1 | /* |
1 | /* |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
2 | * libecb - http://software.schmorp.de/pkg/libecb |
3 | * |
3 | * |
4 | * Copyright (©) 2009-2014 Marc Alexander Lehmann <libecb@schmorp.de> |
4 | * Copyright (©) 2009-2015 Marc Alexander Lehmann <libecb@schmorp.de> |
5 | * Copyright (©) 2011 Emanuele Giaquinta |
5 | * Copyright (©) 2011 Emanuele Giaquinta |
6 | * All rights reserved. |
6 | * All rights reserved. |
7 | * |
7 | * |
8 | * Redistribution and use in source and binary forms, with or without modifica- |
8 | * Redistribution and use in source and binary forms, with or without modifica- |
9 | * tion, are permitted provided that the following conditions are met: |
9 | * tion, are permitted provided that the following conditions are met: |
… | |
… | |
74 | #else |
74 | #else |
75 | #define ECB_PTRSIZE 4 |
75 | #define ECB_PTRSIZE 4 |
76 | #endif |
76 | #endif |
77 | #endif |
77 | #endif |
78 | |
78 | |
|
|
79 | #define ECB_GCC_AMD64 (__amd64 || __amd64__ || __x86_64 || __x86_64__) |
|
|
80 | #define ECB_MSVC_AMD64 (_M_AMD64 || _M_X64) |
|
|
81 | |
79 | /* work around x32 idiocy by defining proper macros */ |
82 | /* work around x32 idiocy by defining proper macros */ |
80 | #if __amd64 || __x86_64 || _M_AMD64 || _M_X64 |
83 | #if ECB_GCC_AMD64 || ECB_MSVC_AMD64 |
81 | #if _ILP32 |
84 | #if _ILP32 |
82 | #define ECB_AMD64_X32 1 |
85 | #define ECB_AMD64_X32 1 |
83 | #else |
86 | #else |
84 | #define ECB_AMD64 1 |
87 | #define ECB_AMD64 1 |
85 | #endif |
88 | #endif |
… | |
… | |
98 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
101 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
99 | #endif |
102 | #endif |
100 | |
103 | |
101 | #define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor))) |
104 | #define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor))) |
102 | |
105 | |
103 | #if __clang__ && defined(__has_builtin) |
106 | #if __clang__ && defined __has_builtin |
104 | #define ECB_CLANG_BUILTIN(x) __has_builtin(x) |
107 | #define ECB_CLANG_BUILTIN(x) __has_builtin (x) |
105 | #else |
108 | #else |
106 | #define ECB_CLANG_BUILTIN(x) 0 |
109 | #define ECB_CLANG_BUILTIN(x) 0 |
107 | #endif |
110 | #endif |
108 | |
111 | |
109 | #if __clang__ && defined(__has_extension) |
112 | #if __clang__ && defined __has_extension |
110 | #define ECB_CLANG_EXTENSION(x) __has_extension(x) |
113 | #define ECB_CLANG_EXTENSION(x) __has_extension (x) |
111 | #else |
114 | #else |
112 | #define ECB_CLANG_EXTENSION(x) 0 |
115 | #define ECB_CLANG_EXTENSION(x) 0 |
113 | #endif |
116 | #endif |
114 | |
117 | |
115 | #define ECB_CPP (__cplusplus+0) |
118 | #define ECB_CPP (__cplusplus+0) |
… | |
… | |
153 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
156 | #if ECB_GCC_VERSION(2,5) || defined __INTEL_COMPILER || (__llvm__ && __GNUC__) || __SUNPRO_C >= 0x5110 || __SUNPRO_CC >= 0x5110 |
154 | #if __i386 || __i386__ |
157 | #if __i386 || __i386__ |
155 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
158 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("lock; orb $0, -1(%%esp)" : : : "memory") |
156 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
159 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
157 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
160 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
158 | #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
161 | #elif ECB_GCC_AMD64 |
159 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
162 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
160 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
163 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
161 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
164 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
162 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
165 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
163 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
166 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
… | |
… | |
278 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
281 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
279 | #endif |
282 | #endif |
280 | |
283 | |
281 | /*****************************************************************************/ |
