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: |
… | |
… | |
40 | |
40 | |
41 | #ifndef ECB_H |
41 | #ifndef ECB_H |
42 | #define ECB_H |
42 | #define ECB_H |
43 | |
43 | |
44 | /* 16 bits major, 16 bits minor */ |
44 | /* 16 bits major, 16 bits minor */ |
45 | #define ECB_VERSION 0x00010003 |
45 | #define ECB_VERSION 0x00010004 |
46 | |
46 | |
47 | #ifdef _WIN32 |
47 | #ifdef _WIN32 |
48 | typedef signed char int8_t; |
48 | typedef signed char int8_t; |
49 | typedef unsigned char uint8_t; |
49 | typedef unsigned char uint8_t; |
50 | typedef signed short int16_t; |
50 | typedef signed short int16_t; |
… | |
… | |
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 |
|
|
110 | #endif |
|
|
111 | |
|
|
112 | #if __clang__ && defined __has_extension |
|
|
113 | #define ECB_CLANG_EXTENSION(x) __has_extension (x) |
|
|
114 | #else |
|
|
115 | #define ECB_CLANG_EXTENSION(x) 0 |
107 | #endif |
116 | #endif |
108 | |
117 | |
109 | #define ECB_CPP (__cplusplus+0) |
118 | #define ECB_CPP (__cplusplus+0) |
110 | #define ECB_CPP11 (__cplusplus >= 201103L) |
119 | #define ECB_CPP11 (__cplusplus >= 201103L) |
111 | |
120 | |
… | |
… | |
147 | #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 |
148 | #if __i386 || __i386__ |
157 | #if __i386 || __i386__ |
149 | #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") |
150 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
159 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
151 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
160 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
152 | #elif __amd64 || __amd64__ || __x86_64 || __x86_64__ |
161 | #elif ECB_GCC_AMD64 |
153 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
162 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("mfence" : : : "memory") |
154 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
163 | #define ECB_MEMORY_FENCE_ACQUIRE __asm__ __volatile__ ("" : : : "memory") |
155 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
164 | #define ECB_MEMORY_FENCE_RELEASE __asm__ __volatile__ ("") |
156 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
165 | #elif __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ |
157 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
166 | #define ECB_MEMORY_FENCE __asm__ __volatile__ ("sync" : : : "memory") |
… | |
… | |
195 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
204 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
196 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
205 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
197 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
206 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
198 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
207 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
199 | |
208 | |
200 | /* The __has_feature syntax from clang is so misdesigned that we cannot use it |
209 | #elif ECB_CLANG_EXTENSION(c_atomic) |
201 | * without risking compile time errors with other compilers. We *could* |
|
|
202 | * define our own ecb_clang_has_feature, but I just can't be bothered to work |
|
|
203 | * around this shit time and again. |
|
|
204 | * #elif defined __clang && __has_feature (cxx_atomic) |
|
|
205 | * // see comment below (stdatomic.h) about the C11 memory model. |
210 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
206 | * #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
211 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
207 | * #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
212 | #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
208 | * #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
213 | #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
209 | */ |
|
|
210 | |
214 | |
211 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
215 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
212 | #define ECB_MEMORY_FENCE __sync_synchronize () |
216 | #define ECB_MEMORY_FENCE __sync_synchronize () |
213 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
217 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
214 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
218 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
… | |
… | |
277 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
281 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
278 | #endif |
282 | #endif |
279 | |
283 | |
280 | /*****************************************************************************/ |
284 | /*****************************************************************************/ |
281 | |
285 | |
282 | #if __cplusplus |
286 | #if ECB_CPP |
283 | #define ecb_inline static inline |
287 | #define ecb_inline static inline |
284 | #elif ECB_GCC_VERSION(2,5) |
