ANNOUNCEMENT: NAG Launches PINEAPL Library
The Numerical Algorithms Group Ltd (NAG) is pleased to announce the
availability of Release 1 of the PINEAPL Library. This library has been
developed in the HPCN Fourth Framework project on Parallel Industrial
NumErical Applications and Portable Libraries (PINEAPL), in which NAG is
currently coordinating. This Library is the result of a coordinated
effort by the Center for Research on Parallel Computing and Supercomputers
National Research Council (Naples, Italy), NAG Ltd (Oxford, UK)
and the Centre for Novel Computing of the University
of Manchester (Manchester, UK).
One of the main goals of the project is to
increase the suitability of the existing NAG Parallel Library for
dealing with computationally intensive industrial applications by
appropriately extending the range of library routines. Additionally,
several industrial applications are being ported onto parallel
computers within the PINEAPL project by replacing sequential code
sections with calls to appropriate parallel library routines.
Consequently, the coverage of the PINEAPL Library is to a large
extent driven by the demands of the end-user applications
with the project and focuses on three major numerical areas:
- Sparse Linear Algebra,
- Fast Fourier Transforms, and
- Non-Linear Optimization.
Additional utility routines for data distribution, input/output and
process management purposes shield users to a large extent from having
to deal explicitly with the message-passing system - which may be either
MPI or PVM - on which the library is based.
The PINEAPL Library enables users to take advantage of the increased
computing power and memory capacity offered by parallel and distributed
computing environments. Targeted primarily at distributed memory
computers and networks of workstations, the PINEAPL Library also
performs well on shared memory computers whenever efficient
implementations of MPI or PVM are available. It offers increased speed
of execution over sequential numerical software on these systems and
allows problems to be solved which are beyond the memory capacity of
single processor systems.
The Library Contents
Sparse Linear Algebra:
The sparse linear algebra section includes iterative solvers for
symmetric and unsymmetric linear systems of equations, domain
decomposition and SSOR preconditioners, as well as a range of
auxiliary subprograms, for instance, a matrix-vector multiplication
routine. An important application area of this section is the
solution of partial differential equations using finite difference,
finite volume or finite element methods. All sparse linear algebra
routines can be used with either regular or irregular data
distributions, thereby supporting both problems with regular and
unstructured grids. Additional graph partitioning software enables
users to generate data distributions which are tailored to the
structure of the particular problem at hand.
FFTs:
The routines in the Fast Fourier Transform section calculate the
Discrete Fourier Transform (DFT) or the inverse DFT of a one-, two-
or three-dimensional sequence of complex data values.
Optimization:
The constraint non-linear optimization routines enable users to minimize
a given objective function based on either an easy-to-use quasi-Newton
method or a sequential quadratic programming algorithm. The independent
variables are constrained by simple upper and lower bounds in the
quasi-Newton method, whereas a mixture of linear and nonlinear equality or
inequality constraints can be imposed in the sequential quadratic
programming algorithm.
Support/Utility Facilities:
These utility routines shield users to a large extent from having
to deal explicitly with the message-passing system. Routines are provided
for process Management routines (under PVM or MPI), Input/Output
(for distributed data), data distribution
(dense and sparse matrices) and basic sparse matrix operations
(eg matrix-vector multiplication, transposed matrix multiplication, etc.)
Contact
Please contact Mishi Derakhshan for details of how to obtain a copy of
the library. The use of the library is initially free for a
trial period, but is subject to a signed collaboration agreement.
Mishi Derakhshan
NAG Ltd, Wilkinson House, Jordan Hill Rd, OXFORD, OX2 8DR, UK
E-mail: mishi@nag.co.uk, OR na.mderakhshan@na-net.ornl.gov
Tel: +44 (1865) 511 245 Fax: +44 (1865) 310 139