byte order to the specified
738     endianness.
739    
740     =back
741    
742     In C++ the following additional functions are supported:
743    
744     =over 4
745    
746     =item T ecb_be_to_host (T v)
747    
748     =item T ecb_le_to_host (T v)
749    
750     =item T ecb_host_to_be (T v)
751    
752     =item T ecb_host_to_le (T v)
753    
754     These work like their C counterparts, above, but use templates for the
755     type, which make them useful in generic code.
756    
757     C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>
758     (so unlike their C counterparts, there is a version for C<uint8_t>, which
759     again can be useful in generic code).
760    
761     =head2 UNALIGNED LOAD/STORE
762    
763     These function load or store unaligned multi-byte values.
764    
765     =over 4
766    
767     =item uint_fast16_t ecb_peek_u16_u (const void *ptr)
768    
769     =item uint_fast32_t ecb_peek_u32_u (const void *ptr)
770    
771     =item uint_fast64_t ecb_peek_u64_u (const void *ptr)
772    
773     These functions load an unaligned, unsigned 16, 32 or 64 bit value from
774     memory.
775    
776     =item uint_fast16_t ecb_peek_be_u16_u (const void *ptr)
777    
778     =item uint_fast32_t ecb_peek_be_u32_u (const void *ptr)
779    
780     =item uint_fast64_t ecb_peek_be_u64_u (const void *ptr)
781    
782     =item uint_fast16_t ecb_peek_le_u16_u (const void *ptr)
783    
784     =item uint_fast32_t ecb_peek_le_u32_u (const void *ptr)
785    
786     =item uint_fast64_t ecb_peek_le_u64_u (const void *ptr)
787    
788     Like above, but additionally convert from big endian (C<be>) or little
789     endian (C<le>) byte order to host byte order while doing so.
790    
791     =item ecb_poke_u16_u (void *ptr, uint16_t v)
792    
793     =item ecb_poke_u32_u (void *ptr, uint32_t v)
794    
795     =item ecb_poke_u64_u (void *ptr, uint64_t v)
796    
797     These functions store an unaligned, unsigned 16, 32 or 64 bit value to
798     memory.
799    
800     =item ecb_poke_be_u16_u (void *ptr, uint_fast16_t v)
801    
802     =item ecb_poke_be_u32_u (void *ptr, uint_fast32_t v)
803    
804     =item ecb_poke_be_u64_u (void *ptr, uint_fast64_t v)
805    
806     =item ecb_poke_le_u16_u (void *ptr, uint_fast16_t v)
807    
808     =item ecb_poke_le_u32_u (void *ptr, uint_fast32_t v)
809    
810     =item ecb_poke_le_u64_u (void *ptr, uint_fast64_t v)
811    
812     Like above, but additionally convert from host byte order to big endian
813     (C<be>) or little endian (C<le>) byte order while doing so.
814    
815     =back
816    
817     In C++ the following additional functions are supported:
818    
819     =over 4
820    
821     =item T ecb_peek (const void *ptr)
822    
823     =item T ecb_peek_be (const void *ptr)
824    
825     =item T ecb_peek_le (const void *ptr)
826    
827     =item T ecb_peek_u (const void *ptr)
828    
829     =item T ecb_peek_be_u (const void *ptr)
830    
831     =item T ecb_peek_le_u (const void *ptr)
832    
833     Similarly to their C counterparts, these functions load an unsigned 8, 16,
834     32 or 64 bit value from memory, with optional conversion from big/little
835     endian.
836    
837     Since the type cannot be deduced, it has top be specified explicitly, e.g.
838    
839     uint_fast16_t v = ecb_peek<uint16_t> (ptr);
840    
841     C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>.
842    
843     Unlike their C counterparts, these functions support 8 bit quantities
844     (C<uint8_t>) and also have an aligned version (without the C<_u> prefix),
845     all of which hopefully makes them more useful in generic code.
846    
847     =item ecb_poke (void *ptr, T v)
848    
849     =item ecb_poke_be (void *ptr, T v)
850    
851     =item ecb_poke_le (void *ptr, T v)
852    
853     =item ecb_poke_u (void *ptr, T v)
854    
855     =item ecb_poke_be_u (void *ptr, T v)
856    
857     =item ecb_poke_le_u (void *ptr, T v)
858    
859     Again, similarly to their C counterparts, these functions store an
860     unsigned 8, 16, 32 or z64 bit value to memory, with optional conversion to
861     big/little endian.
