Prince Pereira | c1c21d6 | 2021-04-22 08:38:15 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2016-present, Yann Collet, Facebook, Inc. |
| 3 | * All rights reserved. |
| 4 | * |
| 5 | * This source code is licensed under both the BSD-style license (found in the |
| 6 | * LICENSE file in the root directory of this source tree) and the GPLv2 (found |
| 7 | * in the COPYING file in the root directory of this source tree). |
| 8 | * You may select, at your option, one of the above-listed licenses. |
| 9 | */ |
| 10 | |
| 11 | #ifndef ZSTD_COMPILER_H |
| 12 | #define ZSTD_COMPILER_H |
| 13 | |
| 14 | /*-******************************************************* |
| 15 | * Compiler specifics |
| 16 | *********************************************************/ |
| 17 | /* force inlining */ |
| 18 | |
| 19 | #if !defined(ZSTD_NO_INLINE) |
| 20 | #if defined (__GNUC__) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */ |
| 21 | # define INLINE_KEYWORD inline |
| 22 | #else |
| 23 | # define INLINE_KEYWORD |
| 24 | #endif |
| 25 | |
| 26 | #if defined(__GNUC__) |
| 27 | # define FORCE_INLINE_ATTR __attribute__((always_inline)) |
| 28 | #elif defined(_MSC_VER) |
| 29 | # define FORCE_INLINE_ATTR __forceinline |
| 30 | #else |
| 31 | # define FORCE_INLINE_ATTR |
| 32 | #endif |
| 33 | |
| 34 | #else |
| 35 | |
| 36 | #define INLINE_KEYWORD |
| 37 | #define FORCE_INLINE_ATTR |
| 38 | |
| 39 | #endif |
| 40 | |
| 41 | /** |
| 42 | * FORCE_INLINE_TEMPLATE is used to define C "templates", which take constant |
| 43 | * parameters. They must be inlined for the compiler to eliminate the constant |
| 44 | * branches. |
| 45 | */ |
| 46 | #define FORCE_INLINE_TEMPLATE static INLINE_KEYWORD FORCE_INLINE_ATTR |
| 47 | /** |
| 48 | * HINT_INLINE is used to help the compiler generate better code. It is *not* |
| 49 | * used for "templates", so it can be tweaked based on the compilers |
| 50 | * performance. |
| 51 | * |
| 52 | * gcc-4.8 and gcc-4.9 have been shown to benefit from leaving off the |
| 53 | * always_inline attribute. |
| 54 | * |
| 55 | * clang up to 5.0.0 (trunk) benefit tremendously from the always_inline |
| 56 | * attribute. |
| 57 | */ |
| 58 | #if !defined(__clang__) && defined(__GNUC__) && __GNUC__ >= 4 && __GNUC_MINOR__ >= 8 && __GNUC__ < 5 |
| 59 | # define HINT_INLINE static INLINE_KEYWORD |
| 60 | #else |
| 61 | # define HINT_INLINE static INLINE_KEYWORD FORCE_INLINE_ATTR |
| 62 | #endif |
| 63 | |
| 64 | /* force no inlining */ |
| 65 | #ifdef _MSC_VER |
| 66 | # define FORCE_NOINLINE static __declspec(noinline) |
| 67 | #else |
| 68 | # ifdef __GNUC__ |
| 69 | # define FORCE_NOINLINE static __attribute__((__noinline__)) |
| 70 | # else |
| 71 | # define FORCE_NOINLINE static |
| 72 | # endif |
| 73 | #endif |
| 74 | |
| 75 | /* target attribute */ |
| 76 | #ifndef __has_attribute |
| 77 | #define __has_attribute(x) 0 /* Compatibility with non-clang compilers. */ |
| 78 | #endif |
| 79 | #if defined(__GNUC__) |
| 80 | # define TARGET_ATTRIBUTE(target) __attribute__((__target__(target))) |
