Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.programming.threads > #2325 > unrolled thread
| Started by | aminer <aminer@toto.net> |
|---|---|
| First post | 2014-05-11 19:28 -0700 |
| Last post | 2014-05-11 18:13 -0700 |
| Articles | 2 — 2 participants |
Back to article view | Back to comp.programming.threads
A new node based scalable and fair and starvation-free lock is coming... aminer <aminer@toto.net> - 2014-05-11 19:28 -0700
Re: A new node based scalable and fair and starvation-free lock is coming... "Chris M. Thomasson" <no@spam.invalid> - 2014-05-11 18:13 -0700
| From | aminer <aminer@toto.net> |
|---|---|
| Date | 2014-05-11 19:28 -0700 |
| Subject | A new node based scalable and fair and starvation-free lock is coming... |
| Message-ID | <lkp138$s1h$1@news.albasani.net> |
Hello, A new node based fair and starvation-free lock is coming... I have invented this lock today and i will post it today or tomorrow... It is scalable and fair and starvation-free. It's node based so it uses less memory ressources than my scalable AMLock but it is 50% slower than my scalable array based lock called AMLock , but even if it is slower than AMLock it minimizes efficiently the cache-coherence traffic and it's scalable and fair and starvation-free. Thank you, Amine Moulay Ramdane.
[toc] | [next] | [standalone]
| From | "Chris M. Thomasson" <no@spam.invalid> |
|---|---|
| Date | 2014-05-11 18:13 -0700 |
| Message-ID | <lkp782$kb5$1@speranza.aioe.org> |
| In reply to | #2325 |
"aminer" wrote in message news:lkp138$s1h$1@news.albasani.net... > Hello, > A new node based fair and starvation-free lock is coming... > I have invented this lock today and i will post it today or tomorrow... [...] Starvation-free node based queue. Humm... I am interested in the algorithm! :^) Can you please try to run the algorithm against a verification tool like Relacy and/or ThreadSanitizer! If I like it, I might code it up in C++ and run it through Relacy myself. ;^) Also, have you read up on any of the hybrid versions? Each node contains an array. So, you can get unbounded access, and still get decent performance from using the per-node array... Also, using spinlocks in user code is a bit sketchy... All of the calls to Sleep(0) make be a bit uneasy to say the least. FWIW, this is why I use a set of eventcounts to handle any conditional waiting. Therefore, I do not need to spin. BTW, using spinlocks in user land code can be a bit sketchy... You can use an eventcount to get rid of any nasty spin- waiting.
[toc] | [prev] | [standalone]
Back to top | Article view | comp.programming.threads
csiph-web