288 | #elif ECB_GCC_VERSION(2,5) |
285 | #define ecb_inline static __inline__ |
289 | #define ecb_inline static __inline__ |
286 | #elif ECB_C99 |
290 | #elif ECB_C99 |
287 | #define ecb_inline static inline |
291 | #define ecb_inline static inline |
… | |
… | |
301 | |
305 | |
302 | #define ECB_CONCAT_(a, b) a ## b |
306 | #define ECB_CONCAT_(a, b) a ## b |
303 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
307 | #define ECB_CONCAT(a, b) ECB_CONCAT_(a, b) |
304 | #define ECB_STRINGIFY_(a) # a |
308 | #define ECB_STRINGIFY_(a) # a |
305 | #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)) |
306 | |
311 | |
307 | #define ecb_function_ ecb_inline |
312 | #define ecb_function_ ecb_inline |
308 | |
313 | |
309 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
314 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
310 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
315 | #define ecb_attribute(attrlist) __attribute__ (attrlist) |
311 | #else |
316 | #else |
312 | #define ecb_attribute(attrlist) |
317 | #define ecb_attribute(attrlist) |
313 | #endif |
318 | #endif |
314 | |
319 | |
315 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p) |
320 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p) |
… | |
… | |
333 | #else |
338 | #else |
334 | #define ecb_prefetch(addr,rw,locality) |
339 | #define ecb_prefetch(addr,rw,locality) |
335 | #endif |
340 | #endif |
336 | |
341 | |
337 | /* no emulation for ecb_decltype */ |
342 | /* no emulation for ecb_decltype */ |
338 | #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; }; |
339 | #define ecb_decltype(x) __decltype(x) |
346 | #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type |
340 | #elif ECB_GCC_VERSION(3,0) |
347 | #elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8) |
341 | #define ecb_decltype(x) __typeof(x) |
348 | #define ecb_decltype(x) __typeof__ (x) |
342 | #endif |
349 | #endif |
343 | |
350 | |
344 | #if _MSC_VER >= 1300 |
351 | #if _MSC_VER >= 1300 |
345 | #define ecb_deprecated __declspec(deprecated) |
352 | #define ecb_deprecated __declspec (deprecated) |
346 | #else |
353 | #else |
347 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
354 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
348 | #endif |
355 | #endif |
349 | |
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 |
350 | #define ecb_noinline ecb_attribute ((__noinline__)) |
368 | #define ecb_noinline ecb_attribute ((__noinline__)) |
|
|
369 | #endif |
|
|
370 | |
351 | #define ecb_unused ecb_attribute ((__unused__)) |
371 | #define ecb_unused ecb_attribute ((__unused__)) |
352 | #define ecb_const ecb_attribute ((__const__)) |
372 | #define ecb_const ecb_attribute ((__const__)) |
353 | #define ecb_pure ecb_attribute ((__pure__)) |
373 | #define ecb_pure ecb_attribute ((__pure__)) |
354 | |
374 | |
355 | /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */ |
375 | #if ECB_C11 || __IBMC_NORETURN |
356 | #if ECB_C11 |
376 | /* http://pic.dhe.ibm.com/infocenter/compbg/v121v141/topic/com.ibm.xlcpp121.bg.doc/language_ref/noreturn.html */ |
357 | #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) |
358 | #else |
383 | #else |
359 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
384 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
360 | #endif |
385 | #endif |
361 | |
386 | |
362 | #if ECB_GCC_VERSION(4,3) |
387 | #if ECB_GCC_VERSION(4,3) |
… | |
… | |
389 | #define ecb_ctz32(x) __builtin_ctz (x) |
414 | #define ecb_ctz32(x) __builtin_ctz (x) |
390 | #define ecb_ctz64(x) __builtin_ctzll (x) |
415 | #define ecb_ctz64(x) __builtin_ctzll (x) |
391 | #define ecb_popcount32(x) __builtin_popcount (x) |
416 | #define ecb_popcount32(x) __builtin_popcount (x) |
392 | /* no popcountll */ |
417 | /* no popcountll */ |
393 | #else |
418 | #else |
394 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
419 | ecb_function_ ecb_const int ecb_ctz32 (uint32_t x); |
395 | ecb_function_ int |
420 | ecb_function_ ecb_const int |
396 | ecb_ctz32 (uint32_t x) |
421 | ecb_ctz32 (uint32_t x) |
397 | { |
422 | { |
398 | int r = 0; |
423 | int r = 0; |
399 | |
424 | |
400 | x &= ~x + 1; /* this isolates the lowest bit */ |
425 | x &= ~x + 1; /* this isolates the lowest bit */ |
… | |
… | |
414 | #endif |
439 | #endif |
415 | |
440 | |
416 | return r; |
441 | return r; |
417 | } |
442 | } |
418 | |
443 | |
419 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
444 | ecb_function_ ecb_const int ecb_ctz64 (uint64_t x); |
420 | ecb_function_ int |
445 | ecb_function_ ecb_const int |
421 | ecb_ctz64 (uint64_t x) |
446 | ecb_ctz64 (uint64_t x) |
422 | { |
447 | { |
423 | int shift = x & 0xffffffffU ? 0 : 32; |
448 | int shift = x & 0xffffffffU ? 0 : 32; |
424 | return ecb_ctz32 (x >> shift) + shift; |
449 | return ecb_ctz32 (x >> shift) + shift; |
425 | } |
450 | } |
426 | |
451 | |
427 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
452 | ecb_function_ ecb_const int ecb_popcount32 (uint32_t x); |
428 | ecb_function_ int |
453 | ecb_function_ ecb_const int |
429 | ecb_popcount32 (uint32_t x) |
454 | ecb_popcount32 (uint32_t x) |
430 | { |
455 | { |
431 | x -= (x >> 1) & 0x55555555; |
456 | x -= (x >> 1) & 0x55555555; |
432 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
457 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
433 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
458 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
434 | x *= 0x01010101; |
459 | x *= 0x01010101; |
435 | |
460 | |
436 | return x >> 24; |
461 | return x >> 24; |
437 | } |
462 | } |
438 | |
463 | |
439 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
464 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x); |
440 | ecb_function_ int ecb_ld32 (uint32_t x) |
465 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x) |
441 | { |
466 | { |
442 | int r = 0; |
467 | int r = 0; |
443 | |
468 | |
444 | if (x >> 16) { x >>= 16; r += 16; } |
469 | if (x >> 16) { x >>= 16; r += 16; } |
445 | if (x >> 8) { x >>= 8; r += 8; } |
470 | if (x >> 8) { x >>= 8; r += 8; } |
… | |
… | |
448 | if (x >> 1) { r += 1; } |
473 | if (x >> 1) { r += 1; } |
449 | |
474 | |
450 | return r; |
475 | return r; |
451 | } |
476 | } |
452 | |
477 | |
453 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
478 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x); |
454 | ecb_function_ int ecb_ld64 (uint64_t x) |
479 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x) |
455 | { |
480 | { |
456 | int r = 0; |
481 | int r = 0; |
457 | |
482 | |
458 | if (x >> 32) { x >>= 32; r += 32; } |
483 | if (x >> 32) { x >>= 32; r += 32; } |
459 | |
484 | |
460 | return r + ecb_ld32 (x); |
485 | return r + ecb_ld32 (x); |
461 | } |
486 | } |
462 | #endif |
487 | #endif |
463 | |
488 | |
464 | 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); |
465 | 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)); } |
466 | 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); |
467 | 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)); } |
468 | |
493 | |
469 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
494 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x); |
470 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
495 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x) |
471 | { |
496 | { |
472 | return ( (x * 0x0802U & 0x22110U) |
497 | return ( (x * 0x0802U & 0x22110U) |
473 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
498 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
474 | } |
499 | } |
475 | |
500 | |
476 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
501 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x); |
477 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
502 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x) |
478 | { |
503 | { |
479 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
504 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
480 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
505 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
481 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
506 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
482 | x = ( x >> 8 ) | ( x << 8); |
507 | x = ( x >> 8 ) | ( x << 8); |
483 | |
508 | |
484 | return x; |
509 | return x; |
485 | } |
510 | } |
486 | |
511 | |
487 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
512 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x); |
488 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
513 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x) |
489 | { |
514 | { |
490 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
515 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
491 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
516 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
492 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
517 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
493 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
518 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
… | |
… | |
496 | return x; |
521 | return x; |
497 | } |
522 | } |
498 | |
523 | |
499 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
524 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
500 | /* so for this version we are lazy */ |
525 | /* so for this version we are lazy */ |
501 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
526 | ecb_function_ ecb_const int ecb_popcount64 (uint64_t x); |
502 | ecb_function_ int |
527 | ecb_function_ ecb_const int |
503 | ecb_popcount64 (uint64_t x) |
528 | ecb_popcount64 (uint64_t x) |
504 | { |
529 | { |
505 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
530 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
506 | } |
531 | } |
507 | |
532 | |
508 | 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); |
509 | 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); |
510 | 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); |
511 | 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); |
512 | 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); |
513 | 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); |
514 | 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); |
515 | 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); |
516 | |
541 | |
517 | 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); } |
518 | 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); } |
519 | 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); } |
520 | 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); } |
521 | 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); } |
522 | 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); } |
523 | 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); } |
524 | 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); } |
525 | |
550 | |
526 | #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)) |
|
|
552 | #if ECB_GCC_VERSION(4,8) || ECB_CLANG_BUILTIN(__builtin_bswap16) |
|
|
553 | #define ecb_bswap16(x) __builtin_bswap16 (x) |
|
|
554 | #else |
527 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
555 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
|
|
556 | #endif |
528 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
557 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
529 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
558 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
|
|
559 | #elif _MSC_VER |
|
|
560 | #include <stdlib.h> |
|
|
561 | #define ecb_bswap16(x) ((uint16_t)_byteswap_ushort ((uint16_t)(x))) |
|
|
562 | #define ecb_bswap32(x) ((uint32_t)_byteswap_ulong ((uint32_t)(x))) |
|
|
563 | #define ecb_bswap64(x) ((uint64_t)_byteswap_uint64 ((uint64_t)(x))) |
530 | #else |
564 | #else |
531 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
565 | ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x); |
532 | ecb_function_ uint16_t |
566 | ecb_function_ ecb_const uint16_t |
533 | ecb_bswap16 (uint16_t x) |
567 | ecb_bswap16 (uint16_t x) |
534 | { |
568 | { |
535 | return ecb_rotl16 (x, 8); |
569 | return ecb_rotl16 (x, 8); |
536 | } |
570 | } |
537 | |
571 | |
538 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
572 | ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x); |
539 | ecb_function_ uint32_t |
573 | ecb_function_ ecb_const uint32_t |
540 | ecb_bswap32 (uint32_t x) |
574 | ecb_bswap32 (uint32_t x) |
541 | { |
575 | { |
542 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
576 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
543 | } |
577 | } |
544 | |
578 | |
545 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
579 | ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x); |
546 | ecb_function_ uint64_t |
580 | ecb_function_ ecb_const uint64_t |
547 | ecb_bswap64 (uint64_t x) |
581 | ecb_bswap64 (uint64_t x) |
548 | { |
582 | { |
549 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
583 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
550 | } |
584 | } |
551 | #endif |
585 | #endif |
552 | |
586 | |
553 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
587 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
554 | #define ecb_unreachable() __builtin_unreachable () |
588 | #define ecb_unreachable() __builtin_unreachable () |
555 | #else |
589 | #else |
556 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
590 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
557 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
591 | ecb_inline ecb_noreturn void ecb_unreachable (void); |
558 | ecb_inline void ecb_unreachable (void) { } |
592 | ecb_inline ecb_noreturn void ecb_unreachable (void) { } |
559 | #endif |
593 | #endif |
560 | |
594 | |
561 | /* try to tell the compiler that some condition is definitely true */ |
595 | /* try to tell the compiler that some condition is definitely true */ |
562 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
596 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
563 | |
597 | |
564 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
598 | ecb_inline ecb_const unsigned char ecb_byteorder_helper (void); |
565 | ecb_inline unsigned char |
599 | ecb_inline ecb_const unsigned char |
566 | ecb_byteorder_helper (void) |
600 | ecb_byteorder_helper (void) |
567 | { |
601 | { |
568 | /* the union code still generates code under pressure in gcc, */ |