| 81 | #else |
| 82 | # define TARGET_ATTRIBUTE(target) |
| 83 | #endif |
| 84 | |
| 85 | /* Enable runtime BMI2 dispatch based on the CPU. |
| 86 | * Enabled for clang & gcc >=4.8 on x86 when BMI2 isn't enabled by default. |
| 87 | */ |
| 88 | #ifndef DYNAMIC_BMI2 |
| 89 | #if ((defined(__clang__) && __has_attribute(__target__)) \ |
| 90 | || (defined(__GNUC__) \ |
| 91 | && (__GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)))) \ |
| 92 | && (defined(__x86_64__) || defined(_M_X86)) \ |
| 93 | && !defined(__BMI2__) |
| 94 | # define DYNAMIC_BMI2 1 |
| 95 | #else |
| 96 | # define DYNAMIC_BMI2 0 |
| 97 | #endif |
| 98 | #endif |
| 99 | |
| 100 | /* prefetch |
| 101 | * can be disabled, by declaring NO_PREFETCH build macro */ |
| 102 | #if defined(NO_PREFETCH) |
| 103 | # define PREFETCH_L1(ptr) (void)(ptr) /* disabled */ |
| 104 | # define PREFETCH_L2(ptr) (void)(ptr) /* disabled */ |
| 105 | #else |
| 106 | # if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_I86)) /* _mm_prefetch() is not defined outside of x86/x64 */ |
| 107 | # include <mmintrin.h> /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */ |
| 108 | # define PREFETCH_L1(ptr) _mm_prefetch((const char*)(ptr), _MM_HINT_T0) |
| 109 | # define PREFETCH_L2(ptr) _mm_prefetch((const char*)(ptr), _MM_HINT_T1) |
| 110 | # elif defined(__GNUC__) && ( (__GNUC__ >= 4) || ( (__GNUC__ == 3) && (__GNUC_MINOR__ >= 1) ) ) |
| 111 | # define PREFETCH_L1(ptr) __builtin_prefetch((ptr), 0 /* rw==read */, 3 /* locality */) |
| 112 | # define PREFETCH_L2(ptr) __builtin_prefetch((ptr), 0 /* rw==read */, 2 /* locality */) |
| 113 | # else |
| 114 | # define PREFETCH_L1(ptr) (void)(ptr) /* disabled */ |
| 115 | # define PREFETCH_L2(ptr) (void)(ptr) /* disabled */ |
| 116 | # endif |
| 117 | #endif /* NO_PREFETCH */ |
| 118 | |
| 119 | #define CACHELINE_SIZE 64 |
| 120 | |
| 121 | #define PREFETCH_AREA(p, s) { \ |
| 122 | const char* const _ptr = (const char*)(p); \ |
| 123 | size_t const _size = (size_t)(s); \ |
| 124 | size_t _pos; \ |
| 125 | for (_pos=0; _pos<_size; _pos+=CACHELINE_SIZE) { \ |
| 126 | PREFETCH_L2(_ptr + _pos); \ |
| 127 | } \ |
| 128 | } |
| 129 | |
| 130 | /* vectorization */ |
| 131 | #if !defined(__clang__) && defined(__GNUC__) |
| 132 | # define DONT_VECTORIZE __attribute__((optimize("no-tree-vectorize"))) |
| 133 | #else |
| 134 | # define DONT_VECTORIZE |
| 135 | #endif |
| 136 | |
| 137 | /* disable warnings */ |
| 138 | #ifdef _MSC_VER /* Visual Studio */ |
| 139 | # include <intrin.h> /* For Visual 2005 */ |
| 140 | # pragma warning(disable : 4100) /* disable: C4100: unreferenced formal parameter */ |
| 141 | # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ |
| 142 | # pragma warning(disable : 4204) /* disable: C4204: non-constant aggregate initializer */ |
| 143 | # pragma warning(disable : 4214) /* disable: C4214: non-int bitfields */ |
| 144 | # pragma warning(disable : 4324) /* disable: C4324: padded structure */ |
| 145 | #endif |
| 146 | |
| 147 | #endif /* ZSTD_COMPILER_H */ |