284 | /*****************************************************************************/ |
282 | |
285 | |
283 | #if __cplusplus |
286 | #if ECB_CPP |
284 | #define ecb_inline static inline |
287 | #define ecb_inline static inline |
285 | #elif ECB_GCC_VERSION(2,5) |
288 | #elif ECB_GCC_VERSION(2,5) |
286 | #define ecb_inline static __inline__ |
289 | #define ecb_inline static __inline__ |
287 | #elif ECB_C99 |
290 | #elif ECB_C99 |
288 | #define ecb_inline static inline |
291 | #define ecb_inline static inline |
… | |
… | |
302 | |
305 | |
303 | #define ECB_CONCAT_(a, b) a ## b |
306 | #define ECB_CONCAT_(a, b) a ## b |
304 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
307 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
305 | #define ECB_STRINGIFY_(a) # a |
308 | #define ECB_STRINGIFY_(a) # a |
306 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
309 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
|
|
310 | #define ECB_STRINGIFY_EXPR(expr) ((expr), ECB_STRINGIFY_ (expr)) |
307 | |
311 | |
308 | #define ecb_function_ ecb_inline |
312 | #define ecb_function_ ecb_inline |
309 | |
313 | |
310 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
314 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
311 | #define ecb_attribute(attrlist) __attribute__ (attrlist) |
315 | #define ecb_attribute(attrlist) __attribute__ (attrlist) |
… | |
… | |
334 | #else |
338 | #else |
335 | #define ecb_prefetch(addr,rw,locality) |
339 | #define ecb_prefetch(addr,rw,locality) |
336 | #endif |
340 | #endif |
337 | |
341 | |
338 | /* no emulation for ecb_decltype */ |
342 | /* no emulation for ecb_decltype */ |
339 | #if ECB_GCC_VERSION(4,5) |
343 | #if ECB_CPP11 |
|
|
344 | // older implementations might have problems with decltype(x)::type, work around it |
|
|
345 | template<class T> struct ecb_decltype_t { typedef T type; }; |
340 | #define ecb_decltype(x) __decltype (x) |
346 | #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type |
341 | #elif ECB_GCC_VERSION(3,0) |
347 | #elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8) |
342 | #define ecb_decltype(x) __typeof (x) |
348 | #define ecb_decltype(x) __typeof__ (x) |
343 | #elif ECB_CPP11 |
|
|
344 | #define ecb_decltype(x) decltype (x) |
|
|
345 | #endif |
349 | #endif |
346 | |
350 | |
347 | #if _MSC_VER >= 1300 |
351 | #if _MSC_VER >= 1300 |
348 | #define ecb_deprecated __declspec(deprecated) |
352 | #define ecb_deprecated __declspec (deprecated) |
349 | #else |
353 | #else |
350 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
354 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
351 | #endif |
355 | #endif |
352 | |
356 | |
|
|
357 | #if _MSC_VER >= 1500 |
|
|
358 | #define ecb_deprecated_message(msg) __declspec (deprecated (msg)) |
|
|
359 | #elif ECB_GCC_VERSION(4,5) |
|
|
360 | #define ecb_deprecated_message(msg) ecb_attribute ((__deprecated__ (msg)) |
|
|
361 | #else |
|
|
362 | #define ecb_deprecated_message(msg) ecb_deprecated |
|
|
363 | #endif |
|
|
364 | |
|
|
365 | #if _MSC_VER >= 1400 |
|
|
366 | #define ecb_noinline __declspec (noinline) |
|
|
367 | #else |
353 | #define ecb_noinline ecb_attribute ((__noinline__)) |
368 | #define ecb_noinline ecb_attribute ((__noinline__)) |
|
|
369 | #endif |
|
|
370 | |
354 | #define ecb_unused ecb_attribute ((__unused__)) |
371 | #define ecb_unused ecb_attribute ((__unused__)) |
355 | #define ecb_const ecb_attribute ((__const__)) |
372 | #define ecb_const ecb_attribute ((__const__)) |
356 | #define ecb_pure ecb_attribute ((__pure__)) |
373 | #define ecb_pure ecb_attribute ((__pure__)) |
357 | |
374 | |
358 | /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */ |
375 | #if ECB_C11 || __IBMC_NORETURN |
359 | #if ECB_C11 |
376 | /* http://pic.dhe.ibm.com/infocenter/compbg/v121v141/topic/com.ibm.xlcpp121.bg.doc/language_ref/noreturn.html */ |
360 | #define ecb_noreturn _Noreturn |
377 | #define ecb_noreturn _Noreturn |
|
|
378 | #elif ECB_CPP11 |
|
|
379 | #define ecb_noreturn [[noreturn]] |
|
|
380 | #elif _MSC_VER >= 1200 |
|
|
381 | /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx */ |
|
|
382 | #define ecb_noreturn __declspec (noreturn) |
361 | #else |
383 | #else |
362 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