602 | /* the union code still generates code under pressure in gcc, */ |
569 | /* but less than using pointers, and always seems to */ |
603 | /* but less than using pointers, and always seems to */ |
570 | /* successfully return a constant. */ |
604 | /* successfully return a constant. */ |
571 | /* the reason why we have this horrible preprocessor mess */ |
605 | /* the reason why we have this horrible preprocessor mess */ |
572 | /* is to avoid it in all cases, at least on common architectures */ |
606 | /* is to avoid it in all cases, at least on common architectures */ |
573 | /* or when using a recent enough gcc version (>= 4.6) */ |
607 | /* or when using a recent enough gcc version (>= 4.6) */ |
574 | #if __i386 || __i386__ || _M_X86 || __amd64 || __amd64__ || _M_X64 |
608 | #if ((__i386 || __i386__) && !__VOS__) || _M_IX86 || ECB_GCC_AMD64 || ECB_MSVC_AMD64 |
575 | return 0x44; |
609 | return 0x44; |
576 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ |
610 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ |
577 | return 0x44; |
611 | return 0x44; |
578 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ |
612 | #elif __BYTE_ORDER__ && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ |
579 | return 0x11; |
613 | return 0x11; |
… | |
… | |
585 | } u = { 0x11223344 }; |
619 | } u = { 0x11223344 }; |
586 | return u.c; |
620 | return u.c; |
587 | #endif |
621 | #endif |
588 | } |
622 | } |
589 | |
623 | |
590 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
624 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
591 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
625 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
592 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
626 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
593 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
627 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
594 | |
628 | |
595 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
629 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
596 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
630 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
597 | #else |
631 | #else |
598 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
632 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
599 | #endif |
633 | #endif |
600 | |
634 | |
601 | #if __cplusplus |
635 | #if ECB_CPP |
602 | template<typename T> |
636 | template<typename T> |
603 | static inline T ecb_div_rd (T val, T div) |
637 | static inline T ecb_div_rd (T val, T div) |
604 | { |
638 | { |
605 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
639 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
606 | } |
640 | } |
… | |
… | |
630 | |
664 | |
631 | /* basically, everything uses "ieee pure-endian" floating point numbers */ |
665 | /* basically, everything uses "ieee pure-endian" floating point numbers */ |
632 | /* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
666 | /* the only noteworthy exception is ancient armle, which uses order 43218765 */ |
633 | #if 0 \ |
667 | #if 0 \ |
634 | || __i386 || __i386__ \ |
668 | || __i386 || __i386__ \ |
635 | || __amd64 || __amd64__ || __x86_64 || __x86_64__ \ |
669 | || ECB_GCC_AMD64 \ |
636 | || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
670 | || __powerpc__ || __ppc__ || __powerpc64__ || __ppc64__ \ |
637 | || defined __s390__ || defined __s390x__ \ |
671 | || defined __s390__ || defined __s390x__ \ |
638 | || defined __mips__ \ |
672 | || defined __mips__ \ |
639 | || defined __alpha__ \ |
673 | || defined __alpha__ \ |
640 | || defined __hppa__ \ |
674 | || defined __hppa__ \ |
641 | || defined __ia64__ \ |
675 | || defined __ia64__ \ |
642 | || defined __m68k__ \ |
676 | || defined __m68k__ \ |
643 | || defined __m88k__ \ |
677 | || defined __m88k__ \ |
644 | || defined __sh__ \ |
678 | || defined __sh__ \ |
645 | || defined _M_IX86 || defined _M_AMD64 || defined _M_IA64 \ |
679 | || defined _M_IX86 || defined ECB_MSVC_AMD64 || defined _M_IA64 \ |
646 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
680 | || (defined __arm__ && (defined __ARM_EABI__ || defined __EABI__ || defined __VFP_FP__ || defined _WIN32_WCE || defined __ANDROID__)) \ |
647 | || defined __aarch64__ |
681 | || defined __aarch64__ |
648 | #define ECB_STDFP 1 |
682 | #define ECB_STDFP 1 |
649 | #include <string.h> /* for memcpy */ |
683 | #include <string.h> /* for memcpy */ |
650 | #else |
684 | #else |
… | |
… | |
666 | #define ECB_NAN NAN |
700 | #define ECB_NAN NAN |
667 | #else |
701 | #else |
668 | #define ECB_NAN ECB_INFINITY |
702 | #define ECB_NAN ECB_INFINITY |
669 | #endif |
703 | #endif |
670 | |
704 | |
|
|
705 | #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L |
|
|
706 | #define ecb_ldexpf(x,e) ldexpf ((x), (e)) |
