#ifndef AWS_COMMON_ATOMICS_H #define AWS_COMMON_ATOMICS_H #include /** * Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved. * SPDX-License-Identifier: Apache-2.0. */ /** * struct aws_atomic_var represents an atomic variable - a value which can hold an integer or pointer * that can be manipulated atomically. struct aws_atomic_vars should normally only be manipulated * with atomics methods defined in this header. */ struct aws_atomic_var { void *value; }; /* Helpers for extracting the integer and pointer values from aws_atomic_var. */ #define AWS_ATOMIC_VAR_PTRVAL(var) ((var)->value) #define AWS_ATOMIC_VAR_INTVAL(var) (*(aws_atomic_impl_int_t *)(var)) /* * This enumeration specifies the memory ordering properties requested for a particular * atomic operation. The atomic operation may provide stricter ordering than requested. * Note that, within a single thread, all operations are still sequenced (that is, a thread * sees its own atomic writes and reads happening in program order, but other threads may * disagree on this ordering). * * The behavior of these memory orderings are the same as in the C11 atomics API; however, * we only implement a subset that can be portably implemented on the compilers we target. */ enum aws_memory_order { /** * No particular ordering constraints are guaranteed relative to other * operations at all; we merely ensure that the operation itself is atomic. */ aws_memory_order_relaxed = 0, /* aws_memory_order_consume - not currently implemented */ /** * Specifies acquire ordering. No reads or writes on the current thread can be * reordered to happen before this operation. This is typically paired with a release * ordering; any writes that happened on the releasing operation will be visible * after the paired acquire operation. * * Acquire ordering is only meaningful on load or load-store operations. */ aws_memory_order_acquire = 2, /* leave a spot for consume if we ever add it */ /** * Specifies release order. No reads or writes can be reordered to come after this * operation. Typically paired with an acquire operation. * * Release ordering is only meaningful on store or load-store operations. */ aws_memory_order_release, /** * Specifies acquire-release order; if this operation acts as a load, it acts as an * acquire operation; if it acts as a store, it acts as a release operation; if it's * a load-store, it does both. */ aws_memory_order_acq_rel, /* * Specifies sequentially consistent order. This behaves as acq_rel, but in addition, * all seq_cst operations appear to occur in some globally consistent order. * * TODO: Figure out how to correctly implement this in MSVC. It appears that interlocked * functions provide only acq_rel ordering. */ aws_memory_order_seq_cst }; /** * Statically initializes an aws_atomic_var to a given size_t value. */ #define AWS_ATOMIC_INIT_INT(x) \ { .value = (void *)(uintptr_t)(x) } /** * Statically initializes an aws_atomic_var to a given void * value. */ #define AWS_ATOMIC_INIT_PTR(x) \ { .value = (void *)(x) } AWS_EXTERN_C_BEGIN /* * Note: We do not use the C11 atomics API; this is because we want to make sure the representation * (and behavior) of atomic values is consistent, regardless of what --std= flag you pass to your compiler. * Since C11 atomics can silently introduce locks, we run the risk of creating such ABI inconsistencies * if we decide based on compiler features which atomics API to use, and in practice we expect to have * either the GNU or MSVC atomics anyway. * * As future work, we could test to see if the C11 atomics API on this platform behaves consistently * with the other APIs and use it if it does. */ /** * Initializes an atomic variable with an integer value. This operation should be done before any * other operations on this atomic variable, and must be done before attempting any parallel operations. * * This operation does not imply a barrier. Ensure that you use an acquire-release barrier (or stronger) * when communicating the fact that initialization is complete to the other thread. Launching the thread * implies a sufficiently strong barrier. */ AWS_STATIC_IMPL void aws_atomic_init_int(volatile struct aws_atomic_var *var, size_t n); /** * Initializes an atomic variable with a pointer value. This operation should be done before any * other operations on this atomic variable, and must be done before attempting any parallel operations. * * This operation does not imply a barrier. Ensure that you use an acquire-release barrier (or stronger) * when communicating the fact that initialization is complete to the other thread. Launching the thread * implies a sufficiently strong barrier. */ AWS_STATIC_IMPL void aws_atomic_init_ptr(volatile struct aws_atomic_var *var, void *p); /** * Reads an atomic var as an integer, using the specified ordering, and returns the result. */ AWS_STATIC_IMPL size_t aws_atomic_load_int_explicit(volatile const struct aws_atomic_var *var, enum aws_memory_order memory_order); /** * Reads an atomic var as an integer, using sequentially consistent ordering, and returns the result. */ AWS_STATIC_IMPL size_t aws_atomic_load_int(volatile const struct aws_atomic_var *var); /** * Reads an atomic var as a pointer, using the specified ordering, and returns the result. */ AWS_STATIC_IMPL void *aws_atomic_load_ptr_explicit(volatile const struct aws_atomic_var *var, enum aws_memory_order memory_order); /** * Reads an atomic var as a pointer, using sequentially consistent ordering, and returns the result. */ AWS_STATIC_IMPL void *aws_atomic_load_ptr(volatile const struct aws_atomic_var *var); /** * Stores an integer into an atomic var, using the specified ordering. */ AWS_STATIC_IMPL void aws_atomic_store_int_explicit(volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order memory_order); /** * Stores an integer into an atomic var, using sequentially consistent ordering. */ AWS_STATIC_IMPL void aws_atomic_store_int(volatile struct aws_atomic_var *var, size_t n); /** * Stores a pointer into an atomic var, using the specified ordering. */ AWS_STATIC_IMPL void aws_atomic_store_ptr_explicit(volatile struct aws_atomic_var *var, void *p, enum aws_memory_order memory_order); /** * Stores a pointer into an atomic var, using sequentially consistent ordering. */ AWS_STATIC_IMPL void aws_atomic_store_ptr(volatile struct aws_atomic_var *var, void *p); /** * Exchanges an integer with the value in an atomic_var, using the specified ordering. * Returns the value that was previously in the atomic_var. */ AWS_STATIC_IMPL size_t aws_atomic_exchange_int_explicit( volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order memory_order); /** * Exchanges an integer with the value in an atomic_var, using sequentially consistent ordering. * Returns the value that was previously in the atomic_var. */ AWS_STATIC_IMPL size_t aws_atomic_exchange_int(volatile struct aws_atomic_var *var, size_t n); /** * Exchanges a pointer with the value in an atomic_var, using the specified ordering. * Returns the value that was previously in the atomic_var. */ AWS_STATIC_IMPL void *aws_atomic_exchange_ptr_explicit( volatile struct aws_atomic_var *var, void *p, enum aws_memory_order memory_order); /** * Exchanges an integer with the value in an atomic_var, using sequentially consistent ordering. * Returns the value that was previously in the atomic_var. */ AWS_STATIC_IMPL void *aws_atomic_exchange_ptr(volatile struct aws_atomic_var *var, void *p); /** * Atomically compares *var to *expected; if they are equal, atomically sets *var = desired. Otherwise, *expected is set * to the value in *var. On success, the memory ordering used was order_success; otherwise, it was order_failure. * order_failure must be no stronger than order_success, and must not be release or acq_rel. * Returns true if the compare was successful and the variable updated to desired. */ AWS_STATIC_IMPL bool aws_atomic_compare_exchange_int_explicit( volatile struct aws_atomic_var *var, size_t *expected, size_t desired, enum aws_memory_order order_success, enum aws_memory_order order_failure); /** * Atomically compares *var to *expected; if they are equal, atomically sets *var = desired. Otherwise, *expected is set * to the value in *var. Uses sequentially consistent memory ordering, regardless of success or failure. * Returns true if the compare was successful and the variable updated to desired. */ AWS_STATIC_IMPL bool aws_atomic_compare_exchange_int(volatile struct aws_atomic_var *var, size_t *expected, size_t desired); /** * Atomically compares *var to *expected; if they are equal, atomically sets *var = desired. Otherwise, *expected is set * to the value in *var. On success, the memory ordering used was order_success; otherwise, it was order_failure. * order_failure must be no stronger than order_success, and must not be release or acq_rel. * Returns true if the compare was successful and the variable updated to desired. */ AWS_STATIC_IMPL bool aws_atomic_compare_exchange_ptr_explicit( volatile struct aws_atomic_var *var, void **expected, void *desired, enum aws_memory_order order_success, enum aws_memory_order order_failure); /** * Atomically compares *var to *expected; if they are equal, atomically sets *var = desired. Otherwise, *expected is set * to the value in *var. Uses sequentially consistent memory ordering, regardless of success or failure. * Returns true if the compare was successful and the variable updated to desired. */ AWS_STATIC_IMPL bool aws_atomic_compare_exchange_ptr(volatile struct aws_atomic_var *var, void **expected, void *desired); /** * Atomically adds n to *var, and returns the previous value of *var. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_add_explicit(volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order order); /** * Atomically subtracts n from *var, and returns the previous value of *var. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_sub_explicit(volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order order); /** * Atomically ORs n with *var, and returns the previous value of *var. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_or_explicit(volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order order); /** * Atomically ANDs n with *var, and returns the previous value of *var. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_and_explicit(volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order order); /** * Atomically XORs n with *var, and returns the previous value of *var. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_xor_explicit(volatile struct aws_atomic_var *var, size_t n, enum aws_memory_order order); /** * Atomically adds n to *var, and returns the previous value of *var. * Uses sequentially consistent ordering. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_add(volatile struct aws_atomic_var *var, size_t n); /** * Atomically subtracts n from *var, and returns the previous value of *var. * Uses sequentially consistent ordering. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_sub(volatile struct aws_atomic_var *var, size_t n); /** * Atomically ands n into *var, and returns the previous value of *var. * Uses sequentially consistent ordering. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_and(volatile struct aws_atomic_var *var, size_t n); /** * Atomically ors n into *var, and returns the previous value of *var. * Uses sequentially consistent ordering. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_or(volatile struct aws_atomic_var *var, size_t n); /** * Atomically xors n into *var, and returns the previous value of *var. * Uses sequentially consistent ordering. */ AWS_STATIC_IMPL size_t aws_atomic_fetch_xor(volatile struct aws_atomic_var *var, size_t n); /** * Provides the same reordering guarantees as an atomic operation with the specified memory order, without * needing to actually perform an atomic operation. */ AWS_STATIC_IMPL void aws_atomic_thread_fence(enum aws_memory_order order); #ifndef AWS_NO_STATIC_IMPL # include #endif /* AWS_NO_STATIC_IMPL */ AWS_EXTERN_C_END #endif