… | |
… | |
491 | /* the following is ecb.h embedded into libev - use update_ev_c to update from an external copy */ |
491 | /* the following is ecb.h embedded into libev - use update_ev_c to update from an external copy */ |
492 | /* ECB.H BEGIN */ |
492 | /* ECB.H BEGIN */ |
493 | /* |
493 | /* |
494 | * libecb - http://software.schmorp.de/pkg/libecb |
494 | * libecb - http://software.schmorp.de/pkg/libecb |
495 | * |
495 | * |
496 | * Copyright (©) 2009-2014 Marc Alexander Lehmann <libecb@schmorp.de> |
496 | * Copyright (©) 2009-2015 Marc Alexander Lehmann <libecb@schmorp.de> |
497 | * Copyright (©) 2011 Emanuele Giaquinta |
497 | * Copyright (©) 2011 Emanuele Giaquinta |
498 | * All rights reserved. |
498 | * All rights reserved. |
499 | * |
499 | * |
500 | * Redistribution and use in source and binary forms, with or without modifica- |
500 | * Redistribution and use in source and binary forms, with or without modifica- |
501 | * tion, are permitted provided that the following conditions are met: |
501 | * tion, are permitted provided that the following conditions are met: |
… | |
… | |
532 | |
532 | |
533 | #ifndef ECB_H |
533 | #ifndef ECB_H |
534 | #define ECB_H |
534 | #define ECB_H |
535 | |
535 | |
536 | /* 16 bits major, 16 bits minor */ |
536 | /* 16 bits major, 16 bits minor */ |
537 | #define ECB_VERSION 0x00010003 |
537 | #define ECB_VERSION 0x00010004 |
538 | |
538 | |
539 | #ifdef _WIN32 |
539 | #ifdef _WIN32 |
540 | typedef signed char int8_t; |
540 | typedef signed char int8_t; |
541 | typedef unsigned char uint8_t; |
541 | typedef unsigned char uint8_t; |
542 | typedef signed short int16_t; |
542 | typedef signed short int16_t; |
… | |
… | |
582 | * causing enormous grief in return for some better fake benchmark numbers. |
582 | * causing enormous grief in return for some better fake benchmark numbers. |
583 | * or so. |
583 | * or so. |
584 | * we try to detect these and simply assume they are not gcc - if they have |
584 | * we try to detect these and simply assume they are not gcc - if they have |
585 | * an issue with that they should have done it right in the first place. |
585 | * an issue with that they should have done it right in the first place. |
586 | */ |
586 | */ |
587 | #ifndef ECB_GCC_VERSION |
|
|
588 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
587 | #if !defined __GNUC_MINOR__ || defined __INTEL_COMPILER || defined __SUNPRO_C || defined __SUNPRO_CC || defined __llvm__ || defined __clang__ |
589 | #define ECB_GCC_VERSION(major,minor) 0 |
588 | #define ECB_GCC_VERSION(major,minor) 0 |
590 | #else |
589 | #else |
591 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
590 | #define ECB_GCC_VERSION(major,minor) (__GNUC__ > (major) || (__GNUC__ == (major) && __GNUC_MINOR__ >= (minor))) |
592 | #endif |
591 | #endif |
|
|
592 | |
|
|
593 | #define ECB_CLANG_VERSION(major,minor) (__clang_major__ > (major) || (__clang_major__ == (major) && __clang_minor__ >= (minor))) |
|
|
594 | |
|
|
595 | #if __clang__ && defined __has_builtin |
|
|
596 | #define ECB_CLANG_BUILTIN(x) __has_builtin (x) |
|
|
597 | #else |
|
|
598 | #define ECB_CLANG_BUILTIN(x) 0 |
|
|
599 | #endif |
|
|
600 | |
|
|
601 | #if __clang__ && defined __has_extension |
|
|
602 | #define ECB_CLANG_EXTENSION(x) __has_extension (x) |
|
|
603 | #else |
|
|
604 | #define ECB_CLANG_EXTENSION(x) 0 |
593 | #endif |
605 | #endif |
594 | |
606 | |
595 | #define ECB_CPP (__cplusplus+0) |
607 | #define ECB_CPP (__cplusplus+0) |
596 | #define ECB_CPP11 (__cplusplus >= 201103L) |
608 | #define ECB_CPP11 (__cplusplus >= 201103L) |
597 | |
609 | |
… | |
… | |
681 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
693 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
682 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
694 | #define ECB_MEMORY_FENCE __atomic_thread_fence (__ATOMIC_SEQ_CST) |
683 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
