Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming.threads > #1902

Re: About Amdahl's law 1 ...

Cross-posted to 2 groups.

Show all headers | View raw

On Thu, 10 Oct 2013 14:30:59 -0700, aminer <aminer@toto.net> wrote:

>
>
>I have wrote:
>"But Amine, you have said that the Amdahl's law predict scalability
>of the worst case contention scenario , so the Amdahl's calculation in 
>parallel computing looks like the calculation of  the worst case 
>complexity in computer science?
>
>That's right, cause what is, in your opinion , the purpose of the 
>Amdahl's law if it's not to predict worst case scalability by 
>calculating the maximum bound of time time inside the critical sections 
>and the maximum bound of the time inside the parallel part and doing the 
>Amdahl's calculation to predict the worst case contention scenario
>hence the worst case scalability? so i think this will predict that the 
>scalability will be equal or greater to the result that we have
>found with the Amdahl's law. This is the purpose of Amdahl's law:
>it's also to predict the worst case scalability."
>
>
>To do a better scalability prediction, you have to use the
>same number of threads than the number of cores, and each
>thread must run a a separate core, to avoid context switches
>and you have to use Locks that lower better the cache-coherence traffic,
>such us the scalable Anderson Lock or the scalable MCS Lock
>or my scalable and distributed fair Lock, and try to find a maximum
>bound of the time inside your critical sections and maximum bound
>inside of the time inside the parallel part , and do the Amdahl's 
>calculation with them to find the result of the worst case scalability, 
>so the scalability will equal or be greater to this result. I think 
>this is the purpose of the Amdahl's law: to predict the worst case 
>scalability.


Please don't top-post.

As has been pointed out to you several times, Amdahl's law predicts
the *best* possible speedup from parallelization.  All those messy
real world considerations, like contention for locks, will cause the
*actual* speed up to be worse.

The major point of Amdahl's law is to put a limit on the amount of
useful parallelism in a workload.  If 10% of your workload is serial,
then you will never run faster than ten times the performance of a
single core, no matter if you build a machine with ten processors or a
million.

Which does not apply to independent workloads running at the same
time.  Most compilers, for example, are fairly serial in nature -
almost everything they do depends on the prior bits being done.  So
most compilers can make little use of more than one core.  OTOH, a
parallel make tool can certainly run several copies of the compiler,
compiling different object modules, in parallel, greatly improving
your build speeds.  But in the latter case Amdahl's law still applies
to the whole build system, and while the individual compiles
parallelize well, the limiting serial components are often the
analysis phase for the make tool and the link step, as well as the
biggest single compile.  And again, if you are building multiple
executables, several independent link steps might parallelize as well.

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

Thread

About Amdahl's law 1 ... aminer <aminer@toto.net> - 2013-10-10 13:08 -0700
  Re: About Amdahl's law 1 ... aminer <aminer@toto.net> - 2013-10-10 13:57 -0700
  Re: About Amdahl's law 1 ... Ben Bacarisse <ben.usenet@bsb.me.uk> - 2013-10-10 19:14 +0100
    Re: About Amdahl's law 1 ... aminer <aminer@toto.net> - 2013-10-10 14:45 -0700
      Re: About Amdahl's law 1 ... Ben Bacarisse <ben.usenet@bsb.me.uk> - 2013-10-10 20:56 +0100
  Re: About Amdahl's law 1 ... aminer <aminer@toto.net> - 2013-10-10 14:30 -0700
    Re: About Amdahl's law 1 ... Robert Wessel <robertwessel2@yahoo.com> - 2013-10-10 23:12 -0500

csiph-web