HPC Applications Sectors
HPC Applications Sectors
High Performance Computers are used for many applications, and it
would be futile to try to list them all. Some of the major market sectors
are:
- Numerically Intensive Simulations, including:
Computational Fluid Dynamics (CFD);
Computational Electromagnetics (CEM);
Oil Reservoir Simulations;
Combustion Modelling;
Molecular Dynamics Modelling;
Quantum Chromodynamics;
Quantum Chemistry;...
- Graphical Visualisation including:
Image Processing;
- Data Base Operations and Information Systems including:
Client and Inventory Database Management;
Datamining;
Online Transaction Processing;
Management Information Systems;
Geographic Information Systems;
Seismic Data Processing;...
- Real Time Systems and Control Applications including:
Hardware and Robotics Control;
Speech Processing;
Pattern Recognition;...
Furthermore, these sectors do not split cleanly, but are cross linked
amongst specialist areas of knowledge. An attempt can be made to
break down applications by the generic class of method that is
required. These include:
- Linear algebra: including:
dense matrix manipulation;
sparse matrix manipulation;
eigensystems;...
- Stochastic or Monte Carlo Simulations.
- Data Histogramming.
- Filtering
or the finest-grained category of all is perhaps by algorithm with
examples like:
- Fast Fourier Transforms.
- Legendre Transforms.
- MultiGrid Methods.
- Fast Multipole.
- N-Body Particle Dynamics.
- Particle in Cell dynamics.
- Structured Grid Methods.
- Unstructured Grid Methods.
- Adaptive Grid Methods.
- Clustering algorithms.
- Text Searching.
It is generally the case that many parallel and high performance
processing systems have a reasonably well known subset of elemental
algorithms that they can run very efficiently. Unfortunately, real
applications often have many algorithmic components and a tradeoff
analysis is necessary to find the best system for a particular
application.
Ken Hawick
Saleh Elmohamed