compiler.h

  1/* SPDX-License-Identifier: GPL-2.0+ OR BSD-3-Clause */
  2/*
  3 * Copyright (c) Meta Platforms, Inc. and affiliates.
  4 * All rights reserved.
  5 *
  6 * This source code is licensed under both the BSD-style license (found in the
  7 * LICENSE file in the root directory of this source tree) and the GPLv2 (found
  8 * in the COPYING file in the root directory of this source tree).
  9 * You may select, at your option, one of the above-listed licenses.
 10 */
 11
 12#ifndef ZSTD_COMPILER_H
 13#define ZSTD_COMPILER_H
 14
 15#include <linux/types.h>
 16
 17#include "portability_macros.h"
 18
 19/*-*******************************************************
 20*  Compiler specifics
 21*********************************************************/
 22/* force inlining */
 23
 24#if !defined(ZSTD_NO_INLINE)
 25#if (defined(__GNUC__) && !defined(__STRICT_ANSI__)) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
 26#  define INLINE_KEYWORD inline
 27#else
 28#  define INLINE_KEYWORD
 29#endif
 30
 31#define FORCE_INLINE_ATTR __attribute__((always_inline))
 32
 33#else
 34
 35#define INLINE_KEYWORD
 36#define FORCE_INLINE_ATTR
 37
 38#endif
 39
 40/*
 41  On MSVC qsort requires that functions passed into it use the __cdecl calling conversion(CC).
 42  This explicitly marks such functions as __cdecl so that the code will still compile
 43  if a CC other than __cdecl has been made the default.
 44*/
 45#define WIN_CDECL
 46
 47/* UNUSED_ATTR tells the compiler it is okay if the function is unused. */
 48#define UNUSED_ATTR __attribute__((unused))
 49
 50/*
 51 * FORCE_INLINE_TEMPLATE is used to define C "templates", which take constant
 52 * parameters. They must be inlined for the compiler to eliminate the constant
 53 * branches.
 54 */
 55#define FORCE_INLINE_TEMPLATE static INLINE_KEYWORD FORCE_INLINE_ATTR UNUSED_ATTR
 56/*
 57 * HINT_INLINE is used to help the compiler generate better code. It is *not*
 58 * used for "templates", so it can be tweaked based on the compilers
 59 * performance.
 60 *
 61 * gcc-4.8 and gcc-4.9 have been shown to benefit from leaving off the
 62 * always_inline attribute.
 63 *
 64 * clang up to 5.0.0 (trunk) benefit tremendously from the always_inline
 65 * attribute.
 66 */
 67#if !defined(__clang__) && defined(__GNUC__) && __GNUC__ >= 4 && __GNUC_MINOR__ >= 8 && __GNUC__ < 5
 68#  define HINT_INLINE static INLINE_KEYWORD
 69#else
 70#  define HINT_INLINE FORCE_INLINE_TEMPLATE
 71#endif
 72
 73/* "soft" inline :
 74 * The compiler is free to select if it's a good idea to inline or not.
 75 * The main objective is to silence compiler warnings
 76 * when a defined function in included but not used.
 77 *
 78 * Note : this macro is prefixed `MEM_` because it used to be provided by `mem.h` unit.
 79 * Updating the prefix is probably preferable, but requires a fairly large codemod,
 80 * since this name is used everywhere.
 81 */
 82#ifndef MEM_STATIC  /* already defined in Linux Kernel mem.h */
 83#define MEM_STATIC static __inline UNUSED_ATTR
 84#endif
 85
 86/* force no inlining */
 87#define FORCE_NOINLINE static __attribute__((__noinline__))
 88
 89
 90/* target attribute */
 91#define TARGET_ATTRIBUTE(target) __attribute__((__target__(target)))
 92
 93/* Target attribute for BMI2 dynamic dispatch.
 94 * Enable lzcnt, bmi, and bmi2.
 95 * We test for bmi1 & bmi2. lzcnt is included in bmi1.