384 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
363 | #endif |
385 | #endif |
364 | |
386 | |
365 | #if ECB_GCC_VERSION(4,3) |
387 | #if ECB_GCC_VERSION(4,3) |
… | |
… | |
392 | #define ecb_ctz32(x) __builtin_ctz (x) |
414 | #define ecb_ctz32(x) __builtin_ctz (x) |
393 | #define ecb_ctz64(x) __builtin_ctzll (x) |
415 | #define ecb_ctz64(x) __builtin_ctzll (x) |
394 | #define ecb_popcount32(x) __builtin_popcount (x) |
416 | #define ecb_popcount32(x) __builtin_popcount (x) |
395 | /* no popcountll */ |
417 | /* no popcountll */ |
396 | #else |
418 | #else |
397 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
419 | ecb_function_ ecb_const int ecb_ctz32 (uint32_t x); |
398 | ecb_function_ int |
420 | ecb_function_ ecb_const int |
399 | ecb_ctz32 (uint32_t x) |
421 | ecb_ctz32 (uint32_t x) |
400 | { |
422 | { |
401 | int r = 0; |
423 | int r = 0; |
402 | |
424 | |
403 | x &= ~x + 1; /* this isolates the lowest bit */ |
425 | x &= ~x + 1; /* this isolates the lowest bit */ |
… | |
… | |
417 | #endif |
439 | #endif |
418 | |
440 | |
419 | return r; |
441 | return r; |
420 | } |
442 | } |
421 | |
443 | |
422 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
444 | ecb_function_ ecb_const int ecb_ctz64 (uint64_t x); |
423 | ecb_function_ int |
445 | ecb_function_ ecb_const int |
424 | ecb_ctz64 (uint64_t x) |
446 | ecb_ctz64 (uint64_t x) |
425 | { |
447 | { |
426 | int shift = x & 0xffffffffU ? 0 : 32; |
448 | int shift = x & 0xffffffffU ? 0 : 32; |
427 | return ecb_ctz32 (x >> shift) + shift; |
449 | return ecb_ctz32 (x >> shift) + shift; |
428 | } |
450 | } |
429 | |
451 | |
430 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
452 | ecb_function_ ecb_const int ecb_popcount32 (uint32_t x); |
431 | ecb_function_ int |
453 | ecb_function_ ecb_const int |
432 | ecb_popcount32 (uint32_t x) |
454 | ecb_popcount32 (uint32_t x) |
433 | { |
455 | { |
434 | x -= (x >> 1) & 0x55555555; |
456 | x -= (x >> 1) & 0x55555555; |
435 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
457 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
436 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
458 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
437 | x *= 0x01010101; |
459 | x *= 0x01010101; |
438 | |
460 | |
439 | return x >> 24; |
461 | return x >> 24; |
440 | } |
462 | } |
441 | |
463 | |
442 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
464 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x); |
443 | ecb_function_ int ecb_ld32 (uint32_t x) |
465 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x) |
444 | { |
466 | { |
445 | int r = 0; |
467 | int r = 0; |
446 | |
468 | |
447 | if (x >> 16) { x >>= 16; r += 16; } |
469 | if (x >> 16) { x >>= 16; r += 16; } |
448 | if (x >> 8) { x >>= 8; r += 8; } |
470 | if (x >> 8) { x >>= 8; r += 8; } |
… | |
… | |
451 | if (x >> 1) { r += 1; } |
473 | if (x >> 1) { r += 1; } |
452 | |
474 | |
453 | return r; |
475 | return r; |
454 | } |
476 | } |
455 | |
477 | |
456 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
478 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x); |
457 | ecb_function_ int ecb_ld64 (uint64_t x) |
479 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x) |
458 | { |
480 | { |
459 | int r = 0; |
481 | int r = 0; |
460 | |
482 | |
461 | if (x >> 32) { x >>= 32; r += 32; } |
483 | if (x >> 32) { x >>= 32; r += 32; } |
462 | |
484 | |
463 | return r + ecb_ld32 (x); |
485 | return r + ecb_ld32 (x); |
464 | } |
486 | } |
465 | #endif |
487 | #endif |
466 | |
488 | |
467 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
489 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x); |
468 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
490 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
469 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
491 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x); |
470 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
492 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
471 | |
493 | |
472 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
494 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x); |
473 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
495 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x) |
