# utk/cs_dept/research/jones

There have been
5555 accesses to this library. (Last updated Fri Aug 26 13:33:57 EDT 1994)

utk/cs_dept/research/jones
# utk/cs_dept/research/jones

There have been
5555 accesses to this library. (Last updated Fri Aug 26 13:33:57 EDT 1994)

utk/cs_dept/research/jones
# utk/cs_dept/research/jones

There have been
5555 accesses to this library. (Last updated Fri Aug 26 13:33:57 EDT 1994)

Unstructured Mesh Computation
#
Parallel Algorithms for Unstructured
Mesh Computation

Many scientific and engineering applications demand the solution
of problems modeled on complex physical domains.
With the advent of massively parallel computation,
we have the computational power to solve problems
that have been computationally intractable on sequential machines.
However, the development of algorithms and software to
solve these complex problems has not kept pace with
these hardware developments.

To address these concerns we have developed parallel algorithms
and software in several areas to help solve problems arising
from complex domains.

## BlockSolve

BlockSolve
is a software library for solving large, sparse systems of linear
equations on massively parallel computers.
The matrices must be symmetric, but may have an
arbitrary sparsity structure.
BlockSolve
is a portable package that is
compatible with several different message-passing paradigms.
The BlockSolve project is a collaboration between
Mark Jones (CS, Tennessee)
and
Paul Plassmann (MCS, Argonne).

## Mesh Algorithms

Many scientific calculations have the property that the solution
changes rapidly in small regions of the total domain being modeled.
To efficiently model such problems, one can use an adaptive mesh
strategy. With an adaptive mesh, the number of grid points in a
region is adjusted so that solution is well approximated in that
region. For example, an adaptive strategy may use many grid points
in a region where the solution is changing rapidly, and many fewer
points in regions where the solution is almost constant.
Click on the following image to watch an adaptive mesh being
generated to model a Gaussian charge on a circular domain.

In this project we are developing parallel algorithms and software for
adaptively refining meshes according to user-specified criterion.
The mesh refinement project is a collaboration between
Mark Jones (CS, Tennessee),
Paul Plassmann (MCS, Argonne),
and
Lori Freitag (MCS, Argonne).