862    
863     C<T> must be one of C<uint8_t>, C<uint16_t>, C<uint32_t> or C<uint64_t>.
864    
865     Unlike their C counterparts, these functions support 8 bit quantities
866     (C<uint8_t>) and also have an aligned version (without the C<_u> prefix),
867     all of which hopefully makes them more useful in generic code.
868    
869     =back
870    
871 root 1.50 =head2 FLOATING POINT FIDDLING
872    
873     =over 4
874    
875 root 1.71 =item ECB_INFINITY [-UECB_NO_LIBM]
876 root 1.62
877     Evaluates to positive infinity if supported by the platform, otherwise to
878     a truly huge number.
879    
880 root 1.71 =item ECB_NAN [-UECB_NO_LIBM]
881 root 1.62
882     Evaluates to a quiet NAN if supported by the platform, otherwise to
883     C<ECB_INFINITY>.
884    
885 root 1.71 =item float ecb_ldexpf (float x, int exp) [-UECB_NO_LIBM]
886 root 1.62
887     Same as C<ldexpf>, but always available.
888    
889 root 1.71 =item uint32_t ecb_float_to_binary16 (float x) [-UECB_NO_LIBM]
890    
891 root 1.50 =item uint32_t ecb_float_to_binary32 (float x) [-UECB_NO_LIBM]
892    
893     =item uint64_t ecb_double_to_binary64 (double x) [-UECB_NO_LIBM]
894    
895     These functions each take an argument in the native C<float> or C<double>
896 root 1.71 type and return the IEEE 754 bit representation of it (binary16/half,
897     binary32/single or binary64/double precision).
898 root 1.50
899     The bit representation is just as IEEE 754 defines it, i.e. the sign bit
900     will be the most significant bit, followed by exponent and mantissa.
901    
902     This function should work even when the native floating point format isn't
903     IEEE compliant, of course at a speed and code size penalty, and of course
904     also within reasonable limits (it tries to convert NaNs, infinities and
905     denormals, but will likely convert negative zero to positive zero).
906    
907     On all modern platforms (where C<ECB_STDFP> is true), the compiler should
908     be able to optimise away this function completely.
909    
910     These functions can be helpful when serialising floats to the network - you
911 root 1.71 can serialise the return value like a normal uint16_t/uint32_t/uint64_t.
912 root 1.50
913     Another use for these functions is to manipulate floating point values
914     directly.
915    
916     Silly example: toggle the sign bit of a float.
917    
918     /* On gcc-4.7 on amd64, */
919     /* this results in a single add instruction to toggle the bit, and 4 extra */
920     /* instructions to move the float value to an integer register and back. */
921    
922     x = ecb_binary32_to_float (ecb_float_to_binary32 (x) ^ 0x80000000U)
923    
924 root 1.58 =item float ecb_binary16_to_float (uint16_t x) [-UECB_NO_LIBM]
925    
926 root 1.50 =item float ecb_binary32_to_float (uint32_t x) [-UECB_NO_LIBM]
927    
928 root 1.70 =item double ecb_binary64_to_double (uint64_t x) [-UECB_NO_LIBM]
929 root 1.50
930 sf-exg 1.59 The reverse operation of the previous function - takes the bit
931 root 1.71 representation of an IEEE binary16, binary32 or binary64 number (half,
932     single or double precision) and converts it to the native C<float> or
933     C<double> format.
934 root 1.50
935     This function should work even when the native floating point format isn't
936     IEEE compliant, of course at a speed and code size penalty, and of course
937     also within reasonable limits (it tries to convert normals and denormals,
938     and might be lucky for infinities, and with extraordinary luck, also for
939     negative zero).
940    
941     On all modern platforms (where C<ECB_STDFP> is true), the compiler should
942     be able to optimise away this function completely.
943    
944 root 1.71 =item uint16_t ecb_binary32_to_binary16 (uint32_t x)
945    
946     =item uint32_t ecb_binary16_to_binary32 (uint16_t x)
947    
948     Convert a IEEE binary32/single precision to binary16/half format, and vice
949 root 1.72 versa, handling all details (round-to-nearest-even, subnormals, infinity
950     and NaNs) correctly.
951 root 1.71
952     These are functions are available under C<-DECB_NO_LIBM>, since
953     they do not rely on the platform floating point format. The
954     C<ecb_float_to_binary16> and C<ecb_binary16_to_float> functions are
955     usually what you want.
