Circular lock-free buffer

后端 未结 18 751
悲哀的现实
悲哀的现实 2020-11-29 15:00

I\'m in the process of designing a system which connects to one or more stream of data feeds and do some analysis on the data than trigger events based on the result. In a t

18条回答
  •  广开言路
    2020-11-29 15:39

    The implementation in the boost library is worth considering. It's easy to use and fairly high performance. I wrote a test & ran it on a quad core i7 laptop (8 threads) and get ~4M enqueue/dequeue operations a second. Another implementation not mentioned so far is the MPMC queue at http://moodycamel.com/blog/2014/detailed-design-of-a-lock-free-queue. I have done some simple testing with this implementation on the same laptop with 32 producers and 32 consumers. It is, as advertised, faster that the boost lockless queue.

    As most of the other answers state lockless programming is hard. Most implementations will have hard to detect corner cases that take a lot of testing & debugging to fix. These are typically fixed with careful placement of memory barriers in the code. You will also find proofs of correctness published in many of the academic articles. I prefer testing these implementations with a brute force tool. Any lockless algorithm you plan on using in production should be checked for correctness using a tool like http://research.microsoft.com/en-us/um/people/lamport/tla/tla.html.

提交回复
热议问题