Subject: | |
From: | |
Reply To: | |
Date: | Mon, 4 Sep 2006 18:47:56 +0200 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Dirk Craeynest a écrit :
> Most often, and in this announcement as well, only "add-on" approaches
> (such as OpenMP) are mentioned but not those built-in in a language
> (such as Ada tasking and protected objects). I quote:
Indeed. Most of the time people want parallelism with OpenMP/MPI. This
is quite low level compared to Ada. Maybe a good approach to this is to
point out that OpenMP/MPI is the procedural way to handle parallelism,
Ada goes far beyond as it has unified the OO paradigm and the
parallelism/concurrency. So in a sense the Ada solution (and Java if I
can pronounce this word here :) is the "revolution" with have seen with
the OO paradigm in the nineties.
Also note that OpenMP is a solution to vector based parallelism, and
this is far from being 100% of the problem. Most scientist applications
are designed with 3 stages : read data, do some computation (not always
vector based), write some data. OpenMP can't be used on stage 1 and 3.
Ada is not imposing any way of designing/implementing the parallelism
and can do great job at the three stages.
I don't know what to do at the moment with that... but I'll work very
soon on converting some sequential applications to parallel applications.
Pascal.
--
--|------------------------------------------------------
--| Pascal Obry Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--| http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595
|
|
|