695 | #define ECB_MEMORY_FENCE_ACQUIRE __atomic_thread_fence (__ATOMIC_ACQUIRE) |
684 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
696 | #define ECB_MEMORY_FENCE_RELEASE __atomic_thread_fence (__ATOMIC_RELEASE) |
685 | |
697 | |
686 | /* The __has_feature syntax from clang is so misdesigned that we cannot use it |
698 | #elif ECB_CLANG_EXTENSION(c_atomic) |
687 | * without risking compile time errors with other compilers. We *could* |
|
|
688 | * define our own ecb_clang_has_feature, but I just can't be bothered to work |
|
|
689 | * around this shit time and again. |
|
|
690 | * #elif defined __clang && __has_feature (cxx_atomic) |
|
|
691 | * // see comment below (stdatomic.h) about the C11 memory model. |
699 | /* see comment below (stdatomic.h) about the C11 memory model. */ |
692 | * #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
700 | #define ECB_MEMORY_FENCE __c11_atomic_thread_fence (__ATOMIC_SEQ_CST) |
693 | * #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
701 | #define ECB_MEMORY_FENCE_ACQUIRE __c11_atomic_thread_fence (__ATOMIC_ACQUIRE) |
694 | * #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
702 | #define ECB_MEMORY_FENCE_RELEASE __c11_atomic_thread_fence (__ATOMIC_RELEASE) |
695 | */ |
|
|
696 | |
703 | |
697 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
704 | #elif ECB_GCC_VERSION(4,4) || defined __INTEL_COMPILER || defined __clang__ |
698 | #define ECB_MEMORY_FENCE __sync_synchronize () |
705 | #define ECB_MEMORY_FENCE __sync_synchronize () |
699 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
706 | #elif _MSC_VER >= 1500 /* VC++ 2008 */ |
700 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
707 | /* apparently, microsoft broke all the memory barrier stuff in Visual Studio 2008... */ |
… | |
… | |
763 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
770 | #define ECB_MEMORY_FENCE_RELEASE ECB_MEMORY_FENCE |
764 | #endif |
771 | #endif |
765 | |
772 | |
766 | /*****************************************************************************/ |
773 | /*****************************************************************************/ |
767 | |
774 | |
768 | #if __cplusplus |
775 | #if ECB_CPP |
769 | #define ecb_inline static inline |
776 | #define ecb_inline static inline |
770 | #elif ECB_GCC_VERSION(2,5) |
777 | #elif ECB_GCC_VERSION(2,5) |
771 | #define ecb_inline static __inline__ |
778 | #define ecb_inline static __inline__ |
772 | #elif ECB_C99 |
779 | #elif ECB_C99 |
773 | #define ecb_inline static inline |
780 | #define ecb_inline static inline |
… | |
… | |
790 | #define ECB_STRINGIFY_(a) # a |
797 | #define ECB_STRINGIFY_(a) # a |
791 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
798 | #define ECB_STRINGIFY(a) ECB_STRINGIFY_(a) |
792 | |
799 | |
793 | #define ecb_function_ ecb_inline |
800 | #define ecb_function_ ecb_inline |
794 | |
801 | |
795 | #if ECB_GCC_VERSION(3,1) |
802 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_VERSION(2,8) |
796 | #define ecb_attribute(attrlist) __attribute__(attrlist) |
803 | #define ecb_attribute(attrlist) __attribute__ (attrlist) |
797 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
|
|
798 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
799 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
|
|
800 | #else |
804 | #else |
801 | #define ecb_attribute(attrlist) |
805 | #define ecb_attribute(attrlist) |
|
|
806 | #endif |
802 | |
807 | |
|
|
808 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_constant_p) |
|
|
809 | #define ecb_is_constant(expr) __builtin_constant_p (expr) |
|
|
810 | #else |
803 | /* possible C11 impl for integral types |
811 | /* possible C11 impl for integral types |
804 | typedef struct ecb_is_constant_struct ecb_is_constant_struct; |
812 | typedef struct ecb_is_constant_struct ecb_is_constant_struct; |
805 | #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */ |
813 | #define ecb_is_constant(expr) _Generic ((1 ? (struct ecb_is_constant_struct *)0 : (void *)((expr) - (expr)), ecb_is_constant_struct *: 0, default: 1)) */ |
806 | |
814 | |
807 | #define ecb_is_constant(expr) 0 |
815 | #define ecb_is_constant(expr) 0 |
|
|
816 | #endif |
|
|
817 | |
|
|
818 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_expect) |
|
|
819 | #define ecb_expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
820 | #else |
808 | #define ecb_expect(expr,value) (expr) |
821 | #define ecb_expect(expr,value) (expr) |
|
|
822 | #endif |
|
|
823 | |
|
|
824 | #if ECB_GCC_VERSION(3,1) || ECB_CLANG_BUILTIN(__builtin_prefetch) |
|
|
825 | #define ecb_prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality) |
|
|
826 | #else |
809 | #define ecb_prefetch(addr,rw,locality) |
827 | #define ecb_prefetch(addr,rw,locality) |
810 | #endif |
828 | #endif |
811 | |
829 | |
812 | /* no emulation for ecb_decltype */ |
830 | /* no emulation for ecb_decltype */ |
813 | #if ECB_GCC_VERSION(4,5) |
831 | #if ECB_CPP11 |
|
|
832 | // older implementations might have problems with decltype(x)::type, work around it |
|
|
833 | template<class T> struct ecb_decltype_t { typedef T type; }; |
814 | #define ecb_decltype(x) __decltype(x) |
834 | #define ecb_decltype(x) ecb_decltype_t<decltype (x)>::type |
815 | #elif ECB_GCC_VERSION(3,0) |
835 | #elif ECB_GCC_VERSION(3,0) || ECB_CLANG_VERSION(2,8) |
816 | #define ecb_decltype(x) __typeof(x) |
836 | #define ecb_decltype(x) __typeof__ (x) |
817 | #endif |
837 | #endif |
818 | |
838 | |
819 | #if _MSC_VER >= 1300 |
839 | #if _MSC_VER >= 1300 |
820 | #define ecb_deprecated __declspec(deprecated) |
840 | #define ecb_deprecated __declspec (deprecated) |
821 | #else |
841 | #else |
822 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
842 | #define ecb_deprecated ecb_attribute ((__deprecated__)) |
823 | #endif |
843 | #endif |
824 | |
844 | |
825 | #define ecb_noinline ecb_attribute ((__noinline__)) |
845 | #define ecb_noinline ecb_attribute ((__noinline__)) |
826 | #define ecb_unused ecb_attribute ((__unused__)) |
846 | #define ecb_unused ecb_attribute ((__unused__)) |
827 | #define ecb_const ecb_attribute ((__const__)) |
847 | #define ecb_const ecb_attribute ((__const__)) |
828 | #define ecb_pure ecb_attribute ((__pure__)) |
848 | #define ecb_pure ecb_attribute ((__pure__)) |
829 | |
849 | |
830 | /* http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */ |
850 | /* TODO http://msdn.microsoft.com/en-us/library/k6ktzx3s.aspx __declspec(noreturn) */ |
831 | #if ECB_C11 |
851 | #if ECB_C11 || __IBMC_NORETURN |
|
|
852 | /* http://pic.dhe.ibm.com/infocenter/compbg/v121v141/topic/com.ibm.xlcpp121.bg.doc/language_ref/noreturn.html */ |
832 | #define ecb_noreturn _Noreturn |
853 | #define ecb_noreturn _Noreturn |
833 | #else |
854 | #else |
834 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
855 | #define ecb_noreturn ecb_attribute ((__noreturn__)) |
835 | #endif |
856 | #endif |
836 | |
857 | |
… | |
… | |
852 | /* for compatibility to the rest of the world */ |
873 | /* for compatibility to the rest of the world */ |
853 | #define ecb_likely(expr) ecb_expect_true (expr) |
874 | #define ecb_likely(expr) ecb_expect_true (expr) |
854 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
875 | #define ecb_unlikely(expr) ecb_expect_false (expr) |
855 | |
876 | |
856 | /* count trailing zero bits and count # of one bits */ |
877 | /* count trailing zero bits and count # of one bits */ |
857 | #if ECB_GCC_VERSION(3,4) |
878 | #if ECB_GCC_VERSION(3,4) \ |
|
|
879 | || (ECB_CLANG_BUILTIN(__builtin_clz) && ECB_CLANG_BUILTIN(__builtin_clzll) \ |
|
|
880 | && ECB_CLANG_BUILTIN(__builtin_ctz) && ECB_CLANG_BUILTIN(__builtin_ctzll) \ |
|
|
881 | && ECB_CLANG_BUILTIN(__builtin_popcount)) |
858 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
882 | /* we assume int == 32 bit, long == 32 or 64 bit and long long == 64 bit */ |