474 | { |
496 | { |
475 | return ( (x * 0x0802U & 0x22110U) |
497 | return ( (x * 0x0802U & 0x22110U) |
476 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
498 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
477 | } |
499 | } |
478 | |
500 | |
479 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
501 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x); |
480 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
502 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x) |
481 | { |
503 | { |
482 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
504 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
483 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
505 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
484 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
506 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
485 | x = ( x >> 8 ) | ( x << 8); |
507 | x = ( x >> 8 ) | ( x << 8); |
486 | |
508 | |
487 | return x; |
509 | return x; |
488 | } |
510 | } |
489 | |
511 | |
490 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
512 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x); |
491 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
513 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x) |
492 | { |
514 | { |
493 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
515 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
494 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
516 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
495 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
517 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
496 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
518 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
… | |
… | |
499 | return x; |
521 | return x; |
500 | } |
522 | } |
501 | |
523 | |
502 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
524 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
503 | /* so for this version we are lazy */ |
525 | /* so for this version we are lazy */ |
504 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
526 | ecb_function_ ecb_const int ecb_popcount64 (uint64_t x); |
505 | ecb_function_ int |
527 | ecb_function_ ecb_const int |
506 | ecb_popcount64 (uint64_t x) |
528 | ecb_popcount64 (uint64_t x) |
507 | { |
529 | { |
508 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
530 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
509 | } |
531 | } |
510 | |
532 | |
511 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
533 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count); |
512 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
534 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count); |
513 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
535 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count); |
514 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
536 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count); |
515 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
537 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count); |
516 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
538 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count); |
517 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
539 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count); |
518 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
540 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count); |
519 | |
541 | |
520 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
542 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
521 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
543 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
522 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
544 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
523 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
545 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
524 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
546 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
525 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
547 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
526 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
548 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
527 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
549 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
528 | |
550 | |
529 | #if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64)) |
551 | #if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64)) |
530 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
552 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
531 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
553 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
532 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
554 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
533 | #else |
555 | #else |
534 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
556 | ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x); |
535 | ecb_function_ uint16_t |
557 | ecb_function_ ecb_const uint16_t |
536 | ecb_bswap16 (uint16_t x) |
558 | ecb_bswap16 (uint16_t x) |
537 | { |
559 | { |
538 | return ecb_rotl16 (x, 8); |
560 | return ecb_rotl16 (x, 8); |
539 | } |
561 | } |
540 | |
562 | |
541 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
563 | ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x); |
542 | ecb_function_ uint32_t |
564 | ecb_function_ ecb_const uint32_t |
543 | ecb_bswap32 (uint32_t x) |
565 | ecb_bswap32 (uint32_t x) |
544 | { |
566 | { |
545 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
567 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
546 | } |
568 | } |
547 | |
569 | |
548 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
570 | ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x); |
549 | ecb_function_ uint64_t |
571 | ecb_function_ ecb_const uint64_t |
550 | ecb_bswap64 (uint64_t x) |
572 | ecb_bswap64 (uint64_t x) |
551 | { |
573 | { |
552 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
574 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
553 | } |
575 | } |
554 | #endif |
576 | #endif |
555 | |
577 | |
556 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
578 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
557 | #define ecb_unreachable() __builtin_unreachable () |
579 | #define ecb_unreachable() __builtin_unreachable () |
558 | #else |
580 | #else |
559 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
581 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
560 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
582 | ecb_inline ecb_noreturn void ecb_unreachable (void); |
561 | ecb_inline void ecb_unreachable (void) { } |
583 | ecb_inline ecb_noreturn void ecb_unreachable (void) { } |
562 | #endif |
584 | #endif |
563 | |
585 | |
564 | /* try to tell the compiler that some condition is definitely true */ |
586 | /* try to tell the compiler that some condition is definitely true */ |
565 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
587 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
566 | |
588 | |
567 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
589 | ecb_inline ecb_const unsigned char ecb_byteorder_helper (void); |
568 | ecb_inline unsigned char |
590 | ecb_inline ecb_const unsigned char |
569 | ecb_byteorder_helper (void) |
591 | ecb_byteorder_helper (void) |
570 | { |
592 | { |
571 | /* the union code still generates code under pressure in gcc, */ |
593 | /* the union code still generates code under pressure in gcc, */ |
572 | /* but less than using pointers, and always seems to */ |
594 | /* but less than using pointers, and always seems to */ |
573 | /* successfully return a constant. */ |
595 | /* successfully return a constant. */ |
574 | /* the reason why we have this horrible preprocessor mess */ |
596 | /* the reason why we have this horrible preprocessor mess */ |
575 | /* is to avoid it in all cases, at least on common architectures */ |
597 | /* is to avoid it in all cases, at least on common architectures */ |
576 | /* or when using a recent enough gcc version (>= 4.6) */ |
598 | /* or when using a recent enough gcc version (>= 4.6) */ |
577 | #if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64 |
599 | #if ((__i386 || __i386__) && !__VOS__) || _M_IX86 || ECB_GCC_AMD64 || ECB_MSVC_AMD64 |
578 | return 0x44; |
600 | return 0x44; |
579 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ |
601 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ |
580 | return 0x44; |
602 | return 0x44; |
581 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ |
603 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ |
582 | return 0x11; |
604 | return 0x11; |
… | |
… | |
588 | } u = { 0x11223344 }; |
610 | } u = { 0x11223344 }; |
589 | return u.c; |
611 | return u.c; |
590 | #endif |
612 | #endif |
591 | } |
613 | } |
592 | |
614 | |
593 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
615 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
594 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
616 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
595 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
617 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
596 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
618 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
597 | |
619 | |
598 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
620 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
599 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
621 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
600 | #else |
622 | #else |
601 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
623 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
602 | #endif |
624 | #endif |
603 | |
625 | |
604 | #if __cplusplus |
626 | #if ECB_CPP |
605 | template<typename T> |
627 | template<typename T> |
606 | static inline T ecb_div_rd (T val, T div) |
628 | static inline T ecb_div_rd (T val, T div) |
607 | { |
629 | { |
608 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
630 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
609 | } |
631 | } |
… | |
… | |
633 | |
655 | |
634 | /* basically, everything uses "ieee pure-endian" floating point numbers */ |
656 | /* basically, everything uses "ieee pure-endian" floating point numbers */ |
635 | /* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
657 | /* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
636 | #if 0 \ |
658 | #if 0 \ |
637 | || __i386 || __i386__ \ |
659 | || __i386 || __i386__ \ |
638 | || __amd64 || __amd64__ || __x86_64 || __x86_64__ \ |
660 | || ECB_GCC_AMD64 \ |
639 | || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
661 | || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
640 | || defined __s390__ || defined __s390x__ \ |
662 | || defined __s390__ || defined __s390x__ \ |
641 | || defined __mips__ \ |
663 | || defined __mips__ \ |
642 | || defined __alpha__ \ |
664 | || defined __alpha__ \ |
643 | || defined __hppa__ \ |
665 | || defined __hppa__ \ |
644 | || defined __ia64__ \ |
666 | || defined __ia64__ \ |
645 | || defined __m68k__ \ |
667 | || defined __m68k__ \ |
646 | || defined __m88k__ \ |
668 | || defined __m88k__ \ |
647 | || defined __sh__ \ |
669 | || defined __sh__ \ |
648 | || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 \ |
670 | || defined _M_IX86 || defined ECB_MSVC_AMD64 || defined _M_IA64 \ |
649 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
671 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
650 | || defined __aarch64__ |
672 | || defined __aarch64__ |
651 | #define ECB_STDFP 1 |
673 | #define ECB_STDFP 1 |
652 | #include <string.h> /* for memcpy */ |
674 | #include <string.h> /* for memcpy */ |
653 | #else |
675 | #else |
… | |
… | |
669 | #define ECB_NAN NAN |
691 | #define ECB_NAN NAN |
670 | #else |
692 | #else |
671 | #define ECB_NAN ECB_INFINITY |
693 | #define ECB_NAN ECB_INFINITY |
672 | #endif |
694 | #endif |
673 | |
695 | |
|
|
696 | #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L |
|
|
697 | #define ecb_ldexpf(x,e) ldexpf ((x), (e)) |
|
|
698 | #define ecb_frexpf(x,e) frexpf ((x), (e)) |
|
|
699 | #else |
|
|
700 | #define ecb_ldexpf(x,e) (float) ldexp ((double) (x), (e)) |
|
|
701 | #define ecb_frexpf(x,e) (float) frexp ((double) (x), (e)) |
|
|
702 | #endif |
|
|
703 | |
674 | /* converts an ieee half/binary16 to a float */ |
704 | /* converts an ieee half/binary16 to a float */ |
675 | ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; |
705 | ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x); |
676 | ecb_function_ float |
