Groups > comp.programming.threads > #1038

Re: CAS operations and scalability...

Path	csiph.com!usenet.pasdenom.info!weretis.net!feeder4.news.weretis.net!eternal-september.org!feeder.eternal-september.org!mx04.eternal-september.org!.POSTED!not-for-mail
From	"aminer" <aminer@videotron.ca>
Newsgroups	comp.programming, comp.programming.threads
Subject	Re: CAS operations and scalability...
Date	Sun, 26 Aug 2012 14:08:14 -0500
Organization	A noiseless patient Spider
Lines	50
Message-ID	<k1domk$s2n$1@dont-email.me> (permalink)
References	<k1dira$jt9$1@dont-email.me> <FsqdncaQ9pkkw6fNnZ2dnUVZ_sGdnZ2d@earthlink.com>
Injection-Date	Sun, 26 Aug 2012 18:08:20 +0000 (UTC)
Injection-Info	mx04.eternal-september.org; posting-host="c43ca82f9e8d62a602307fe9d2e9b807"; logging-data="28759"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX1/JQCh7JweICSCA2nu47Tg8"
X-MimeOLE	Produced By Microsoft MimeOLE V6.00.2900.5512
X-RFC2646	Format=Flowed; Response
X-Newsreader	Microsoft Outlook Express 6.00.2900.5512
Cancel-Lock	sha1:+NStfX7s83r2P4FD/sb5Cr6zXVY=
X-Priority	3
X-MSMail-Priority	Normal
Xref	csiph.com comp.programming:2120 comp.programming.threads:1038

Cross-posted to 2 groups.

Show key headers only | View raw


Hello,


You are right Patricia.



Thank you,
Amine Moulay Ramdane.



"Patricia Shanahan" <pats@acm.org> wrote in message 
news:FsqdncaQ9pkkw6fNnZ2dnUVZ_sGdnZ2d@earthlink.com...
> On 8/26/2012 10:28 AM, aminer wrote:
> ...
>> When the CAS operation "goes on the bus", use of CAS can impair 
>> scalability.
>> but CAS can be accomplished locally -- that is, with no bus 
>> transactions --
>> and then it can scale.
> ...
>
> The problem the hardware faces doing CAS in-cache is that it must make
> sure that no other processor can write to the cache line containing the
> CAS target during the critical period of the cache.
>
> If multiple processors are attempting CAS operations on the same target
> at about the same time there are two costs:
>
> 1. The bus traffic to ensure that only one processor is writing to the
> cache line at any time, and that processor has the latest value of the
> cache line.
>
> 2. Repeated CAS operations, due to changes to the CAS target between
> doing whatever reads select the value and doing the CAS itself.
>
> You can do a couple of things to limit these costs. One is to make sure
> each CAS target is by itself in a cache line. That will reduce the risk
> of unnecessary bus traffic due to access to other fields in the same
> line. The other is to structure your code to minimize contention for any
> one CAS target, so that a CAS has a very high probability of the compare
> matching. You can instrument your CAS-using code to see how many CAS
> attempts you are making per successful CAS.
>
> Patricia

Back to comp.programming.threads | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

CAS operations and scalability... "aminer" <aminer@videotron.ca> - 2012-08-26 12:28 -0500
  Re: CAS operations and scalability... "aminer" <aminer@videotron.ca> - 2012-08-26 12:56 -0500
  Re: CAS operations and scalability... Patricia Shanahan <pats@acm.org> - 2012-08-26 10:36 -0700
    Re: CAS operations and scalability... "aminer" <aminer@videotron.ca> - 2012-08-26 14:08 -0500
  Re: CAS operations and scalability... "aminer" <aminer@videotron.ca> - 2012-09-13 19:13 -0500

csiph-web