859 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
883 | #define ecb_ld32(x) (__builtin_clz (x) ^ 31) |
860 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
884 | #define ecb_ld64(x) (__builtin_clzll (x) ^ 63) |
861 | #define ecb_ctz32(x) __builtin_ctz (x) |
885 | #define ecb_ctz32(x) __builtin_ctz (x) |
862 | #define ecb_ctz64(x) __builtin_ctzll (x) |
886 | #define ecb_ctz64(x) __builtin_ctzll (x) |
863 | #define ecb_popcount32(x) __builtin_popcount (x) |
887 | #define ecb_popcount32(x) __builtin_popcount (x) |
864 | /* no popcountll */ |
888 | /* no popcountll */ |
865 | #else |
889 | #else |
866 | ecb_function_ int ecb_ctz32 (uint32_t x) ecb_const; |
890 | ecb_function_ ecb_const int ecb_ctz32 (uint32_t x); |
867 | ecb_function_ int |
891 | ecb_function_ ecb_const int |
868 | ecb_ctz32 (uint32_t x) |
892 | ecb_ctz32 (uint32_t x) |
869 | { |
893 | { |
870 | int r = 0; |
894 | int r = 0; |
871 | |
895 | |
872 | x &= ~x + 1; /* this isolates the lowest bit */ |
896 | x &= ~x + 1; /* this isolates the lowest bit */ |
… | |
… | |
886 | #endif |
910 | #endif |
887 | |
911 | |
888 | return r; |
912 | return r; |
889 | } |
913 | } |
890 | |
914 | |
891 | ecb_function_ int ecb_ctz64 (uint64_t x) ecb_const; |
915 | ecb_function_ ecb_const int ecb_ctz64 (uint64_t x); |
892 | ecb_function_ int |
916 | ecb_function_ ecb_const int |
893 | ecb_ctz64 (uint64_t x) |
917 | ecb_ctz64 (uint64_t x) |
894 | { |
918 | { |
895 | int shift = x & 0xffffffffU ? 0 : 32; |
919 | int shift = x & 0xffffffffU ? 0 : 32; |
896 | return ecb_ctz32 (x >> shift) + shift; |
920 | return ecb_ctz32 (x >> shift) + shift; |
897 | } |
921 | } |
898 | |
922 | |
899 | ecb_function_ int ecb_popcount32 (uint32_t x) ecb_const; |
923 | ecb_function_ ecb_const int ecb_popcount32 (uint32_t x); |
900 | ecb_function_ int |
924 | ecb_function_ ecb_const int |
901 | ecb_popcount32 (uint32_t x) |
925 | ecb_popcount32 (uint32_t x) |
902 | { |
926 | { |
903 | x -= (x >> 1) & 0x55555555; |
927 | x -= (x >> 1) & 0x55555555; |
904 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
928 | x = ((x >> 2) & 0x33333333) + (x & 0x33333333); |
905 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
929 | x = ((x >> 4) + x) & 0x0f0f0f0f; |
906 | x *= 0x01010101; |
930 | x *= 0x01010101; |
907 | |
931 | |
908 | return x >> 24; |
932 | return x >> 24; |
909 | } |
933 | } |
910 | |
934 | |
911 | ecb_function_ int ecb_ld32 (uint32_t x) ecb_const; |
935 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x); |
912 | ecb_function_ int ecb_ld32 (uint32_t x) |
936 | ecb_function_ ecb_const int ecb_ld32 (uint32_t x) |
913 | { |
937 | { |
914 | int r = 0; |
938 | int r = 0; |
915 | |
939 | |
916 | if (x >> 16) { x >>= 16; r += 16; } |
940 | if (x >> 16) { x >>= 16; r += 16; } |
917 | if (x >> 8) { x >>= 8; r += 8; } |
941 | if (x >> 8) { x >>= 8; r += 8; } |
… | |
… | |
920 | if (x >> 1) { r += 1; } |
944 | if (x >> 1) { r += 1; } |
921 | |
945 | |
922 | return r; |
946 | return r; |
923 | } |
947 | } |
924 | |
948 | |
925 | ecb_function_ int ecb_ld64 (uint64_t x) ecb_const; |
949 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x); |
926 | ecb_function_ int ecb_ld64 (uint64_t x) |
950 | ecb_function_ ecb_const int ecb_ld64 (uint64_t x) |
927 | { |
951 | { |
928 | int r = 0; |
952 | int r = 0; |
929 | |
953 | |
930 | if (x >> 32) { x >>= 32; r += 32; } |
954 | if (x >> 32) { x >>= 32; r += 32; } |
931 | |
955 | |
932 | return r + ecb_ld32 (x); |
956 | return r + ecb_ld32 (x); |
933 | } |
957 | } |
934 | #endif |
958 | #endif |
935 | |
959 | |
936 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) ecb_const; |
960 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x); |
937 | ecb_function_ ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
961 | ecb_function_ ecb_const ecb_bool ecb_is_pot32 (uint32_t x) { return !(x & (x - 1)); } |
938 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) ecb_const; |