 96 */
 97#define BMI2_TARGET_ATTRIBUTE TARGET_ATTRIBUTE("lzcnt,bmi,bmi2")
 98
 99/* prefetch
100 * can be disabled, by declaring NO_PREFETCH build macro */
101#if ( (__GNUC__ >= 4) || ( (__GNUC__ == 3) && (__GNUC_MINOR__ >= 1) ) )
102#  define PREFETCH_L1(ptr)  __builtin_prefetch((ptr), 0 /* rw==read */, 3 /* locality */)
103#  define PREFETCH_L2(ptr)  __builtin_prefetch((ptr), 0 /* rw==read */, 2 /* locality */)
104#elif defined(__aarch64__)
105#  define PREFETCH_L1(ptr)  do { __asm__ __volatile__("prfm pldl1keep, %0" ::"Q"(*(ptr))); } while (0)
106#  define PREFETCH_L2(ptr)  do { __asm__ __volatile__("prfm pldl2keep, %0" ::"Q"(*(ptr))); } while (0)
107#else
108#  define PREFETCH_L1(ptr) do { (void)(ptr); } while (0)  /* disabled */
109#  define PREFETCH_L2(ptr) do { (void)(ptr); } while (0)  /* disabled */
110#endif  /* NO_PREFETCH */
111
112#define CACHELINE_SIZE 64
113
114#define PREFETCH_AREA(p, s)                              \
115    do {                                                 \
116        const char* const _ptr = (const char*)(p);       \
117        size_t const _size = (size_t)(s);                \
118        size_t _pos;                                     \
119        for (_pos=0; _pos<_size; _pos+=CACHELINE_SIZE) { \
120            PREFETCH_L2(_ptr + _pos);                    \
121        }                                                \
122    } while (0)
123
124/* vectorization
125 * older GCC (pre gcc-4.3 picked as the cutoff) uses a different syntax,
126 * and some compilers, like Intel ICC and MCST LCC, do not support it at all. */
127#if !defined(__INTEL_COMPILER) && !defined(__clang__) && defined(__GNUC__) && !defined(__LCC__)
128#  if (__GNUC__ == 4 && __GNUC_MINOR__ > 3) || (__GNUC__ >= 5)
129#    define DONT_VECTORIZE __attribute__((optimize("no-tree-vectorize")))
130#  else
131#    define DONT_VECTORIZE _Pragma("GCC optimize(\"no-tree-vectorize\")")
132#  endif
133#else
134#  define DONT_VECTORIZE
135#endif
136
137/* Tell the compiler that a branch is likely or unlikely.
138 * Only use these macros if it causes the compiler to generate better code.
139 * If you can remove a LIKELY/UNLIKELY annotation without speed changes in gcc
140 * and clang, please do.
141 */
142#define LIKELY(x) (__builtin_expect((x), 1))
143#define UNLIKELY(x) (__builtin_expect((x), 0))
144
145#if __has_builtin(__builtin_unreachable) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 5)))
146#  define ZSTD_UNREACHABLE do { assert(0), __builtin_unreachable(); } while (0)
147#else
148#  define ZSTD_UNREACHABLE do { assert(0); } while (0)
149#endif
150
151/* disable warnings */
152
153/* compile time determination of SIMD support */
154
155/* C-language Attributes are added in C23. */
156#if defined(__STDC_VERSION__) && (__STDC_VERSION__ > 201710L) && defined(__has_c_attribute)
157# define ZSTD_HAS_C_ATTRIBUTE(x) __has_c_attribute(x)
158#else
159# define ZSTD_HAS_C_ATTRIBUTE(x) 0
160#endif
161
162/* Only use C++ attributes in C++. Some compilers report support for C++
163 * attributes when compiling with C.
164 */
165#define ZSTD_HAS_CPP_ATTRIBUTE(x) 0
166
167/* Define ZSTD_FALLTHROUGH macro for annotating switch case with the 'fallthrough' attribute.
