Parallel computer

    Project Details


    In 1994 the Ministry of Education called for proposals concerning experiments with parallel computers within the area of scientific computing. Three research centres including DCRS were selected and a Silicon Graphics (SGI) PowerChallenge was procured and installed at DCRS in September 95. It is a Symmetric Multi Processor (SMP) machine with 4 GBytes of RAM and twelve 90 MHz R8000 processors offering 4.3 GFLOPS. DCRS has contributed 232 GB of disk, high-speed data net etc. and undertakes the systems administration. The entire capacity of the computer is allocated DCRS. However, also a few designated projects conducted by the remainder of the institute make use of the computer.
    Typically, a substantial digital signal processing is required to convert remote sensing signals into imagery suitable for interpretation by the end user. Since the development and experimental usage of remote sensing software are often iterative processes it is crucial to minimise the execution time of each program by making the code running on several CPUs simultaneously. Consequently, parallel programming has been given high priority, and now almost all compute-intensive jobs are parallel jobs. The compilers have options allowing parallel code to be generated automatically, but in practice a manual effort is usually needed. Still, the additional effort is moderate. Typical speed-up factors are 4-6 when running on 8 processors.
    In 1996 the parallel computer was connected to DCRS's HP workstations with a high-speed data net based on FDDI and 100VG technology, and in 1997 a 100BaseTX net to the SGI workstations was established.
    In 1997 the three parallel computer projects were subject to an evaluation based on reports and a workshop, and in 1998 it was decided to extend the project period from three to four years. (Related projects - see:
    Effective start/end date28/09/199530/09/1999


    • Unknown


    Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.