[rappel] Exposés de Bjarne Stroustrup

Durant sa visite à l'INRIA Sophia Antipolis, Professeur Bjarne
Stroustrup de Texas A&M University et AT&T fera deux esposés dans le
cadre du :

* colloquium Morgenstern (pour les 20 ans de l'UR)

Titre : Do general-purpose languages have a future?

Résumé :
As the computing world matures, the roles of computer
professionals are becoming more specialized. In particular, a
programmer can spend a whole career doing work in embedded
systems or data analysis without a need to gain expertise in
other fields. Would such programmers be best served by
completely different special-purpose languages? What are the
fundamental and commercial factors that drive language
evolution? What are the roles of program development environments,
libraries, and tools? I think that general-purpose languages
will have a key role in the programming world, but that the role
will evolve and differ from what most people think of today. To
make the discussion a bit concrete, I'll base some of my
observations on examples from current C++ and its possible future
developments.

Lieu : ESSI- Amphithéatre Ouest - 930,
route des Colles - Sophia Antipolis


* séminaire GALAAD + club des utilisateurs du calcul
parallèle en partenariat avec la région PACA

Titre : distributed and parallel computing from a C++ perpsective

Résumé :
in this talk I willl explore two areas: (1) conventional
systems and application programming using threads and RPC
(typically using fairly conventional computers) and, (2)
scientific/numeric computing using high-performance computers
(typically multi-processors or clusters). Our main thesis is
that the programming model presented to the user must be
radically simplified compared to what is provided by most
commercial and academic systems. The problems faced by end
users are simply too difficult to program using explicit
parallelism. Furthermore, the sophisticated mechanisms offered
to programmers cause fragmentation of the user community leading
to the use of tedious low-level and inefficient high-level
facilities. We must develop a few naive programming models and
support them well with optimizing tools. These naive models must be
high-level and domain specific. Only by letting the physicists
program the physics and leave computer concepts, such as
loops and memory hierarchies, to the computer scientists can
we hope for dramatic progress. We will consider language
features, library facilities, type analysis and program
transformation approaches, and adaptive techniques. We
think that our approach requires a language with enough static
structure to enable semantic analysis and enough flexibility to
define convenient domain specific libraries. This talk aims to
state fundamentals and provoke discussion, rather than to
present detailed results or mechanisms.

Lieu : INRIA de Sophia Antipolis - salle Euler Violet (006)
2004, route des lucioles - Sophia Antipolis

-- Gaby