168 * - C23: https://en.cppreference.com/w/c/language/attributes/fallthrough
169 * - CPP17: https://en.cppreference.com/w/cpp/language/attributes/fallthrough
170 * - Else: __attribute__((__fallthrough__))
171 */
172#define ZSTD_FALLTHROUGH fallthrough
173
174/*-**************************************************************
175*  Alignment
176*****************************************************************/
177
178/* @return 1 if @u is a 2^n value, 0 otherwise
179 * useful to check a value is valid for alignment restrictions */
180MEM_STATIC int ZSTD_isPower2(size_t u) {
181    return (u & (u-1)) == 0;
182}
183
184/* this test was initially positioned in mem.h,
185 * but this file is removed (or replaced) for linux kernel
186 * so it's now hosted in compiler.h,
187 * which remains valid for both user & kernel spaces.
188 */
189
190#ifndef ZSTD_ALIGNOF
191/* covers gcc, clang & MSVC */
192/* note : this section must come first, before C11,
193 * due to a limitation in the kernel source generator */
194#  define ZSTD_ALIGNOF(T) __alignof(T)
195
196#endif /* ZSTD_ALIGNOF */
197
198#ifndef ZSTD_ALIGNED
199/* C90-compatible alignment macro (GCC/Clang). Adjust for other compilers if needed. */
200#define ZSTD_ALIGNED(a) __attribute__((aligned(a)))
201#endif /* ZSTD_ALIGNED */
202
203
204/*-**************************************************************
205*  Sanitizer
206*****************************************************************/
207
208/*
209 * Zstd relies on pointer overflow in its decompressor.
210 * We add this attribute to functions that rely on pointer overflow.
211 */
212#ifndef ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
213#  if __has_attribute(no_sanitize)
214#    if !defined(__clang__) && defined(__GNUC__) && __GNUC__ < 8
215       /* gcc < 8 only has signed-integer-overlow which triggers on pointer overflow */
216#      define ZSTD_ALLOW_POINTER_OVERFLOW_ATTR __attribute__((no_sanitize("signed-integer-overflow")))
217#    else
218       /* older versions of clang [3.7, 5.0) will warn that pointer-overflow is ignored. */
219#      define ZSTD_ALLOW_POINTER_OVERFLOW_ATTR __attribute__((no_sanitize("pointer-overflow")))
220#    endif
221#  else
222#    define ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
223#  endif
224#endif
225
226/*
227 * Helper function to perform a wrapped pointer difference without triggering
228 * UBSAN.
229 *
230 * @returns lhs - rhs with wrapping
231 */
232MEM_STATIC
233ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
234ptrdiff_t ZSTD_wrappedPtrDiff(unsigned char const* lhs, unsigned char const* rhs)
235{
236    return lhs - rhs;
237}
238
239/*
240 * Helper function to perform a wrapped pointer add without triggering UBSAN.
241 *
242 * @return ptr + add with wrapping
243 */
244MEM_STATIC
245ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
246unsigned char const* ZSTD_wrappedPtrAdd(unsigned char const* ptr, ptrdiff_t add)
247{
248    return ptr + add;
249}
250
251/*
252 * Helper function to perform a wrapped pointer subtraction without triggering
253 * UBSAN.
254 *
255 * @return ptr - sub with wrapping
256 */
257MEM_STATIC
258ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
259unsigned char const* ZSTD_wrappedPtrSub(unsigned char const* ptr, ptrdiff_t sub)
260{
261    return ptr - sub;
262}
263
264/*
265 * Helper function to add to a pointer that works around C's undefined behavior
266 * of adding 0 to NULL.
267 *
268 * @returns `ptr + add` except it defines `NULL + 0 == NULL`.
269 */
270MEM_STATIC
271unsigned char* ZSTD_maybeNullPtrAdd(unsigned char* ptr, ptrdiff_t add)
272{
273    return add > 0 ? ptr + add : ptr;
274}
275
276/* Issue #3240 reports an ASAN failure on an llvm-mingw build. Out of an
277 * abundance of caution, disable our custom poisoning on mingw. */
278#ifdef __MINGW32__
279#ifndef ZSTD_ASAN_DONT_POISON_WORKSPACE
280#define ZSTD_ASAN_DONT_POISON_WORKSPACE 1
281#endif
282#ifndef ZSTD_MSAN_DONT_POISON_WORKSPACE
283#define ZSTD_MSAN_DONT_POISON_WORKSPACE 1
284#endif
285#endif
286
287
288
289#endif /* ZSTD_COMPILER_H */