[Todos] Fwd: Curso de High Performance Computing

[Ariel Zylberberg]

Gente,
A pedido de Diego Fernández Slezak del depto de computación reenvio
información sobre un curso de High Performance Computing, parte de la
Escuela de Ciencias Informáticas de este año.
Saludos,
Ariel Zylberberg

http://www.dc.uba.ar/eci (curso MT1)


Practical high performance computing for the modern age

Turno:

Mañana (9 a 12 hs): teórica.
Tarde (14 a 17 hs): práctica de laboratorio.
Duración: 30 horas.

Idioma:

Inglés.

Requisitos:

Conocimientos básicos de programación y de arquitecturas de computadoras.

Programa:

As the first decade of the 21st Century draws to a close, the domain
of supercomputing is achieving the milestone of reaching 1 Petaflops
(a million billion floating point operations per second) performance
driven by the exciting requirements of advanced applications in
science, medicine, and technology as well as by the continued
exponential growth in semiconductor technology reflected by Moore's
Law. But the effective exploitation of these unprecedented
technologies is only accomplished through a deep understanding of
their structure, operational properties, and semantics. This one-week
course, Practical High Performance Computing for the Modern Age, is
being presented to provide the attendee with a first detailed
understanding of supercomputing systems and their application to real
world problems. Based on the highly successful graduate level
introductory course, High Performance Computing: Models, Methods, and
Means, presented at Louisiana State University and broadcast in high
definition to other universities in the US and Europe, this short
course will give a comprehensive coverage of the essential elements of
supercomputing so that by its conclusion the participant will be
prepared to make informed decisions about his/her follow on pursuits
within the field. This course is suitable for the non-expert with
minimum prerequisites. Knowledge of basic programming language (e.g.,
C or Fortran) usage and user level exposure to a modern operating
system (e.g., Linux, Windows) is all that is required. This course
will be of particular value to those professionals seeking information
in preparation for computational science, HPC systems research, Linux
cluster administration, and hardware/software engineering. The five
days of lectures will cover:

Monday: Modern HPC Architecture and Systems
Tuesday: SMP Parallel Programming with OpenMP
Wednesday: MPP/Cluster Distributed Programming with MPI
Thursday: Towards the Petaflops Era with Multicore and Accelerators
Friday: Advanced Research Directions in HPC towards Exascale Computing

With the availability of an additional Teaching Assistant, hands-on
recitations will be provided to give the participant direct experience
with the fundamental tools and methods used in high performance
computing applications.

Profesor:

Thomas Sterling, Louisiana State University, USA.

Dr. Thomas Sterling holds a joint appointment as a Principal Scientist
at the NASA Jet Propulsion Laboratory and as a Faculty Associate at
Caltech's Center for Advanced Computing Research. Since receiving his
Ph.D from MIT as a Hertz Fellow in 1984, Dr. Sterling has pursued a
career of applied research in parallel computer architecture and high
performance computing at Harris Corp, the IDA Supercomputing Research
Center, and the NASA Goddard Space Flight Center and University of
Maryland, joining Caltech and JPL in 1996. Thomas Sterling is widely
recognized for his contributions to commodity cluster computing
through the Beowulf Project which he started in 1994 and is the
co-author of two books on clusters including "How to Build a Beowulf"
and "Beowulf Cluster Computing". Dr. Sterling was the PI of the HTMT
Project funded by a number of agencies to conduct the first
multidisciplinary multi-institutional in-depth study of future
Petaflops scale general purpose computers and is a co-author of a book
on "Enabling Technologies for Petaflops Computing." Currently he is
the PI of several projects that together are developing the MIND
architecture, an advanced PIM computing component sponsored by NASA,
Sandia National Laboratory, and DARPA.