706 | ecb_function_ ecb_const float |
677 | ecb_binary16_to_float (uint16_t x) |
707 | ecb_binary16_to_float (uint16_t x) |
678 | { |
708 | { |
679 | int e = (x >> 10) & 0x1f; |
709 | int e = (x >> 10) & 0x1f; |
680 | int m = x & 0x3ff; |
710 | int m = x & 0x3ff; |
681 | float r; |
711 | float r; |
682 | |
712 | |
683 | if (!e ) r = ldexpf (m , -24); |
713 | if (!e ) r = ecb_ldexpf (m , -24); |
684 | else if (e != 31) r = ldexpf (m + 0x400, e - 25); |
714 | else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25); |
685 | else if (m ) r = ECB_NAN; |
715 | else if (m ) r = ECB_NAN; |
686 | else r = ECB_INFINITY; |
716 | else r = ECB_INFINITY; |
687 | |
717 | |
688 | return x & 0x8000 ? -r : r; |
718 | return x & 0x8000 ? -r : r; |
689 | } |
719 | } |
690 | |
720 | |
691 | /* convert a float to ieee single/binary32 */ |
721 | /* convert a float to ieee single/binary32 */ |
692 | ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
722 | ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x); |
693 | ecb_function_ uint32_t |
723 | ecb_function_ ecb_const uint32_t |
694 | ecb_float_to_binary32 (float x) |
724 | ecb_float_to_binary32 (float x) |
695 | { |
725 | { |
696 | uint32_t r; |
726 | uint32_t r; |
697 | |
727 | |
698 | #if ECB_STDFP |
728 | #if ECB_STDFP |
… | |
… | |
705 | if (x == 0e0f ) return 0x00000000U; |
735 | if (x == 0e0f ) return 0x00000000U; |
706 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
736 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
707 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
737 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
708 | if (x != x ) return 0x7fbfffffU; |
738 | if (x != x ) return 0x7fbfffffU; |
709 | |
739 | |
710 | m = frexpf (x, &e) * 0x1000000U; |
740 | m = ecb_frexpf (x, &e) * 0x1000000U; |
711 | |
741 | |
712 | r = m & 0x80000000U; |
742 | r = m & 0x80000000U; |
713 | |
743 | |
714 | if (r) |
744 | if (r) |
715 | m = -m; |
745 | m = -m; |
… | |
… | |
727 | |
757 | |
728 | return r; |
758 | return r; |
729 | } |
759 | } |
730 | |
760 | |
731 | /* converts an ieee single/binary32 to a float */ |
761 | /* converts an ieee single/binary32 to a float */ |
732 | ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
762 | ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x); |
733 | ecb_function_ float |
763 | ecb_function_ ecb_const float |
734 | ecb_binary32_to_float (uint32_t x) |
764 | ecb_binary32_to_float (uint32_t x) |
735 | { |
765 | { |
736 | float r; |
766 | float r; |
737 | |
767 | |
738 | #if ECB_STDFP |
768 | #if ECB_STDFP |
… | |
… | |
748 | x |= 0x800000U; |
778 | x |= 0x800000U; |
749 | else |
779 | else |
750 | e = 1; |
780 | e = 1; |
751 | |
781 | |
752 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
782 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
753 | r = ldexpf (x * (0.5f / 0x800000U), e - 126); |
783 | r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126); |
754 | |
784 | |
755 | r = neg ? -r : r; |
785 | r = neg ? -r : r; |
756 | #endif |
786 | #endif |
757 | |
787 | |
758 | return r; |
788 | return r; |
759 | } |
789 | } |
760 | |
790 | |
761 | /* convert a double to ieee double/binary64 */ |
791 | /* convert a double to ieee double/binary64 */ |
762 | ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
792 | ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x); |
763 | ecb_function_ uint64_t |
793 | ecb_function_ ecb_const uint64_t |
764 | ecb_double_to_binary64 (double x) |
794 | ecb_double_to_binary64 (double x) |
765 | { |
795 | { |
766 | uint64_t r; |
796 | uint64_t r; |
767 | |
797 | |
768 | #if ECB_STDFP |
798 | #if ECB_STDFP |
… | |
… | |
797 | |
827 | |
798 | return r; |
828 | return r; |
799 | } |
829 | } |
800 | |
830 | |
801 | /* converts an ieee double/binary64 to a double */ |
831 | /* converts an ieee double/binary64 to a double */ |
802 | ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; |
832 | ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x); |
803 | ecb_function_ double |
833 | ecb_function_ ecb_const double |
804 | ecb_binary64_to_double (uint64_t x) |
834 | ecb_binary64_to_double (uint64_t x) |
805 | { |
835 | { |
806 | double r; |
836 | double r; |
807 | |
837 | |
808 | #if ECB_STDFP |
838 | #if ECB_STDFP |