ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
(Generate patch)

Comparing libecb/ecb.pod (file contents):
Revision 1.23 by sf-exg, Fri May 27 01:35:46 2011 UTC vs.
Revision 1.35 by root, Fri Jun 17 21:38:31 2011 UTC

… …
15It mainly provides a number of wrappers around GCC built-ins, together 15It mainly provides a number of wrappers around GCC built-ins, together
16with replacement functions for other compilers. In addition to this, 16with replacement functions for other compilers. In addition to this,
17it provides a number of other lowlevel C utilities, such as endianness 17it provides a number of other lowlevel C utilities, such as endianness
18detection, byte swapping or bit rotations. 18detection, byte swapping or bit rotations.
19 19
20Or in other words, things that should be built-in into any standard C 20Or in other words, things that should be built into any standard C system,
21system, but aren't. 21but aren't, implemented as efficient as possible with GCC, and still
22correct with other compilers.
22 23
23More might come. 24More might come.
24 25
25=head2 ABOUT THE HEADER 26=head2 ABOUT THE HEADER
26 27
… …
56refers to any kind of boolean value, not a specific type. 57refers to any kind of boolean value, not a specific type.
57 58
58=head2 GCC ATTRIBUTES 59=head2 GCC ATTRIBUTES
59 60
60A major part of libecb deals with GCC attributes. These are additional 61A major part of libecb deals with GCC attributes. These are additional
61attributes that you cna assign to functions, variables and sometimes even 62attributes that you can assign to functions, variables and sometimes even
62types - much like C<const> or C<volatile> in C. 63types - much like C<const> or C<volatile> in C.
63 64
64While GCC allows declarations to show up in many surprising places, 65While GCC allows declarations to show up in many surprising places,
65but not in many expeted places, the safest way is to put attribute 66but not in many expected places, the safest way is to put attribute
66declarations before the whole declaration: 67declarations before the whole declaration:
67 68
68 ecb_const int mysqrt (int a); 69 ecb_const int mysqrt (int a);
69 ecb_unused int i; 70 ecb_unused int i;
70 71
… …
100 #else 101 #else
101 return 0; 102 return 0;
102 #endif 103 #endif
103 } 104 }
104 105
106=item ecb_inline
107
108This is not actually an attribute, but you use it like one. It expands
109either to C<static inline> or to just C<static>, if inline isn't
110supported. It should be used to declare functions that should be inlined,
111for code size or speed reasons.
112
113Example: inline this function, it surely will reduce codesize.
114
115 ecb_inline int
116 negmul (int a, int b)
117 {
118 return - (a * b);
119 }
120
105=item ecb_noinline 121=item ecb_noinline
106 122
107Prevent a function from being inlined - it might be optimised away, but 123Prevent a function from being inlined - it might be optimised away, but
108not inlined into other functions. This is useful if you know your function 124not inlined into other functions. This is useful if you know your function
109is rarely called and large enough for inlining not to be helpful. 125is rarely called and large enough for inlining not to be helpful.
… …
183 199
184In addition to placing cold functions together (or at least away from hot 200In addition to placing cold functions together (or at least away from hot
185functions), this knowledge can be used in other ways, for example, the 201functions), this knowledge can be used in other ways, for example, the
186function will be optimised for size, as opposed to speed, and codepaths 202function will be optimised for size, as opposed to speed, and codepaths
187leading to calls to those functions can automatically be marked as if 203leading to calls to those functions can automatically be marked as if
188C<ecb_unlikely> had been used to reach them. 204C<ecb_expect_false> had been used to reach them.
189 205
190Good examples for such functions would be error reporting functions, or 206Good examples for such functions would be error reporting functions, or
191functions only called in exceptional or rare cases. 207functions only called in exceptional or rare cases.
192 208
193=item ecb_artificial 209=item ecb_artificial
… …
255 271
256Evaluates C<expr> and returns it. In addition, it tells the compiler that 272Evaluates C<expr> and returns it. In addition, it tells the compiler that
257the C<expr> evaluates to C<value> a lot, which can be used for static 273the C<expr> evaluates to C<value> a lot, which can be used for static
258branch optimisations. 274branch optimisations.
259 275
260Usually, you want to use the more intuitive C<ecb_likely> and 276Usually, you want to use the more intuitive C<ecb_expect_true> and
261C<ecb_unlikely> functions instead. 277C<ecb_expect_false> functions instead.
