wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (05/05/89)
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
SCIENTIFIC COMPUTATION SEMINAR
-Thursday, May 11, 1989
Professor Johnny Peterson from IBM Bergen Scientific center,
will speak on ``Parallel Seismic Wave Propagation''.
TIME: 3:30 p.m.
ROOM: DC 1304
ABSTRACT
Seismic modelling to produce realistic seismic traces
is a computationally intensive problem. A 2D synthetic
wave propagation code using explicit finite differences
with absorbing boundary conditions has been implemented
on an Intel Hypercube with 32 processors. The
algorithms is highly parallel with good load balancing
between processors and speed-up proportional to the
number of processors being used. A vectorized version
of the code has been used to evaluate the performance
of the Intel IPSC/1-VX 5d vector hypercube.wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (09/02/89)
will speak on
``Adaptive Mesh Methods for the Fourier Solution of
Certain Wave Problems.''
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
SCIENTIFIC COMPUTATION SEMINAR
-Friday, September 8, 1989
Dimpy Pathria, graduate student, Dept. of Computer
Science will speak on ``Adaptive Mesh Methods For The
Fourier Solution of Certain Wave Problems.''
TIME: 10:30 a.m.
ROOM: DC 3540
ABSTRACT
In a previous talk we described the pseudospectral
split-step method for solving certain nonlinear
Schrodinger equations. Pseudospectral approximations
are often used for discretizing the spatial component
of evolutionary problems. In this talk we consider
enhancing the performance of such methods by two
adaptive grid generation schemes. The first scheme
addresses the temporal nonuniformity commonly exhibited
by solutions of nonlinear evolutionary equations. By
integrating the problem independently on interleaved
grids and comparing the Fourier coefficients as
calculated on each, it provides a mechanism for
automatically varying the degree of the interpolating
trigonometric polynomial. The second scheme addresses
the spatial nonuniformity of the solutions, in which
the resolution requirements vary over the domain. It
is an adaptive meshing algorithm for Fourier
interpolation, where the collocation nodes are
dynamically located in accordance with the needs of the
problem. Numerical experiments demonstrate the
advantages of the adaptive discretization schemes.wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (09/28/89)
McMaster University, will speak on
``An Adaptive Algorithm and Parallel Architecture for
Linear Associative Memory Problem.''
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
SCIENTIFIC COMPUTATION SEMINAR
-Thursday, October 5, 1989
Sanzheng Qiao, Communications Research Laboratory,
McMaster University, will speak on ``An Adaptive
Algorithm and Parallel Architecture for Linear
Associative Memory Problem''
TIME: 3:30 p.m.
ROOM: DC 1304
ABSTRACT
ABSTRACT
Neural networks have become popular since they can be
``trained'' to perform complex tasks and they are
suitable for massively parallel computing. The linear
associative memory problem is to find a weight matrix
after the memory has been exposed to pairs of
associated input and output patterns. This model can
then be used to ``recall'' memory from an input. This
paper presents an adaptive algorithm for computing the
weight matrix. A parallel implementation on a Warp
systolic computer shows the actual speedup.wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (02/07/90)
Amsterdam.
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
SCIENTIFIC COMPUTATION SEMINAR
Professor Rob H. Bisseling, Koninklijke/Shell-
Laboratorium, Amsterdam will speak on ``Parallel Sparse
Linear System Solving on a Transputer Network.
TIME: 3:30 p.m.
ROOM: DC 1304
DATE: Monday, February 12, 1990 *NOTE DATE!*
ABSTRACT
Large sparse systems of linear equations are important
in many areas, such as e.g. PDE-solving and Linear
Programming. This talk presents a parallel algorithm
for the direct solution of a sparse linear system
Ax = b on a local memory MIMD computer. The algorithm
-- - -
consists of the LU decomposition of A, followed by the
-
solution of two triangular systems.
The matrix A is distributed across the processors
-
according to the grid (i.e., scattered) distribution.
Each processor represents the nonzeroes of its part of
the sparse matrix by a two-dimensional linked list.
The pivot elements of the LU decomposition are
determined by a Markowitz criterion to preserve
sparsity, with threshold pivoting to maintain numerical
stability. The algorithm has been implemented in Occam
2, and experimental results have been obtained on a
mesh of 36 transputers.
This algorithm is part of the PARPACK library of
parallel programs for dense and sparse matrix
computations which is currently under development at
KSLA.wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (02/07/90)
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
SCIENTIFIC COMPUTATION SEMINAR
-February 15, 1990
Dr. George D. Byrne, Exxon Research & Eng. Co.,
Annandale, NJ will speak on ``The Solution of a
Co-polymerization Model with VODPK - Is That All There
Is?''
TIME: 3:30 p.m.
ROOM: DC 1302
ABSTRACT
In this presentation, we describe the process of taking
a chemical engineering problem, working with the
engineer to develop a reasonable mathematical model,
and indicate the process of developing a strategy for
solving it. The strategy includes the use of a
prototypical software package (VODPK), with a backup
procedure. The objective of this presentation is to
indicate the give and take process of developing a
model, developing a physical feel for a problem,
looking forward to an efficient computational solution,
and developing a bullet-proof code.
The clean-cut objective of the project is the numerical
solution of a co-polymerization model by the numerical
method of lines and the development of a turn key
production code. The mathematical model is a mixed
system of partial differential equations (PDEs),
algebraic equations, and quadratures. The PDEs are
also of mixed type - parabolic, hyperbolic, and
elliptic. With the algebraic equations, the PDEs
describe the growth of polymers within a catalyst
pellet in the presence of two monomers. The basic idea
is that the monomers join the polymer chains, which
grow and expand the spherical catalyst pellet. Some of
the reactions occur on a very fast time scale, while
others are on a much slower time scale. Consequently,
the pseudo-steady state assumption appears to be valid.
The quadratures are used to calculate several
parameters that are useful in the laboratory. There,
actual catalyst pellets are dissected and their
February 6, 1990
- 2 -
contents analyzed to tune the several parameters for
the manufacturing process and further numerical
simulation. By the way, polymers are composed of long
molecular chains of monomers and are used to make
plastics and resins.
Computationally, this is an interesting problem, since
we are calculating the solution of this system on the
interior of the spherical pellet, whose boundary is
moving outward as the polymerization process evolves in
time. Moreover, much of the action is near this moving
boundary. So a non-uniform grid is required for the
accurate resolution of this model. To conserve
computer time and storage, we plan to develop a black
box preconditioner and use it with a variant of GMRES
(SPIGMR) in the prototypical ODE solver VODPK (written
by Brown, Byrne, and Hindmarsh). The backup procedure
involves the use of a band solver in VODE and a variant
of the orthonormalization process in SPIGMR (written by
Brown and Hindmarsh).rbutterworth@watmath.waterloo.edu (Ray Butterworth) (02/22/90)
In article <3015@water.waterloo.edu> wlrush@water.waterloo.edu (Wenchantress Wench Wendall) writes: >Laval University will speak on ``Numerical Solution of >Viscous Compressible Flows.'' Um, is that "Mr. University", "Dr. University", "Ms. University", or what?
wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (02/23/90)
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
SCIENTIFIC COMPUTATION SEMINAR
-Friday, March 2, 1990
Professor Michel Fortin, Dept. of Math., Laval
University will speak on ``Numerical Solution of
Viscous Compressible Flows.''
TIME: 10:30 a.m. NOTE TIME
ROOM: DC 1304
ABSTRACT
This talk will focus on finite element simulation of
viscous compressible flows, including problems related
to the choice of elements and to treatment of shocks by
artificial viscosity methods. Numerical examples will
be presented.