Atomic memory access

最新推荐文章于 2024-02-19 16:16:04 发布
最新推荐文章于 2024-02-19 16:16:04 发布
阅读量315 收藏
点赞数
CC 4.0 BY-SA版权
分类专栏： C++ GCC
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.youkuaiyun.com/old__/article/details/8250720
本文介绍了GCC提供的内置原子操作函数，这些函数通过内存屏障(memory barrier)确保了在多核CPU环境下内存访问的原子性。文章解释了full barrier的概念，并探讨了如何在多进程环境中使用这些函数来保证内存访问的同步。

摘要生成于 C知道 ，由 DeepSeek-R1 满血版支持， 前往体验 >

Gcc 提供了一些built in的函数,以实现内存访问的原子性:具体函数在 http://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/Atomic-Builtins.html

In most cases, these builtins are considered a full barrier. That is, no memory operand will be moved across the operation, either forward or backward. Further, instructions will be issued as necessary to prevent the processor from speculating loads across the operation and from queuing stores after the operation. 

full barrier 就是说明了这些函数是通过cpu以memory barrier方式访问内存的，所以即使在多核CPU下，这些函数也可保证其原子性. 

也就隐含的说明了，不管是单进程还是多进程只要是对函数中ptr参数的内存地址的访问，都是原子进行的——如ptr是在共享内存中，那多个进程对ptr所指向的内存的访问都是原子进行的，保证了在多个进程间的同步，即使在多核cpu下.

Memory barrier: http://en.wikipedia.org/wiki/Memory_barrier

C++ NUMA (Non-Uniform Memory Access) 架构优化：跨内存节点访问
weixin_41455464的博客
07-12 269
通过理解NUMA的原理，并结合C++提供的工具，你可以让你的程序在多核服务器上发挥出最大的潜力。记住，数据本地化和线程绑定是NUMA优化的核心，合理的数据划分和避免跨节点访问是提高性能的关键。好的，各位观众老爷，今天咱们来聊聊C++里的NUMA架构优化，这玩意儿听起来高大上，其实就是怎么让你的程序在多核服务器上跑得飞起，不卡顿。别担心，我保证用最接地气的方式，把这背后的门道给你们讲明白。如果你的程序不考虑NUMA，那很可能出现“远水解不了近渴”的情况，线程们抢着访问同一个远程内存节点，导致性能瓶颈。
原子访问(Atomic Access)
luccs624061082的专栏
06-06 237
在编程中，原子动作是一个突然有效地发生。一个原子动作在中途是不能停止的：要不完全执行完，要不一点也不执行。直到原子操作完全执行，它的副作用才会可见。 我们已经看见了自增长表达式，比如c++，不描述一个原子操作。甚至非常简单的表达式可以定义复杂的动作，可以分解为其他的动作。可是，有些动作你可以指定为原子的： 1 引用变量和大多数基本数据类型（除了lon...
Atomic Builtins - Using the GNU Compiler Collection (GCC) GCC 提供的原子操作
weixin_33860737的博客
08-06 195
http://gcc.gnu.org/onlinedocs/gcc-4.4.3/gcc/Atomic-Builtins.html   gcc从4.1.2提供了__sync_*系列的built-in函数，用于提供加减和逻辑运算的原子操作。   5.47 Built-in functions for atomic memory access The following builtins ar...
Atomic-accesses
wmzjzwlzs的专栏
04-25 587
转自：Documentation – Arm Developer Atomic accesses An atomic access is a term for a series of accesses to a memory region. Atomic accesses are used by managers when they would like to perform a sequence of accesses to a particular memory region, while be
原子访问（Atomic Access）
平凡
07-05 4085
在编程中，原子操作是指在一次发生中生效。一个原子操作不能在中途停止：它要么完整发生，要么不发生。没有副作用的原子操作是可见的，直到完成。   我们见过自增表达式了，如c++，并不是原子操作，甚至非常简单的表达式会定义复杂的操作，这些可以被分解为其它的操作。有一些操作可以确认是原子操作，如： 对引用类型变量和大部分基本类型变量（除long和double）的读、写是原子的 所有声明为volatile变量
RDMA（Remote Direct Memory Access）技术
07-02
### RDMA（Remote Direct Memory Access）技术详解 #### 一、引言 随着现代数据中心对高性能计算(HPC)和云计算需求的增长，网络架构和技术也在不断发展以满足这些需求。RDMA（远程直接内存访问）作为一种先进的...
About Cache Coherence, Atomic Operation, Memory Ordering, Memory Barrier, Volatile
Meditation
04-06 7536
写这篇文章的起因是看到何登成博士的一个微博问题，自己想but
Summary of Resistive Random Access Memory
Blog of Lucky H6
07-16 801
Summary for Resistive Random Access Memory Part 1 Basic Understanding of RRAM physics Concepts and Terminologies HRS: high resistance state LRS: low resistance state Structure: Metal - Insulator (Oxide) - Metal Set: The switching event from HRS to LRS
C++ 11 开发中的 Atomic 原子操作.rar_C++11、原子_atomic c++_c++ atomic使用_c++
09-23
6. **原子成员访问（Atomic Member Access）** 对于包含原子成员的复杂类型，C++11提供了一种原子成员访问的机制。这允许我们只对对象的某个特定成员进行原子操作，而不需要整个对象都是原子的。 7. **原子类型...
Atomic Access　in thread
Java天空
01-15 1141
 !Atomic Access(an atomic action cannot stop in the middle;no side effects of an atomic action are visible until the action is complete):         !Reads and writes are atomic for reference variables a
Chapter6：Atomic Accesses
北枫凉的博客
08-14 428
AXI的原子访问
Built-in functions for atomic memory access
innobase的博客
08-08 428
5.44 Built-in functions for atomic memory accessThe following builtins are intended to be compatible with those described in the Intel Itanium Processor-specific Application Binary Interface, section 7
The Java™ Tutorials — Concurrency ：Atomic Access 原子访问
FIRE_TRAY的Android之路
02-15 674
The Java™ Tutorials — Concurrency ：Atomic Access 原子访问 原文地址：https://docs.oracle.com/javase/tutorial/essential/concurrency/atomic.html 关键点 原子操作：不可打断的操作原子操作类型  对于引用变量，和大部分的原始数据类型的读写（除long
Atomic access 与 Exclusive access 的区别
power1952的专栏
02-19 1381
Atomic access（原子访问）和Exclusive access（独占访问）是两种不同的访问机制，它们主要用于处理并发访问和多线程环境中的资源访问问题。
Armv8.x-A扩展特性
Hacker
02-06 1329
文章目录Armv8.1-AArmv8.2-AArmv8.3-AArmv8.4-A如何判断当前CPU使用的Armv8.x的.x是多少呢Cortex-A Processors supporting Armv8.0-ACortex-A Processors supporting Armv8.2-A 备忘。 Armv8.1-A Atomic memory access instructions (AArch64) Limited Order regions (AArch64) Increased Virtua.
浅析atomic原子性
tobemlw的博客
01-11 2340
什么是atomic java.util.concurrent.atomic包中的很多类使用很高效的机器级指令（而不是锁）来保证操作的原子性。 有哪些常用的类 查看api可以知道atomic包下含有以下 比较常用的有AtomicBoolean、AtomicInteger、AtomicIntegerArray、AtomicLong等。 看看怎么使用 以AtomicLong举例，得到
atomic
最新发布
03-08
### Atomic Operations in Programming and Concurrency In concurrent programming, atomic operations play a crucial role in ensuring data integrity when multiple threads access shared resources simultaneously. An operation is considered atomic if it completes instantly from the perspective of other concurrently executing threads or processes. Atomicity guarantees that an operation will not be interrupted once started until completion. This prevents race conditions where two or more threads attempt to modify the same variable at nearly the same time leading to unpredictable outcomes[^1]. #### Usage of Atomics To handle synchronization without traditional locks, C++ provides `std::atomic` types within its standard library. These allow performing read-modify-write actions atomically: ```cpp #include <atomic> #include <thread> std::atomic<int> counter{0}; void increment_counter() { for (int i = 0; i < 1000; ++i) { counter.fetch_add(1, std::memory_order_relaxed); } } int main() { const unsigned thread_count = 10; std::vector<std::thread> threads; for (unsigned i = 0; i < thread_count; ++i) { threads.emplace_back(increment_counter); } for (auto& t : threads) { t.join(); } // Expected value should be close to 10 * 1000 = 10000 std::cout << "Counter: " << counter.load() << '\n'; } ``` The above example demonstrates how `fetch_add()` method performs addition as an indivisible action across different threads safely avoiding any potential conflicts during updates[^2]. #### Memory Ordering Constraints When working with atomics, memory ordering constraints dictate what changes become visible to other threads after an atomic operation has completed. The options range from very weak (`memory_order_relaxed`) allowing maximum performance but minimal visibility guarantees up to fully synchronized (`memory_order_seq_cst`) providing strictest consistency between all involved parties[^3]. For instance, setting appropriate barriers ensures correct program behavior even under relaxed orderings by preventing certain instructions from being reordered around these points thus preserving intended semantics despite optimizations performed by compilers or hardware architectures. #### Benefits Over Traditional Synchronization Mechanisms Using atomic variables offers several advantages over conventional locking mechanisms such as mutexes including better scalability due to reduced contention among threads accessing non-overlapping parts of state space independently alongside lower overhead costs associated with acquiring/releases compared to full-fledged mutual exclusions constructs[^4]. However, developers must carefully design algorithms taking into account possible limitations imposed by chosen levels of coarseness versus fineness regarding granularity choices made concerning protected regions since overly broad scopes could negate some benefits derived otherwise through finer control offered here instead[^5].
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值