262 278
279=item bool ecb_expect_true (cond)
280
263=item bool ecb_likely (cond) 281=item bool ecb_expect_false (cond)
264
265=item bool ecb_unlikely (cond)
266 282
267These two functions expect a expression that is true or false and return 283These two functions expect a expression that is true or false and return
268C<1> or C<0>, respectively, so when used in the condition of an C<if> or 284C<1> or C<0>, respectively, so when used in the condition of an C<if> or
269other conditional statement, it will not change the program: 285other conditional statement, it will not change the program:
270 286
271 /* these two do the same thing */ 287 /* these two do the same thing */
272 if (some_condition) ...; 288 if (some_condition) ...;
273 if (ecb_likely (some_condition)) ...; 289 if (ecb_expect_true (some_condition)) ...;
274 290
275However, by using C<ecb_likely>, you tell the compiler that the condition 291However, by using C<ecb_expect_true>, you tell the compiler that the
276is likely to be true (and for C<ecb_unlikely>, that it is unlikely to be 292condition is likely to be true (and for C<ecb_expect_false>, that it is
277true). 293unlikely to be true).
278 294
279For example, when you check for a null pointer and expect this to be a 295For example, when you check for a null pointer and expect this to be a
280rare, exceptional, case, then use C<ecb_unlikely>: 296rare, exceptional, case, then use C<ecb_expect_false>:
281 297
282 void my_free (void *ptr) 298 void my_free (void *ptr)
283 { 299 {
284 if (ecb_unlikely (ptr == 0)) 300 if (ecb_expect_false (ptr == 0))
285 return; 301 return;
286 } 302 }
287 303
288Consequent use of these functions to mark away exceptional cases or to 304Consequent use of these functions to mark away exceptional cases or to
289tell the compiler what the hot path through a function is can increase 305tell the compiler what the hot path through a function is can increase
290performance considerably. 306performance considerably.
307
308You might know these functions under the name C<likely> and C<unlikely>
309- while these are common aliases, we find that the expect name is easier
310to understand when quickly skimming code. If you wish, you can use
311C<ecb_likely> instead of C<ecb_expect_true> and C<ecb_unlikely> instead of
312C<ecb_expect_false> - these are simply aliases.
291 313
292A very good example is in a function that reserves more space for some 314A very good example is in a function that reserves more space for some
293memory block (for example, inside an implementation of a string stream) - 315memory block (for example, inside an implementation of a string stream) -
294each time something is added, you have to check for a buffer overrun, but 316each time something is added, you have to check for a buffer overrun, but
295you expect that most checks will turn out to be false: 317you expect that most checks will turn out to be false:
296 318
297 /* make sure we have "size" extra room in our buffer */ 319 /* make sure we have "size" extra room in our buffer */
298 ecb_inline void 320 ecb_inline void
299 reserve (int size) 321 reserve (int size)
300 { 322 {
301 if (ecb_unlikely (current + size > end)) 323 if (ecb_expect_false (current + size > end))
302 real_reserve_method (size); /* presumably noinline */ 324 real_reserve_method (size); /* presumably noinline */
303 } 325 }
304 326
305=item bool ecb_assume (cond) 327=item bool ecb_assume (cond)
306 328
… …
309 331
310This can be used to teach the compiler about invariants or other 332This can be used to teach the compiler about invariants or other
311conditions that might improve code generation, but which are impossible to 333conditions that might improve code generation, but which are impossible to
312deduce form the code itself. 334deduce form the code itself.
313 335
314For example, the example reservation function from the C<ecb_unlikely> 336For example, the example reservation function from the C<ecb_expect_false>
315description could be written thus (only C<ecb_assume> was added): 337description could be written thus (only C<ecb_assume> was added):
316 338
317 ecb_inline void 339 ecb_inline void
318 reserve (int size) 340 reserve (int size)
319 { 341 {
320 if (ecb_unlikely (current + size > end)) 342 if (ecb_expect_false (current + size > end))
321 real_reserve_method (size); /* presumably noinline */ 343 real_reserve_method (size); /* presumably noinline */
322 344
323 ecb_assume (current + size <= end); 345 ecb_assume (current + size <= end);
324 } 346 }
325 347
… …
386 408
387These two functions return true if the byte order is big endian 409These two functions return true if the byte order is big endian
388(most-significant byte first) or little endian (least-significant byte 410(most-significant byte first) or little endian (least-significant byte
389first) respectively. 411first) respectively.
390 412
413On systems that are neither, their return values are unspecified.
414
391=item int ecb_ctz32 (uint32_t x) 415=item int ecb_ctz32 (uint32_t x)
392 416
417=item int ecb_ctz64 (uint64_t x)
418
393Returns the index of the least significant bit set in C<x> (or 419Returns the index of the least significant bit set in C<x> (or
394equivalently the number of bits set to 0 before the least significant 420equivalently the number of bits set to 0 before the least significant bit
395bit set), starting from 0. If C<x> is 0 the result is undefined. A 421set), starting from 0. If C<x> is 0 the result is undefined.