956    
957 root 1.50 =back
958    
959 root 1.1 =head2 ARITHMETIC
960    
961 root 1.3 =over 4
962    
963 root 1.14 =item x = ecb_mod (m, n)
964 root 1.3
965 root 1.25 Returns C<m> modulo C<n>, which is the same as the positive remainder
966     of the division operation between C<m> and C<n>, using floored
967     division. Unlike the C remainder operator C<%>, this function ensures that
968     the return value is always positive and that the two numbers I<m> and
969     I<m' = m + i * n> result in the same value modulo I<n> - in other words,
970     C<ecb_mod> implements the mathematical modulo operation, which is missing
971     in the language.
972 root 1.14
973 sf-exg 1.23 C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
974 root 1.14 negatable, that is, both C<m> and C<-m> must be representable in its
975 root 1.30 type (this typically excludes the minimum signed integer value, the same
976 root 1.25 limitation as for C</> and C<%> in C).
977 sf-exg 1.11
978 root 1.24 Current GCC versions compile this into an efficient branchless sequence on
979 root 1.28 almost all CPUs.
980 root 1.24
981     For example, when you want to rotate forward through the members of an
982     array for increasing C<m> (which might be negative), then you should use
983     C<ecb_mod>, as the C<%> operator might give either negative results, or
984     change direction for negative values:
985    
986     for (m = -100; m <= 100; ++m)
987     int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
988    
989 sf-exg 1.37 =item x = ecb_div_rd (val, div)
990    
991     =item x = ecb_div_ru (val, div)
992    
993     Returns C<val> divided by C<div> rounded down or up, respectively.
994     C<val> and C<div> must have integer types and C<div> must be strictly
995 sf-exg 1.38 positive. Note that these functions are implemented with macros in C
996     and with function templates in C++.
997 sf-exg 1.37
998 root 1.3 =back
999 root 1.1
1000     =head2 UTILITY
1001    
1002 root 1.3 =over 4
1003    
1004 sf-exg 1.23 =item element_count = ecb_array_length (name)
1005 root 1.3
1006 sf-exg 1.13 Returns the number of elements in the array C<name>. For example:
1007    
1008     int primes[] = { 2, 3, 5, 7, 11 };
1009     int sum = 0;
1010    
1011     for (i = 0; i < ecb_array_length (primes); i++)
1012     sum += primes [i];
1013    
1014 root 1.3 =back
1015 root 1.1
1016 root 1.43 =head2 SYMBOLS GOVERNING COMPILATION OF ECB.H ITSELF
1017    
1018     These symbols need to be defined before including F<ecb.h> the first time.
1019    
1020     =over 4
1021    
1022 root 1.51 =item ECB_NO_THREADS
1023 root 1.43
1024     If F<ecb.h> is never used from multiple threads, then this symbol can
1025     be defined, in which case memory fences (and similar constructs) are
1026     completely removed, leading to more efficient code and fewer dependencies.
1027    
1028     Setting this symbol to a true value implies C<ECB_NO_SMP>.
1029    
1030     =item ECB_NO_SMP
1031    
1032     The weaker version of C<ECB_NO_THREADS> - if F<ecb.h> is used from
1033     multiple threads, but never concurrently (e.g. if the system the program
1034     runs on has only a single CPU with a single core, no hyperthreading and so
1035     on), then this symbol can be defined, leading to more efficient code and
1036     fewer dependencies.
1037    
1038 root 1.50 =item ECB_NO_LIBM
1039    
1040     When defined to C<1>, do not export any functions that might introduce
1041     dependencies on the math library (usually called F<-lm>) - these are
1042     marked with [-UECB_NO_LIBM].
1043    
1044 sf-exg 1.69 =back
1045    
1046 root 1.68 =head1 UNDOCUMENTED FUNCTIONALITY
1047    
1048     F<ecb.h> is full of undocumented functionality as well, some of which is
1049     intended to be internal-use only, some of which we forgot to document, and
1050     some of which we hide because we are not sure we will keep the interface
1051     stable.
1052    
1053     While you are welcome to rummage around and use whatever you find useful
1054     (we can't stop you), keep in mind that we will change undocumented
1055     functionality in incompatible ways without thinking twice, while we are
1056     considerably more conservative with documented things.
1057    
1058     =head1 AUTHORS
1059    
1060     C<libecb> is designed and maintained by:
1061    
1062     Emanuele Giaquinta <e.giaquinta@glauco.it>
1063     Marc Alexander Lehmann <schmorp@schmorp.de>
1064    
1065 root 1.1