Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.programming.threads > #2324 > unrolled thread
| Started by | aminer <aminer@toto.net> |
|---|---|
| First post | 2014-05-11 19:13 -0700 |
| Last post | 2014-05-11 19:13 -0700 |
| Articles | 1 — 1 participant |
Back to article view | Back to comp.programming.threads
This PH.D doctor don't seem understand aminer <aminer@toto.net> - 2014-05-11 19:13 -0700
| From | aminer <aminer@toto.net> |
|---|---|
| Date | 2014-05-11 19:13 -0700 |
| Subject | This PH.D doctor don't seem understand |
| Message-ID | <lkp06p$qis$1@news.albasani.net> |
Hello, Read this: http://concurrencyfreaks.blogspot.ca/2014/05/c11-atomics-and-mpsc-mutual-exclusion.html They say this: -- "This may remind you of the MCS Lock, and indeed there are some similarities, like the fact that MCS also uses atomic_exchange(), named "swap" in the paper: http://www.cise.ufl.edu/tr/DOC/REP-1992-71.pdf But notice that the node of MCS uses a "locked" field which our lock does not have. Our MPSC lock has a single member on the node which is the "next" field. Moreover, in MCS, the unlock() function is lock-free because it does a compare-and-swap in a loop, while the unlock() in MPSC is wait-free because it does a single atomic_store()." -- This PH.D doctor don't seem understand something important from what i have read, read carefully what they have wrote here: "But notice that the node of MCS uses a "locked" field which our lock does not have. Our MPSC lock has a single member on the node which is the "next" field." Have you noticed with me or not ? they don't seem to understand that the "locked" field of the MCS lock allow the MCS lock to minimize efficiently the cache-coherence traffic, so it allows the MCS lock to minimize efficiently the contention and since it minimizes efficiently the cache-coherence traffic that means also it affects less the speed of the other threads , hence this will allow the MCS lock to scale, and that's not the case with there new lock , there new lock doesn't scale at all, and i have just explain to you why in my previous post. Thank you, Amine Moulay Ramdane.
Back to top | Article view | comp.programming.threads
csiph-web