962 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x); |
939 | ecb_function_ ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
963 | ecb_function_ ecb_const ecb_bool ecb_is_pot64 (uint64_t x) { return !(x & (x - 1)); } |
940 | |
964 | |
941 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) ecb_const; |
965 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x); |
942 | ecb_function_ uint8_t ecb_bitrev8 (uint8_t x) |
966 | ecb_function_ ecb_const uint8_t ecb_bitrev8 (uint8_t x) |
943 | { |
967 | { |
944 | return ( (x * 0x0802U & 0x22110U) |
968 | return ( (x * 0x0802U & 0x22110U) |
945 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
969 | | (x * 0x8020U & 0x88440U)) * 0x10101U >> 16; |
946 | } |
970 | } |
947 | |
971 | |
948 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) ecb_const; |
972 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x); |
949 | ecb_function_ uint16_t ecb_bitrev16 (uint16_t x) |
973 | ecb_function_ ecb_const uint16_t ecb_bitrev16 (uint16_t x) |
950 | { |
974 | { |
951 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
975 | x = ((x >> 1) & 0x5555) | ((x & 0x5555) << 1); |
952 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
976 | x = ((x >> 2) & 0x3333) | ((x & 0x3333) << 2); |
953 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
977 | x = ((x >> 4) & 0x0f0f) | ((x & 0x0f0f) << 4); |
954 | x = ( x >> 8 ) | ( x << 8); |
978 | x = ( x >> 8 ) | ( x << 8); |
955 | |
979 | |
956 | return x; |
980 | return x; |
957 | } |
981 | } |
958 | |
982 | |
959 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) ecb_const; |
983 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x); |
960 | ecb_function_ uint32_t ecb_bitrev32 (uint32_t x) |
984 | ecb_function_ ecb_const uint32_t ecb_bitrev32 (uint32_t x) |
961 | { |
985 | { |
962 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
986 | x = ((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1); |
963 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
987 | x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2); |
964 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
988 | x = ((x >> 4) & 0x0f0f0f0f) | ((x & 0x0f0f0f0f) << 4); |
965 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
989 | x = ((x >> 8) & 0x00ff00ff) | ((x & 0x00ff00ff) << 8); |
… | |
… | |
968 | return x; |
992 | return x; |
969 | } |
993 | } |
970 | |
994 | |
971 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
995 | /* popcount64 is only available on 64 bit cpus as gcc builtin */ |
972 | /* so for this version we are lazy */ |
996 | /* so for this version we are lazy */ |
973 | ecb_function_ int ecb_popcount64 (uint64_t x) ecb_const; |
997 | ecb_function_ ecb_const int ecb_popcount64 (uint64_t x); |
974 | ecb_function_ int |
998 | ecb_function_ ecb_const int |
975 | ecb_popcount64 (uint64_t x) |
999 | ecb_popcount64 (uint64_t x) |
976 | { |
1000 | { |
977 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
1001 | return ecb_popcount32 (x) + ecb_popcount32 (x >> 32); |
978 | } |
1002 | } |
979 | |
1003 | |
980 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) ecb_const; |
1004 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count); |
981 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) ecb_const; |
1005 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count); |
982 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) ecb_const; |
1006 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count); |
983 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) ecb_const; |
1007 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count); |
984 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) ecb_const; |
1008 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count); |
985 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) ecb_const; |
1009 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count); |
