Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.programming > #3469 > unrolled thread
| Started by | "Chris M. Thomasson" <no@spam.invalid> |
|---|---|
| First post | 2013-06-19 17:58 -0700 |
| Last post | 2013-06-19 18:57 -0700 |
| Articles | 2 — 1 participant |
Back to article view | Back to comp.programming
This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by
below is the oldest one visible, not the original post.
Re: Intel SCC 48-core x86 Processor "Chris M. Thomasson" <no@spam.invalid> - 2013-06-19 17:58 -0700
Re: Intel SCC 48-core x86 Processor "Chris M. Thomasson" <no@spam.invalid> - 2013-06-19 18:57 -0700
| From | "Chris M. Thomasson" <no@spam.invalid> |
|---|---|
| Date | 2013-06-19 17:58 -0700 |
| Subject | Re: Intel SCC 48-core x86 Processor |
| Message-ID | <kptk4l$i2f$1@speranza.aioe.org> |
"aminer" wrote in message news:kptjeo$tkt$1@dont-email.me... > Hello, > I have read the following paper: > http://mcg.cs.tau.ac.il/papers/ppopp2013-rwlocks.pdf > So as you have noticed they are using the same method that are using > the Distributed RWLock by Dmitry Vyukov, so we have to change the > "function GetCurrentProcessorNumber:long;" inside MSync library by > function that returns the current NUMA number, by doing this we will > reduce coherence traffic and the distributed RWLock will be NUMA aware. > And that's easy to do. Distribution aside for a moment... Take a look at Dmitry Vyukov asymmetric rwmutex algorithm: https://groups.google.com/forum/?fromgroups#!original/comp.programming.threads/t9O4wI-co8Y/CIy9UAQ0rHoJ This really kicks some serious as$!
[toc] | [next] | [standalone]
| From | "Chris M. Thomasson" <no@spam.invalid> |
|---|---|
| Date | 2013-06-19 18:57 -0700 |
| Message-ID | <kptnja$pgb$1@speranza.aioe.org> |
| In reply to | #3469 |
"aminer" wrote in message news:51C284DA.4040307@toto.net... > Hello, > I don't understand something: > Look at this: > "wait_for_all_readers_notifications();" > How can you wait for all the readers ? this algorithm > do not tell us how to wait for all the readers ? Yikes! I posted to a brief overall sketch of the algorithm. Check this for more a more detailed sketch: https://groups.google.com/forum/?fromgroups#!original/comp.programming.threads/t9O4wI-co8Y/ya3QnzvbajUJ Sorry about that Aminer. Anyway, this is one way to do it, e.g., iterate and observe reader state in writer before the key aspect of asymmetric memory synchronization actually kicks in... http://home.comcast.net/~pjbishop/Dave/Asymmetric-Dekker-Synchronization.txt
[toc] | [prev] | [standalone]
Back to top | Article view | comp.programming
csiph-web