396common use case is to compute the integer binary logarithm, i.e., 422
397floor(log2(n)). For example: 423For example:
398 424
399 ecb_ctz32 (3) = 0 425 ecb_ctz32 (3) = 0
400 ecb_ctz32 (6) = 1 426 ecb_ctz32 (6) = 1
401 427
428=item int ecb_ld32 (uint32_t x)
429
430=item int ecb_ld64 (uint64_t x)
431
432Returns the index of the most significant bit set in C<x>, or the number
433of digits the number requires in binary (so that C<< 2**ld <= x <
4342**(ld+1) >>). If C<x> is 0 the result is undefined. A common use case is
435to compute the integer binary logarithm, i.e. C<floor (log2 (n))>, for
436example to see how many bits a certain number requires to be encoded.
437
438This function is similar to the "count leading zero bits" function, except
439that that one returns how many zero bits are "in front" of the number (in
440the given data type), while C<ecb_ld> returns how many bits the number
441itself requires.
442
402=item int ecb_popcount32 (uint32_t x) 443=item int ecb_popcount32 (uint32_t x)
444
445=item int ecb_popcount64 (uint64_t x)
403 446
404Returns the number of bits set to 1 in C<x>. For example: 447Returns the number of bits set to 1 in C<x>. For example:
405 448
406 ecb_popcount32 (7) = 3 449 ecb_popcount32 (7) = 3
407 ecb_popcount32 (255) = 8 450 ecb_popcount32 (255) = 8
408 451
409=item uint32_t ecb_bswap16 (uint32_t x) 452=item uint32_t ecb_bswap16 (uint32_t x)
410 453
411=item uint32_t ecb_bswap32 (uint32_t x) 454=item uint32_t ecb_bswap32 (uint32_t x)
412 455
456=item uint64_t ecb_bswap64 (uint64_t x)
457
413These two functions return the value of the 16-bit (32-bit) value C<x> 458These functions return the value of the 16-bit (32-bit, 64-bit) value
414after reversing the order of bytes (0x11223344 becomes 0x44332211). 459C<x> after reversing the order of bytes (0x11223344 becomes 0x44332211 in
460C<ecb_bswap32>).
461
462=item uint8_t ecb_rotl8 (uint8_t x, unsigned int count)
463
464=item uint16_t ecb_rotl16 (uint16_t x, unsigned int count)
465
466=item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
467
468=item uint64_t ecb_rotl64 (uint64_t x, unsigned int count)
469
470=item uint8_t ecb_rotr8 (uint8_t x, unsigned int count)
471
472=item uint16_t ecb_rotr16 (uint16_t x, unsigned int count)
415 473
416=item uint32_t ecb_rotr32 (uint32_t x, unsigned int count) 474=item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
417 475
418=item uint32_t ecb_rotl32 (uint32_t x, unsigned int count) 476=item uint64_t ecb_rotr64 (uint64_t x, unsigned int count)
419 477
420These two functions return the value of C<x> after rotating all the bits 478These two families of functions return the value of C<x> after rotating
421by C<count> positions to the right or left respectively. 479all the bits by C<count> positions to the right (C<ecb_rotr>) or left
480(C<ecb_rotl>).
422 481
423Current GCC versions understand these functions and usually compile them 482Current GCC versions understand these functions and usually compile them
424to "optimal" code (e.g. a single C<roll> on x86). 483to "optimal" code (e.g. a single C<rol> or a combination of C<shld> on
484x86).
425 485
426=back 486=back
427 487
428=head2 ARITHMETIC 488=head2 ARITHMETIC
429 489
430=over 4 490=over 4
431 491
432=item x = ecb_mod (m, n) 492=item x = ecb_mod (m, n)
433 493
434Returns the positive remainder of the modulo operation between C<m> and 494Returns C<m> modulo C<n>, which is the same as the positive remainder
495of the division operation between C<m> and C<n>, using floored
435C<n>. Unlike the C modulo operator C<%>, this function ensures that the 496division. Unlike the C remainder operator C<%>, this function ensures that
436return value is always positive - ISO C guarantees very little when 497the return value is always positive and that the two numbers I<m> and
437negative numbers are used with C<%>. 498I<m' = m + i * n> result in the same value modulo I<n> - in other words,
499C<ecb_mod> implements the mathematical modulo operation, which is missing
500in the language.
438 501
439C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be 502C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
440negatable, that is, both C<m> and C<-m> must be representable in its 503negatable, that is, both C<m> and C<-m> must be representable in its
441type. 504type (this typically excludes the minimum signed integer value, the same
505limitation as for C</> and C<%> in C).
506
507Current GCC versions compile this into an efficient branchless sequence on
508almost all CPUs.
509
510For example, when you want to rotate forward through the members of an
511array for increasing C<m> (which might be negative), then you should use
512C<ecb_mod>, as the C<%> operator might give either negative results, or
513change direction for negative values:
514
515 for (m = -100; m <= 100; ++m)
516 int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
442 517
443=back 518=back
444 519
445=head2 UTILITY 520=head2 UTILITY
446 521

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines