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; |
… | |
… | |
90 | * causing enormous grief in return for some better fake benchmark numbers. |
90 | * causing enormous grief in return for some better fake benchmark numbers. |
91 | * or so. |
91 | * or so. |
92 | * we try to detect these and simply assume they are not gcc - if they have |
92 | * we try to detect these and simply assume they are not gcc - if they have |
93 | * an issue with that they should have done it right in the first place. |
93 | * an issue with that they should have done it right in the first place. |
94 | */ |
94 | */ |
95 | #ifndef ECB_GCC_VERSION |
|
|
96 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
95 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
97 | #define ECB_GCC_VERSION(major,minor) 0 |
96 | #define ECB_GCC_VERSION(major,minor) 0 |
98 | #else |
97 | #else |
99 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
98 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
100 | #endif |
99 | #endif |
|
|
100 | |
|
|
101 | #define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor))) |
|
|
102 | |
|
|
103 | #if __clang__ && defined __has_builtin |
|
|
104 | #define ECB_CLANG_BUILTIN(x) __has_builtin (x) |
|
|
105 | #else |
|
|
106 | #define ECB_CLANG_BUILTIN(x) 0 |
|
|
107 | #endif |
|
|
108 | |
|
|
109 | #if __clang__ && defined __has_extension |
|
|
110 | #define ECB_CLANG_EXTENSION(x) __has_extension (x) |
|
|
111 | #else |
|
|
112 | #define ECB_CLANG_EXTENSION(x) 0 |
101 | #endif |
113 | #endif |
102 | |
114 | |
103 | #define ECB_CPP (__cplusplus+0) |
115 | #define ECB_CPP (__cplusplus+0) |
104 | #define ECB_CPP11 (__cplusplus >= 201103L) |
116 | #define ECB_CPP11 (__cplusplus >= 201103L) |
105 | |
117 | |
… | |
… | |
189 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
201 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
190 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
202 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
191 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
203 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
192 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
204 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
193 | |
205 | |
194 | /* The __has_feature syntax from clang is so misdesigned that we cannot use it |
206 | #elif ECB_CLANG_EXTENSION(c_atomic) |
195 | * without risking compile time errors with other compilers. We *could* |
|
|
196 | * define our own ecb_clang_has_feature, but I just can't be bothered to work |
|
|
197 | * around this shit time and again. |
|
|
198 | * #elif defined __clang && __has_feature (cxx_atomic) |
|
|
199 | * // see comment below (stdatomic.h) about the C11 memory model. |
207 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
200 | * #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
208 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
201 | * #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
209 | #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
202 | * #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
210 | #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
203 | */ |
|
|
204 | |
211 | |
205 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
212 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
206 | #define ECB_MEMORY_FENCE __sync_synchronize () |
213 | #define ECB_MEMORY_FENCE __sync_synchronize () |
207 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
214 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
208 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
215 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
… | |
… | |
271 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
278 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
272 | #endif |
279 | #endif |
273 | |
280 | |
274 | /*****************************************************************************/ |
281 | /*****************************************************************************/ |
275 | |
282 | |
276 | #if __cplusplus |
283 | #if ECB_CPP |
277 | #define ecb_inline static inline |
284 | #define ecb_inline static inline |
278 | #elif ECB_GCC_VERSION(2,5) |
285 | #elif ECB_GCC_VERSION(2,5) |
279 | #define ecb_inline static __inline__ |
286 | #define ecb_inline static __inline__ |
280 | #elif ECB_C99 |
287 | #elif ECB_C99 |
281 | #define ecb_inline static inline |
288 | #define ecb_inline static inline |
… | |
… | |
298 | #define ECB_STRINGIFY_(a) # a |
305 | #define ECB_STRINGIFY_(a) # a |
299 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
306 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
300 | |
307 | |
301 | #define ecb_function_ ecb_inline |
308 | #define ecb_function_ ecb_inline |
302 | |
309 | |
303 | #if ECB_GCC_VERSION(3,1) |
310 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
304 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
311 | #define ecb_attribute(attrlist) __attribute__ (attrlist) |
|
|
312 | #else |
|
|
313 | #define ecb_attribute(attrlist) |
|
|
314 | #endif |
|
|
315 | |
|
|
316 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p) |
305 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
317 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
306 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
307 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
|
|
308 | #else |
318 | #else |
309 | #define ecb_attribute(attrlist) |
|
|
310 | |
|
|
311 | /* possible C11 impl for integral types |
319 | /* possible C11 impl for integral types |
312 | typedef struct ecb_is_constant_struct ecb_is_constant_struct; |
320 | typedef struct ecb_is_constant_struct ecb_is_constant_struct; |
313 | #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */ |
321 | #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */ |
314 | |
322 | |
315 | #define ecb_is_constant(expr) 0 |
323 | #define ecb_is_constant(expr) 0 |
|
|
324 | #endif |
|
|
325 | |
|
|
326 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_expect) |
|
|
327 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
328 | #else |
316 | #define ecb_expect(expr,value) (expr) |
329 | #define ecb_expect(expr,value) (expr) |
|
|
330 | #endif |
|
|
331 | |
|
|
332 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_prefetch) |
|
|
333 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
|
|
334 | #else |
317 | #define ecb_prefetch(addr,rw,locality) |
335 | #define ecb_prefetch(addr,rw,locality) |
318 | #endif |
336 | #endif |
319 | |
337 | |
320 | /* no emulation for ecb_decltype */ |
338 | /* no emulation for ecb_decltype */ |
321 | #if ECB_GCC_VERSION(4,5) |
339 | #if ECB_CPP11 |
|
|
340 | // older implementations might have problems with decltype(x)::type, work around it |
|
|
341 | template<class T> struct ecb_decltype_t { typedef T type; }; |
322 | #define ecb_decltype(x) __decltype(x) |
342 | #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type |
323 | #elif ECB_GCC_VERSION(3,0) |
343 | #elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8) |
324 | #define ecb_decltype(x) __typeof(x) |
344 | #define ecb_decltype(x) __typeof__ (x) |
325 | #endif |
345 | #endif |
326 | |
346 | |
327 | #if _MSC_VER >= 1300 |
347 | #if _MSC_VER >= 1300 |
328 | #define ecb_deprecated __declspec(deprecated) |
348 | #define ecb_deprecated __declspec (deprecated) |
329 | #else |
349 | #else |
330 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
350 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
331 | #endif |
351 | #endif |
332 | |
352 | |
333 | #define ecb_noinline ecb_attribute ((__noinline__)) |
353 | #define ecb_noinline ecb_attribute ((__noinline__)) |
334 | #define ecb_unused ecb_attribute ((__unused__)) |
354 | #define ecb_unused ecb_attribute ((__unused__)) |
335 | #define ecb_const ecb_attribute ((__const__)) |
355 | #define ecb_const ecb_attribute ((__const__)) |
336 | #define ecb_pure ecb_attribute ((__pure__)) |
356 | #define ecb_pure ecb_attribute ((__pure__)) |
337 | |
357 | |
338 | /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */ |
358 | /* TODO http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */ |
339 | #if ECB_C11 |
359 | #if ECB_C11 || __IBMC_NORETURN |
|
|
360 | /* http://pic.dhe.ibm.com/infocenter/compbg/v121v141/topic/com.ibm.xlcpp121.bg.doc/language_ref/noreturn.html */ |
340 | #define ecb_noreturn _Noreturn |
361 | #define ecb_noreturn _Noreturn |
|
|
362 | #elif ECB_CPP11 |
|
|
363 | #define ecb_noreturn [[noreturn]] |
|
|
364 | #elif _MSC_VER >= 1200 |
|
|
365 | #define ecb_noreturn __declspec (noreturn) |
341 | #else |
366 | #else |
342 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
367 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
343 | #endif |
368 | #endif |
344 | |
369 | |
345 | #if ECB_GCC_VERSION(4,3) |
370 | #if ECB_GCC_VERSION(4,3) |
… | |
… | |
360 | /* for compatibility to the rest of the world */ |
385 | /* for compatibility to the rest of the world */ |
361 | #define ecb_likely(expr) ecb_expect_true (expr) |
386 | #define ecb_likely(expr) ecb_expect_true (expr) |
362 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
387 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
363 | |
388 | |
364 | /* count trailing zero bits and count # of one bits */ |
389 | /* count trailing zero bits and count # of one bits */ |
365 | #if ECB_GCC_VERSION(3,4) |
390 | #if ECB_GCC_VERSION(3,4) \ |
|
|
391 | || (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \ |
|
|
392 | && ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \ |
|
|
393 | && ECB_CLANG_BUILTIN(__builtin_popcount)) |
366 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
394 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
367 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
395 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
368 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
396 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
369 | #define ecb_ctz32(x) __builtin_ctz (x) |
397 | #define ecb_ctz32(x) __builtin_ctz (x) |
370 | #define ecb_ctz64(x) __builtin_ctzll (x) |
398 | #define ecb_ctz64(x) __builtin_ctzll (x) |
371 | #define ecb_popcount32(x) __builtin_popcount (x) |
399 | #define ecb_popcount32(x) __builtin_popcount (x) |
372 | /* no popcountll */ |
400 | /* no popcountll */ |
373 | #else |
401 | #else |
374 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
402 | ecb_function_ ecb_const int ecb_ctz32 (uint32_t x); |
375 | ecb_function_ int |
403 | ecb_function_ ecb_const int |
376 | ecb_ctz32 (uint32_t x) |
404 | ecb_ctz32 (uint32_t x) |
377 | { |
405 | { |
378 | int r = 0; |
406 | int r = 0; |
379 | |
407 | |
380 | x &= ~x + 1; /* this isolates the lowest bit */ |
408 | x &= ~x + 1; /* this isolates the lowest bit */ |
… | |
… | |
394 | #endif |
422 | #endif |
395 | |
423 | |
396 | return r; |
424 | return r; |
397 | } |
425 | } |
398 | |
426 | |
399 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
427 | ecb_function_ ecb_const int ecb_ctz64 (uint64_t x); |
400 | ecb_function_ int |
428 | ecb_function_ ecb_const int |
401 | ecb_ctz64 (uint64_t x) |
429 | ecb_ctz64 (uint64_t x) |
402 | { |
430 | { |
403 | int shift = x & 0xffffffffU ? 0 : 32; |
431 | int shift = x & 0xffffffffU ? 0 : 32; |
404 | return ecb_ctz32 (x >> shift) + shift; |
432 | return ecb_ctz32 (x >> shift) + shift; |
405 | } |
433 | } |
406 | |
434 | |
407 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
435 | ecb_function_ ecb_const int ecb_popcount32 (uint32_t x); |
408 | ecb_function_ int |
436 | ecb_function_ ecb_const int |
409 | ecb_popcount32 (uint32_t x) |
437 | ecb_popcount32 (uint32_t x) |
410 | { |
438 | { |
411 | x -= (x >> 1) & 0x55555555; |
439 | x -= (x >> 1) & 0x55555555; |
412 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
440 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
413 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
441 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
414 | x *= 0x01010101; |
442 | x *= 0x01010101; |
415 | |
443 | |
416 | return x >> 24; |
444 | return x >> 24; |
417 | } |
445 | } |
418 | |
446 | |
419 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
447 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x); |
420 | ecb_function_ int ecb_ld32 (uint32_t x) |
448 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x) |
421 | { |
449 | { |
422 | int r = 0; |
450 | int r = 0; |
423 | |
451 | |
424 | if (x >> 16) { x >>= 16; r += 16; } |
452 | if (x >> 16) { x >>= 16; r += 16; } |
425 | if (x >> 8) { x >>= 8; r += 8; } |
453 | if (x >> 8) { x >>= 8; r += 8; } |
… | |
… | |
428 | if (x >> 1) { r += 1; } |
456 | if (x >> 1) { r += 1; } |
429 | |
457 | |
430 | return r; |
458 | return r; |
431 | } |
459 | } |
432 | |
460 | |
433 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
461 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x); |
434 | ecb_function_ int ecb_ld64 (uint64_t x) |
462 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x) |
435 | { |
463 | { |
436 | int r = 0; |
464 | int r = 0; |
437 | |
465 | |
438 | if (x >> 32) { x >>= 32; r += 32; } |
466 | if (x >> 32) { x >>= 32; r += 32; } |
439 | |
467 | |
440 | return r + ecb_ld32 (x); |
468 | return r + ecb_ld32 (x); |
441 | } |
469 | } |
442 | #endif |
470 | #endif |
443 | |
471 | |
444 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
472 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x); |
445 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
473 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
446 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
474 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x); |
447 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
475 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
448 | |
476 | |
449 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
477 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x); |
450 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
478 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x) |
451 | { |
479 | { |
452 | return ( (x * 0x0802U & 0x22110U) |
480 | return ( (x * 0x0802U & 0x22110U) |
453 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
481 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
454 | } |
482 | } |
455 | |
483 | |
456 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
484 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x); |
457 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
485 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x) |
458 | { |
486 | { |
459 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
487 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
460 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
488 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
461 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
489 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
462 | x = ( x >> 8 ) | ( x << 8); |
490 | x = ( x >> 8 ) | ( x << 8); |
463 | |
491 | |
464 | return x; |
492 | return x; |
465 | } |
493 | } |
466 | |
494 | |
467 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
495 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x); |
468 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
496 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x) |
469 | { |
497 | { |
470 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
498 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
471 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
499 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
472 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
500 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
473 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
501 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
… | |
… | |
476 | return x; |
504 | return x; |
477 | } |
505 | } |
478 | |
506 | |
479 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
507 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
480 | /* so for this version we are lazy */ |
508 | /* so for this version we are lazy */ |
481 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
509 | ecb_function_ ecb_const int ecb_popcount64 (uint64_t x); |
482 | ecb_function_ int |
510 | ecb_function_ ecb_const int |
483 | ecb_popcount64 (uint64_t x) |
511 | ecb_popcount64 (uint64_t x) |
484 | { |
512 | { |
485 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
513 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
486 | } |
514 | } |
487 | |
515 | |
488 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
516 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count); |
489 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
517 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count); |
490 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
518 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count); |
491 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
519 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count); |
492 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
520 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count); |
493 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
521 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count); |
494 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
522 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count); |
495 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
523 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count); |
496 | |
524 | |
497 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
525 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
498 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
526 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
499 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
527 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
500 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
528 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
501 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
529 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
502 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
530 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
503 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
531 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
504 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
532 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
505 | |
533 | |
506 | #if ECB_GCC_VERSION(4,3) |
534 | #if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64)) |
507 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
535 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
508 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
536 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
509 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
537 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
510 | #else |
538 | #else |
511 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
539 | ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x); |
512 | ecb_function_ uint16_t |
540 | ecb_function_ ecb_const uint16_t |
513 | ecb_bswap16 (uint16_t x) |
541 | ecb_bswap16 (uint16_t x) |
514 | { |
542 | { |
515 | return ecb_rotl16 (x, 8); |
543 | return ecb_rotl16 (x, 8); |
516 | } |
544 | } |
517 | |
545 | |
518 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
546 | ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x); |
519 | ecb_function_ uint32_t |
547 | ecb_function_ ecb_const uint32_t |
520 | ecb_bswap32 (uint32_t x) |
548 | ecb_bswap32 (uint32_t x) |
521 | { |
549 | { |
522 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
550 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
523 | } |
551 | } |
524 | |
552 | |
525 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
553 | ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x); |
526 | ecb_function_ uint64_t |
554 | ecb_function_ ecb_const uint64_t |
527 | ecb_bswap64 (uint64_t x) |
555 | ecb_bswap64 (uint64_t x) |
528 | { |
556 | { |
529 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
557 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
530 | } |
558 | } |
531 | #endif |
559 | #endif |
532 | |
560 | |
533 | #if ECB_GCC_VERSION(4,5) |
561 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
534 | #define ecb_unreachable() __builtin_unreachable () |
562 | #define ecb_unreachable() __builtin_unreachable () |
535 | #else |
563 | #else |
536 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
564 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
537 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
565 | ecb_inline ecb_noreturn void ecb_unreachable (void); |
538 | ecb_inline void ecb_unreachable (void) { } |
566 | ecb_inline ecb_noreturn void ecb_unreachable (void) { } |
539 | #endif |
567 | #endif |
540 | |
568 | |
541 | /* try to tell the compiler that some condition is definitely true */ |
569 | /* try to tell the compiler that some condition is definitely true */ |
542 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
570 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
543 | |
571 | |
544 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
572 | ecb_inline ecb_const unsigned char ecb_byteorder_helper (void); |
545 | ecb_inline unsigned char |
573 | ecb_inline ecb_const unsigned char |
546 | ecb_byteorder_helper (void) |
574 | ecb_byteorder_helper (void) |
547 | { |
575 | { |
548 | /* the union code still generates code under pressure in gcc, */ |
576 | /* the union code still generates code under pressure in gcc, */ |
549 | /* but less than using pointers, and always seems to */ |
577 | /* but less than using pointers, and always seems to */ |
550 | /* successfully return a constant. */ |
578 | /* successfully return a constant. */ |
… | |
… | |
565 | } u = { 0x11223344 }; |
593 | } u = { 0x11223344 }; |
566 | return u.c; |
594 | return u.c; |
567 | #endif |
595 | #endif |
568 | } |
596 | } |
569 | |
597 | |
570 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
598 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
571 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
599 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
572 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
600 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
573 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
601 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
574 | |
602 | |
575 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
603 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
576 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
604 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
577 | #else |
605 | #else |
578 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
606 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
579 | #endif |
607 | #endif |
580 | |
608 | |
581 | #if __cplusplus |
609 | #if ECB_CPP |
582 | template<typename T> |