986 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) ecb_const; |
1010 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count); |
987 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) ecb_const; |
1011 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count); |
988 | |
1012 | |
989 | ecb_inline uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
1013 | ecb_inline ecb_const uint8_t ecb_rotl8 (uint8_t x, unsigned int count) { return (x >> ( 8 - count)) | (x << count); } |
990 | ecb_inline uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
1014 | ecb_inline ecb_const uint8_t ecb_rotr8 (uint8_t x, unsigned int count) { return (x << ( 8 - count)) | (x >> count); } |
991 | ecb_inline uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
1015 | ecb_inline ecb_const uint16_t ecb_rotl16 (uint16_t x, unsigned int count) { return (x >> (16 - count)) | (x << count); } |
992 | ecb_inline uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
1016 | ecb_inline ecb_const uint16_t ecb_rotr16 (uint16_t x, unsigned int count) { return (x << (16 - count)) | (x >> count); } |
993 | ecb_inline uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
1017 | ecb_inline ecb_const uint32_t ecb_rotl32 (uint32_t x, unsigned int count) { return (x >> (32 - count)) | (x << count); } |
994 | ecb_inline uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
1018 | ecb_inline ecb_const uint32_t ecb_rotr32 (uint32_t x, unsigned int count) { return (x << (32 - count)) | (x >> count); } |
995 | ecb_inline uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
1019 | ecb_inline ecb_const uint64_t ecb_rotl64 (uint64_t x, unsigned int count) { return (x >> (64 - count)) | (x << count); } |
996 | ecb_inline uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
1020 | ecb_inline ecb_const uint64_t ecb_rotr64 (uint64_t x, unsigned int count) { return (x << (64 - count)) | (x >> count); } |
997 | |
1021 | |
998 | #if ECB_GCC_VERSION(4,3) |
1022 | #if ECB_GCC_VERSION(4,3) || (ECB_CLANG_BUILTIN(__builtin_bswap32) && ECB_CLANG_BUILTIN(__builtin_bswap64)) |
999 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
1023 | #define ecb_bswap16(x) (__builtin_bswap32 (x) >> 16) |
1000 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
1024 | #define ecb_bswap32(x) __builtin_bswap32 (x) |
1001 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
1025 | #define ecb_bswap64(x) __builtin_bswap64 (x) |
1002 | #else |
1026 | #else |
1003 | ecb_function_ uint16_t ecb_bswap16 (uint16_t x) ecb_const; |
1027 | ecb_function_ ecb_const uint16_t ecb_bswap16 (uint16_t x); |
1004 | ecb_function_ uint16_t |
1028 | ecb_function_ ecb_const uint16_t |
1005 | ecb_bswap16 (uint16_t x) |
1029 | ecb_bswap16 (uint16_t x) |
1006 | { |
1030 | { |
1007 | return ecb_rotl16 (x, 8); |
1031 | return ecb_rotl16 (x, 8); |
1008 | } |
1032 | } |
1009 | |
1033 | |
1010 | ecb_function_ uint32_t ecb_bswap32 (uint32_t x) ecb_const; |
1034 | ecb_function_ ecb_const uint32_t ecb_bswap32 (uint32_t x); |
1011 | ecb_function_ uint32_t |
1035 | ecb_function_ ecb_const uint32_t |
1012 | ecb_bswap32 (uint32_t x) |
1036 | ecb_bswap32 (uint32_t x) |
1013 | { |
1037 | { |
1014 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
1038 | return (((uint32_t)ecb_bswap16 (x)) << 16) | ecb_bswap16 (x >> 16); |
1015 | } |
1039 | } |
1016 | |
1040 | |
1017 | ecb_function_ uint64_t ecb_bswap64 (uint64_t x) ecb_const; |
1041 | ecb_function_ ecb_const uint64_t ecb_bswap64 (uint64_t x); |
1018 | ecb_function_ uint64_t |
1042 | ecb_function_ ecb_const uint64_t |
1019 | ecb_bswap64 (uint64_t x) |
1043 | ecb_bswap64 (uint64_t x) |
1020 | { |
1044 | { |
1021 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
1045 | return (((uint64_t)ecb_bswap32 (x)) << 32) | ecb_bswap32 (x >> 32); |
1022 | } |
1046 | } |
1023 | #endif |
1047 | #endif |
1024 | |
1048 | |
1025 | #if ECB_GCC_VERSION(4,5) |
1049 | #if ECB_GCC_VERSION(4,5) || ECB_CLANG_BUILTIN(__builtin_unreachable) |
1026 | #define ecb_unreachable() __builtin_unreachable () |
1050 | #define ecb_unreachable() __builtin_unreachable () |
1027 | #else |
1051 | #else |
1028 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
1052 | /* this seems to work fine, but gcc always emits a warning for it :/ */ |
1029 | ecb_inline void ecb_unreachable (void) ecb_noreturn; |
1053 | ecb_inline ecb_noreturn void ecb_unreachable (void); |
1030 | ecb_inline void ecb_unreachable (void) { } |
1054 | ecb_inline ecb_noreturn void ecb_unreachable (void) { } |
1031 | #endif |
1055 | #endif |
1032 | |
1056 | |
1033 | /* try to tell the compiler that some condition is definitely true */ |
1057 | /* try to tell the compiler that some condition is definitely true */ |
1034 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
1058 | #define ecb_assume(cond) if (!(cond)) ecb_unreachable (); else 0 |
1035 | |
1059 | |
1036 | ecb_inline unsigned char ecb_byteorder_helper (void) ecb_const; |
1060 | ecb_inline ecb_const unsigned char ecb_byteorder_helper (void); |
1037 | ecb_inline unsigned char |
1061 | ecb_inline ecb_const unsigned char |
1038 | ecb_byteorder_helper (void) |
1062 | ecb_byteorder_helper (void) |
1039 | { |
1063 | { |
1040 | /* the union code still generates code under pressure in gcc, */ |
1064 | /* the union code still generates code under pressure in gcc, */ |
1041 | /* but less than using pointers, and always seems to */ |
1065 | /* but less than using pointers, and always seems to */ |
1042 | /* successfully return a constant. */ |
1066 | /* successfully return a constant. */ |
… | |
… | |
1057 | } u = { 0x11223344 }; |
1081 | } u = { 0x11223344 }; |
1058 | return u.c; |
1082 | return u.c; |
1059 | #endif |
1083 | #endif |
1060 | } |
1084 | } |
1061 | |
1085 | |
1062 | ecb_inline ecb_bool ecb_big_endian (void) ecb_const; |
1086 | ecb_inline ecb_const ecb_bool ecb_big_endian (void); |
1063 | ecb_inline ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
1087 | ecb_inline ecb_const ecb_bool ecb_big_endian (void) { return ecb_byteorder_helper () == 0x11; } |
1064 | ecb_inline ecb_bool ecb_little_endian (void) ecb_const; |
1088 | ecb_inline ecb_const ecb_bool ecb_little_endian (void); |
1065 | ecb_inline ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
1089 | ecb_inline ecb_const ecb_bool ecb_little_endian (void) { return ecb_byteorder_helper () == 0x44; } |
1066 | |
1090 | |
1067 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
1091 | #if ECB_GCC_VERSION(3,0) || ECB_C99 |
1068 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
1092 | #define ecb_mod(m,n) ((m) % (n) + ((m) % (n) < 0 ? (n) : 0)) |
1069 | #else |
1093 | #else |
1070 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
1094 | #define ecb_mod(m,n) ((m) < 0 ? ((n) - 1 - ((-1 - (m)) % (n))) : ((m) % (n))) |
1071 | #endif |
1095 | #endif |
1072 | |
1096 | |
1073 | #if __cplusplus |
1097 | #if ECB_CPP |
1074 | template<typename T> |
1098 | template<typename T> |
1075 | static inline T ecb_div_rd (T val, T div) |
1099 | static inline T ecb_div_rd (T val, T div) |
1076 | { |
1100 | { |
1077 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
1101 | return val < 0 ? - ((-val + div - 1) / div) : (val ) / div; |
1078 | } |
1102 | } |
… | |
… | |
1138 | #define ECB_NAN NAN |
1162 | #define ECB_NAN NAN |
1139 | #else |
1163 | #else |
1140 | #define ECB_NAN ECB_INFINITY |
1164 | #define ECB_NAN ECB_INFINITY |
1141 | #endif |
1165 | #endif |
1142 | |
1166 | |
|
|
1167 | #if ECB_C99 || _XOPEN_VERSION >= 600 || _POSIX_VERSION >= 200112L |
|
|
1168 | #define ecb_ldexpf(x,e) ldexpf ((x), (e)) |
|
|
1169 | #else |
|
|
1170 | #define ecb_ldexpf(x,e) (float) ldexp ((x), (e)) |
|
|
1171 | #endif |
|
|
1172 | |
1143 | /* converts an ieee half/binary16 to a float */ |
1173 | /* converts an ieee half/binary16 to a float */ |
1144 | ecb_function_ float ecb_binary16_to_float (uint16_t x) ecb_const; |
