This is the SECOND ANNOUNCEMENT for the upcoming parallel computation
workshop at MSRI:
http://sage.math.washington.edu/msri07/
TITLE: Interactive Parallel Computation in Support of Research
in Algebra, Geometry and Number Theory
LOCATION: MSRI, Will be January 29 -- February 2, 2007
FUNDING: https://secure.msri.org/forms/workshopreg/workshopreg?event_id=410.0
(Deadline: December 1, 2006 !!)
INVITED SPEAKERS:
David Bailey (Berkeley)
Henry Cohn (Microsoft Research)
Gene Cooperman (Northeastern University)
Brian Granger (Tech-X)
Robert Harrison (Oak Ridge National Lab)
Bill Hart (Warwick)
Yozo Hida (Berkeley)
Jason Martin (James Madison University)
Moreno Maza and Xie (Western Ontario)
Alfred Noel (UMass Boston / MIT)
Yi Qiang (UW)
Jean-Louis Roch (France)
Jan Verschelde (UIC)
ORGANIZERS:
Ifti Burhanuddin (University of Southern California, Computer Science)
James Demmel (UC Berkeley, Mathematics and Computer Science)
Edray Goins (Purdue University, Mathematics)
Erich Kaltofen (North Carolina State University, Mathematics)
Fernando Perez (University of Colorado, Applied Math)
Chair: William Stein (University of Washington, Mathematics)
Helena Verrill (Lousianna State University, Mathematics)
Joe Weening (CCR, Research)
DESCRIPTION: The goal of this workshop is to study and formulate
practical parallel algorithms that support interactive mathematical
research in algebra, geometry, and number theory, and to formulate
strategies to encourage implementation and testing of these ideas.
Computer manufacturers have begun delivering multiprocessor machines
onto desktops; indeed, this seems to be the only means for continuing
the pace of cpu power growth that we have become accustomed to.
At the moment, general purpose mathematical software packages rarely
exploit parallelism, and this is especially true in the areas of
algebra, geometry, number theory, and combinatorics. Dramatic
advances in performance will only be possible if parallelism can be
harnessed in ways that are transparent to users.
We hope to bring together a diverse group of mathematics and computer
science researchers and students to discuss algorithms, assess current
prospects, and suggest ways to move forward. Talks and discussions
will cover new algorithms that exploit parallelism, specific problems
likely to benefit from dramatic speedups from parallelism, and
strategies to encourage implementation of these ideas.
Examples of specific problems that we hope to address include:
(a) Parallel multimodular and p-adic methods for dense, sparse and
black box linear algebra over finite fields, the rational
numbers and rational functions (e.g., linear system solutions,
matrix multiplication, determinants and characteristic
polynomials, kernels, etc.)
(b) Parallel Groebner basis techniques, parallel triangular set construction
(c) Distributed general purpose integer factorization algorithms
(d) Uni- and multivariate polynomial arithmetic on large polynomials
(high degree, many terms) such as multiplication, GCD, factorization,
both for exact and approximate coefficients
(e) Parallel methods for searching for rational points on curves
(f) Distributed computation of large tables (e.g., elliptic curves,
modular forms, data about L-functions, number fields, etc.)
Applications of parallel computation to numerical problems, e.g., in
differential equations, linear algebra, etc., tend to have been more
fully developed than in algebraic areas, so we will invite experts in
those areas in order to hear about techniques that have been
successful at attacking those problems.
In sum, we aim at the following benefits for the mathematical community:
(a) new techniques and algorithms to exploit parallelism,
(b) exposure of areas and problems that may benefit from these ideas,
(c) a strategic assessment of how best to move the state of the art
forward, and
(d) a significant improvement of freely available general purpose
software for mathematical research.
FORMAT: There will be at most 3 hour long lectures a day, and will be
plenty of time for interaction between participants. In addition, we
will have one or two panel discussions about the current state of the
art, and strategies for implementing support for parallel computation.
Before the workshop the organizers will create a preliminary strategic
plan for parallel computation that will focus discussion during the
workshop, and which will hopefully be significantly improved during
the workshop.
REGISTRATION: Participants should register at the MSRI web page (when
that page is available in a few days). Any participant interested in
giving a talk should indicate this, and include an abstract on their
registration form. The deadline for applications to talk is Dec. 1,
and the organizing committee will make its selection and post a
schedule by Dec. 15. Application for financial support can also be
made on the MSRI registration page; we especially encourage graduate
students to apply.