The In-house High Performance Computer

by Joe Lee

Inside the Enterprise Competency Centre (ECC), there stands the high-performance computer, Sun Enterprise 10000 system, with 40 (400MHz, 8MB cache) CPUs, 14GB memory and 216 GB disk space, providing over 17Gflops of processing power.

It is particularly useful for research projects that require intensive CPU power, large amount of memory and massive disk storage. Normally, the machine is divided into two separate and equal domains for different users. Use of the entire machine as a single domain is possible with prior arrangement.

Software Availability

The machine would be useless if not run by powerful software. Sun Microsystems has provided a powerful development environment and the necessary tools, such as Forte 6.0 (including C 6.0, C++ 6.0, Fortran 6.0 compilers) and Sun HPC Clustertools 3.1 (including MPI 4.1, S3L 3.1), under its advanced operating system Solaris 7. For details of these tools, please refer to the documentation Website of Sun Microsystems at URL

Based on research needs, we have further acquired the following:

  • Matlab 6 with different toolboxes including Simulink, Control System, Optimization, Neural Network, Signal Processing and Image Processing
  • GNU products: g77, gcc, g++, c++, cpp, gzip, gunzip, gzexe, make, autoconf, automake, bison, flex, gawk, ghostscript, ghostview, m4, tar, texinfo
  • KAI C++ compiler (OpenMP compliance)
  • NAG f90 compiler and C parallel library
  • SAS
  • CFX 5, CFX 4
  • Others: X11R6.5.1, TeX, LaTex, netCDF, Perl

Training Course

To help researchers get familiar with the high-performance computer, its computing environment and programming tools, the Computing Services Centre (CSC) invited a specialist from the U.S.A. to conduct a 3-day training course on campus during 10-12 April 2001. The following topics were discussed in great details:

  • The memory hierarchy in RISC cache based systems
  • Sequential optimization techniques
  • SUN WorkShop Environment components:
    - The Fortran and C compilers (including key options)
    - The performance analyzer
  • Modulo Scheduling
  • Introduction into Parallelization
  • Shared Memory Parallelization
  • SUN HPC ClusterTool and MPI

Eighteen staff and research students participated the training course. Though the course was a little bit too long for some staff, most of them found that the contents of the training were rich and informative, the topics were very useful and could help them in their research work, and the instructor was fully prepared, professional and attentive.

Research Areas

The availability of this machine was first announced in late December 2000. Since then over thirty applications have been received. Currently, more than twenty research projects are using this machine. They are originated from several departments including AP, CS, EE, MA, MEEM, MS, MGT and the others. Typically, the projects are related to solving large-scale computation and simulation problems, data mining, statistical analysis of massive data collected from online and offline sources, implementation of parallel algorithms for computational electromagnetic and CAD software, computer-aided geometric design (CAGD), feature recognition from 3D models, phase synchronisation in chaotic systems, study of large molecules, analysis of tropical cyclones and many more.

Resource Utilisation

According to systems statistics, the machine has been heavily used and there were just a few percent of CPU time left idle during the past three months. As consumption of disk space kept increasing, more resources will be added when deemed necessary.

For further information of the high-performance computer, please visit the following URL: