Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.programming.threads > #2168
| From | "Chris M. Thomasson" <no@spam.invalid> |
|---|---|
| Newsgroups | comp.programming.threads, comp.programming, comp.arch |
| Subject | Re: More about concurrent FIFO queues |
| Date | 2014-04-14 14:00 -0700 |
| Organization | Aioe.org NNTP Server |
| Message-ID | <lihi8t$t4s$1@speranza.aioe.org> (permalink) |
| References | <lihgs1$v1f$2@news.albasani.net> |
Cross-posted to 3 groups.
>"aminer" wrote in message news:lihgs1$v1f$2@news.albasani.net... >Hello, >We have to be smart more than that... >[...] >> As you have noticed if LOAD(&c.ver) = ver + 1 >> the thread will then pop up the item.. but this is a weakness >> i will explain to you why... >> With a CAS based method the pop() method can use a backoff mechanism >> when the CAS fails under high contention and this will give >> 3x times more throughput than the bakery algorithm in the pop() >> side. You can easily combine the XADD based strict bakery algorithm with a CAS based pop operation. It really depends on what the end users application demands are. Do they demand wait-free? Can they live with some sporadic lock-free operations being executed?
Back to comp.programming.threads | Previous | Next — Previous in thread | Find similar | Unroll thread
More about concurrent FIFO queues aminer <aminer@toto.net> - 2014-04-14 16:36 -0700 Re: More about concurrent FIFO queues "Chris M. Thomasson" <no@spam.invalid> - 2014-04-14 14:00 -0700
csiph-web