1174 | ecb_function_ ecb_const float ecb_binary16_to_float (uint16_t x); |
1145 | ecb_function_ float |
1175 | ecb_function_ ecb_const float |
1146 | ecb_binary16_to_float (uint16_t x) |
1176 | ecb_binary16_to_float (uint16_t x) |
1147 | { |
1177 | { |
1148 | int e = (x >> 10) & 0x1f; |
1178 | int e = (x >> 10) & 0x1f; |
1149 | int m = x & 0x3ff; |
1179 | int m = x & 0x3ff; |
1150 | float r; |
1180 | float r; |
1151 | |
1181 | |
1152 | if (!e ) r = ldexpf (m , -24); |
1182 | if (!e ) r = ecb_ldexpf (m , -24); |
1153 | else if (e != 31) r = ldexpf (m + 0x400, e - 25); |
1183 | else if (e != 31) r = ecb_ldexpf (m + 0x400, e - 25); |
1154 | else if (m ) r = ECB_NAN; |
1184 | else if (m ) r = ECB_NAN; |
1155 | else r = ECB_INFINITY; |
1185 | else r = ECB_INFINITY; |
1156 | |
1186 | |
1157 | return x & 0x8000 ? -r : r; |
1187 | return x & 0x8000 ? -r : r; |
1158 | } |
1188 | } |
1159 | |
1189 | |
1160 | /* convert a float to ieee single/binary32 */ |
1190 | /* convert a float to ieee single/binary32 */ |
1161 | ecb_function_ uint32_t ecb_float_to_binary32 (float x) ecb_const; |
1191 | ecb_function_ ecb_const uint32_t ecb_float_to_binary32 (float x); |
1162 | ecb_function_ uint32_t |
1192 | ecb_function_ ecb_const uint32_t |
1163 | ecb_float_to_binary32 (float x) |
1193 | ecb_float_to_binary32 (float x) |
1164 | { |
1194 | { |
1165 | uint32_t r; |
1195 | uint32_t r; |
1166 | |
1196 | |
1167 | #if ECB_STDFP |
1197 | #if ECB_STDFP |
… | |
… | |
1196 | |
1226 | |
1197 | return r; |
1227 | return r; |
1198 | } |
1228 | } |
1199 | |
1229 | |
1200 | /* converts an ieee single/binary32 to a float */ |
1230 | /* converts an ieee single/binary32 to a float */ |
1201 | ecb_function_ float ecb_binary32_to_float (uint32_t x) ecb_const; |
1231 | ecb_function_ ecb_const float ecb_binary32_to_float (uint32_t x); |
1202 | ecb_function_ float |
1232 | ecb_function_ ecb_const float |
1203 | ecb_binary32_to_float (uint32_t x) |
1233 | ecb_binary32_to_float (uint32_t x) |
1204 | { |
1234 | { |
1205 | float r; |
1235 | float r; |
1206 | |
1236 | |
1207 | #if ECB_STDFP |
1237 | #if ECB_STDFP |
… | |
… | |
1217 | x |= 0x800000U; |
1247 | x |= 0x800000U; |
1218 | else |
1248 | else |
1219 | e = 1; |
1249 | e = 1; |
1220 | |
1250 | |
1221 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
1251 | /* we distrust ldexpf a bit and do the 2**-24 scaling by an extra multiply */ |
1222 | r = ldexpf (x * (0.5f / 0x800000U), e - 126); |
1252 | r = ecb_ldexpf (x * (0.5f / 0x800000U), e - 126); |
1223 | |
1253 | |
1224 | r = neg ? -r : r; |
1254 | r = neg ? -r : r; |
1225 | #endif |
1255 | #endif |
1226 | |
1256 | |
1227 | return r; |
1257 | return r; |
1228 | } |
1258 | } |
1229 | |
1259 | |
1230 | /* convert a double to ieee double/binary64 */ |
1260 | /* convert a double to ieee double/binary64 */ |
1231 | ecb_function_ uint64_t ecb_double_to_binary64 (double x) ecb_const; |
1261 | ecb_function_ ecb_const uint64_t ecb_double_to_binary64 (double x); |
1232 | ecb_function_ uint64_t |
1262 | ecb_function_ ecb_const uint64_t |
1233 | ecb_double_to_binary64 (double x) |
1263 | ecb_double_to_binary64 (double x) |
1234 | { |
1264 | { |
1235 | uint64_t r; |
1265 | uint64_t r; |
1236 | |
1266 | |
1237 | #if ECB_STDFP |
1267 | #if ECB_STDFP |
… | |
… | |
1266 | |
1296 | |
1267 | return r; |
1297 | return r; |
1268 | } |
1298 | } |
1269 | |
1299 | |
1270 | /* converts an ieee double/binary64 to a double */ |
1300 | /* converts an ieee double/binary64 to a double */ |
1271 | ecb_function_ double ecb_binary64_to_double (uint64_t x) ecb_const; |
1301 | ecb_function_ ecb_const double ecb_binary64_to_double (uint64_t x); |
1272 | ecb_function_ double |
1302 | ecb_function_ ecb_const double |
1273 | ecb_binary64_to_double (uint64_t x) |
1303 | ecb_binary64_to_double (uint64_t x) |
1274 | { |
1304 | { |
1275 | double r; |
1305 | double r; |
1276 | |
1306 | |
1277 | #if ECB_STDFP |
1307 | #if ECB_STDFP |