|
|
707 | #define ecb_frexpf(x,e) frexpf ((x), (e)) |
|
|
708 | #else |
|
|
709 | #define ecb_ldexpf(x,e) (float) ldexp ((double) (x), (e)) |
|
|
710 | #define ecb_frexpf(x,e) (float) frexp ((double) (x), (e)) |
|
|
711 | #endif |
|
|
712 | |
671 | /* converts an ieee half/binary16 to a float */ |
713 | /* converts an ieee half/binary16 to a float */ |
672 | ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; |
714 | ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x); |
673 | ecb_function_ float |
715 | ecb_function_ ecb_const float |
674 | ecb_binary16_to_float (uint16_t x) |
716 | ecb_binary16_to_float (uint16_t x) |
675 | { |
717 | { |
676 | int e = (x >> 10) & 0x1f; |
718 | int e = (x >> 10) & 0x1f; |
677 | int m = x & 0x3ff; |
719 | int m = x & 0x3ff; |
678 | float r; |
720 | float r; |
679 | |
721 | |
680 | if (!e ) r = ldexpf (m , -24); |
722 | if (!e ) r = ecb_ldexpf (m , -24); |
681 | else if (e != 31) r = ldexpf (m + 0x400, e - 25); |
723 | else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25); |
682 | else if (m ) r = ECB_NAN; |
724 | else if (m ) r = ECB_NAN; |
683 | else r = ECB_INFINITY; |
725 | else r = ECB_INFINITY; |
684 | |
726 | |
685 | return x & 0x8000 ? -r : r; |
727 | return x & 0x8000 ? -r : r; |
686 | } |
728 | } |
687 | |
729 | |
688 | /* convert a float to ieee single/binary32 */ |
730 | /* convert a float to ieee single/binary32 */ |
689 | ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
731 | ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x); |
690 | ecb_function_ uint32_t |
732 | ecb_function_ ecb_const uint32_t |
691 | ecb_float_to_binary32 (float x) |
733 | ecb_float_to_binary32 (float x) |
692 | { |
734 | { |
693 | uint32_t r; |
735 | uint32_t r; |
694 | |
736 | |
695 | #if ECB_STDFP |
737 | #if ECB_STDFP |
… | |
… | |
702 | if (x == 0e0f ) return 0x00000000U; |
744 | if (x == 0e0f ) return 0x00000000U; |
703 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
745 | if (x > +3.40282346638528860e+38f) return 0x7f800000U; |
704 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
746 | if (x < -3.40282346638528860e+38f) return 0xff800000U; |
705 | if (x != x ) return 0x7fbfffffU; |
747 | if (x != x ) return 0x7fbfffffU; |
706 | |
748 | |
707 | m = frexpf (x, &e) * 0x1000000U; |
749 | m = ecb_frexpf (x, &e) * 0x1000000U; |
708 | |
750 | |
709 | r = m & 0x80000000U; |
751 | r = m & 0x80000000U; |
710 | |
752 | |
711 | if (r) |
753 | if (r) |
712 | m = -m; |
754 | m = -m; |
… | |
… | |
724 | |
766 | |
725 | return r; |
767 | return r; |
726 | } |
768 | } |
727 | |
769 | |
728 | /* converts an ieee single/binary32 to a float */ |
770 | /* converts an ieee single/binary32 to a float */ |
729 | ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
771 | ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x); |
730 | ecb_function_ float |
772 | ecb_function_ ecb_const float |
731 | ecb_binary32_to_float (uint32_t x) |
773 | ecb_binary32_to_float (uint32_t x) |
732 | { |
774 | { |
733 | float r; |
775 | float r; |
734 | |
776 | |
735 | #if ECB_STDFP |
777 | #if ECB_STDFP |
… | |
… | |
745 | x |= 0x800000U; |
787 | x |= 0x800000U; |
746 | else |
788 | else |
747 | e = 1; |
789 | e = 1; |
748 | |
790 | |
749 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
791 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
750 | r = ldexpf (x * (0.5f / 0x800000U), e - 126); |
792 | r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126); |
751 | |
793 | |
752 | r = neg ? -r : r; |
794 | r = neg ? -r : r; |
753 | #endif |
795 | #endif |
754 | |
796 | |
755 | return r; |
797 | return r; |
756 | } |
798 | } |
757 | |
799 | |
758 | /* convert a double to ieee double/binary64 */ |
800 | /* convert a double to ieee double/binary64 */ |
759 | ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
801 | ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x); |
760 | ecb_function_ uint64_t |
802 | ecb_function_ ecb_const uint64_t |
761 | ecb_double_to_binary64 (double x) |
803 | ecb_double_to_binary64 (double x) |
762 | { |
804 | { |
763 | uint64_t r; |
805 | uint64_t r; |
764 | |
806 | |
765 | #if ECB_STDFP |
807 | #if ECB_STDFP |
… | |
… | |
794 | |
836 | |
795 | return r; |
837 | return r; |
796 | } |
838 | } |
797 | |
839 | |
798 | /* converts an ieee double/binary64 to a double */ |
840 | /* converts an ieee double/binary64 to a double */ |
799 | ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; |
841 | ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x); |
800 | ecb_function_ double |
842 | ecb_function_ ecb_const double |
801 | ecb_binary64_to_double (uint64_t x) |
843 | ecb_binary64_to_double (uint64_t x) |
802 | { |
844 | { |
803 | double r; |
845 | double r; |
804 | |
846 | |
805 | #if ECB_STDFP |
847 | #if ECB_STDFP |