610 | template<typename T> |
583 | static inline T ecb_div_rd (T val, T div) |
611 | static inline T ecb_div_rd (T val, T div) |
584 | { |
612 | { |
585 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
613 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
586 | } |
614 | } |
… | |
… | |
646 | #define ECB_NAN NAN |
674 | #define ECB_NAN NAN |
647 | #else |
675 | #else |
648 | #define ECB_NAN ECB_INFINITY |
676 | #define ECB_NAN ECB_INFINITY |
649 | #endif |
677 | #endif |
650 | |
678 | |
|
|
679 | #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L |
|
|
680 | #define ecb_ldexpf(x,e) ldexpf ((x), (e)) |
|
|
681 | #else |
|
|
682 | #define ecb_ldexpf(x,e) (float) ldexp ((x), (e)) |
|
|
683 | #endif |
|
|
684 | |
651 | /* converts an ieee half/binary16 to a float */ |
685 | /* converts an ieee half/binary16 to a float */ |
652 | ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; |
686 | ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x); |
653 | ecb_function_ float |
687 | ecb_function_ ecb_const float |
654 | ecb_binary16_to_float (uint16_t x) |
688 | ecb_binary16_to_float (uint16_t x) |
655 | { |
689 | { |
656 | int e = (x >> 10) & 0x1f; |
690 | int e = (x >> 10) & 0x1f; |
657 | int m = x & 0x3ff; |
691 | int m = x & 0x3ff; |
658 | float r; |
692 | float r; |
659 | |
693 | |
660 | if (!e ) r = ldexpf (m , -24); |
694 | if (!e ) r = ecb_ldexpf (m , -24); |
661 | else if (e != 31) r = ldexpf (m + 0x400, e - 25); |
695 | else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25); |
662 | else if (m ) r = ECB_NAN; |
696 | else if (m ) r = ECB_NAN; |
663 | else r = ECB_INFINITY; |
697 | else r = ECB_INFINITY; |
664 | |
698 | |
665 | return x & 0x8000 ? -r : r; |
699 | return x & 0x8000 ? -r : r; |
666 | } |
700 | } |
667 | |
701 | |
668 | /* convert a float to ieee single/binary32 */ |
702 | /* convert a float to ieee single/binary32 */ |
669 | ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
703 | ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x); |
670 | ecb_function_ uint32_t |
704 | ecb_function_ ecb_const uint32_t |
671 | ecb_float_to_binary32 (float x) |
705 | ecb_float_to_binary32 (float x) |
672 | { |
706 | { |
673 | uint32_t r; |
707 | uint32_t r; |
674 | |
708 | |
675 | #if ECB_STDFP |
709 | #if ECB_STDFP |
… | |
… | |
704 | |
738 | |
705 | return r; |
739 | return r; |
706 | } |
740 | } |
707 | |
741 | |
708 | /* converts an ieee single/binary32 to a float */ |
742 | /* converts an ieee single/binary32 to a float */ |
709 | ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
743 | ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x); |
710 | ecb_function_ float |
744 | ecb_function_ ecb_const float |
711 | ecb_binary32_to_float (uint32_t x) |
745 | ecb_binary32_to_float (uint32_t x) |
712 | { |
746 | { |
713 | float r; |
747 | float r; |
714 | |
748 | |
715 | #if ECB_STDFP |
749 | #if ECB_STDFP |
… | |
… | |
725 | x |= 0x800000U; |
759 | x |= 0x800000U; |
726 | else |
760 | else |
727 | e = 1; |
761 | e = 1; |
728 | |
762 | |
729 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
763 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
730 | r = ldexpf (x * (0.5f / 0x800000U), e - 126); |
764 | r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126); |
731 | |
765 | |
732 | r = neg ? -r : r; |
766 | r = neg ? -r : r; |
733 | #endif |
767 | #endif |
734 | |
768 | |
735 | return r; |
769 | return r; |
736 | } |
770 | } |
737 | |
771 | |
738 | /* convert a double to ieee double/binary64 */ |
772 | /* convert a double to ieee double/binary64 */ |
739 | ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
773 | ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x); |
740 | ecb_function_ uint64_t |
774 | ecb_function_ ecb_const uint64_t |
741 | ecb_double_to_binary64 (double x) |
775 | ecb_double_to_binary64 (double x) |
742 | { |
776 | { |
743 | uint64_t r; |
777 | uint64_t r; |
744 | |
778 | |
745 | #if ECB_STDFP |
779 | #if ECB_STDFP |
… | |
… | |
774 | |
808 | |
775 | return r; |
809 | return r; |
776 | } |
810 | } |
777 | |
811 | |
778 | /* converts an ieee double/binary64 to a double */ |
812 | /* converts an ieee double/binary64 to a double */ |
779 | ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; |
813 | ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x); |
780 | ecb_function_ double |
814 | ecb_function_ ecb_const double |
781 | ecb_binary64_to_double (uint64_t x) |
815 | ecb_binary64_to_double (uint64_t x) |
782 | { |
816 | { |
783 | double r; |
817 | double r; |
784 | |
818 | |
785 | #if ECB_STDFP |
819 | #if ECB_STDFP |