Search

High-performance computing, or supercomputing, is the field of exploiting massive parallelism through advanced hardware technology and architectures to solve large application problems. There have been many important turning points at which the field has made significant shifts and changes. Among these are the golden years of vector supercomputers, the rapid developments of the massively parallel architectures in the late 1980s and early 1990s, the cluster computing era starting from the mid 1990s, the Grid Computing, from the late 1990s, the development of the Earth Simulator in Japan, the U.S. regaining of the leadership of Supercomputing with the introduction the IBM Blue Gene L, then finally the emergence of the first PetaFLOPS machine, the roadrunner, in mid 2008. This talk will consider the progress in this field from an architectural, performance and historical points of view. It will then introduce some of the implications and challenges associated with the latest developments and the needed research directions. It will then introduce some of the ongoing efforts that are likely to produce the next generation of Parallel Supercomputers, such as the DARPA High-Productivity Computing Systems initiative.

Reconfigurable computers (RCs) are those parallel systems that are designed around multiple general-purpose processors and multiple field programmable gate array (FPGA) chips. These systems can leverage the synergism between conventional processors and FPGAs to provide low-level hardware functionality at the same level of programmability as general-purpose computers. RCs have proposed very high processing capabilities for computationally intensive applications such as Image Processing. This is due to the inherently parallel operation paradigm of the FPGA hardware. In this paper we present the design and implementation of image processing kernels for RCs. This library of kernels have been tested and verified for performance on one of the state-of-the-art reconfigurable computers, SRC-6E. This paper shows that RCs are between 8 to 400 times faster than comparable Pentiums for image based tasks.

The Tropical Rainfall Measuring Mission (TRMM) satellite was launched with the first space-borne Precipitation Radar (PR) to collect accurate precipitation measurements. To validate the space data set, well-instrumented and calibrated ground validation (GV) sites are established. The paper is a first attempt to merging rainfall estimates from space-borne and ground based radars. We describe a technique to forecast bogus PR data that occurred before and after the passage of the PR over a GV site. Several simulations are performed at 10-minute intervals of GV reflectivity data at Melbourne, Florida on March 9, 1998 from 8:00 am to 9:00 am to track continuously PR reflectivity which overpass the GV site at 8:30 am. We apply genetic algorithms (GAs) to find the needed transformations to produce a time series of PR reflectivity data. The transformations include translation and rotation. By registering the PR original data following the transformed results, the new PR data are presented 20 minutes before and after the overpass. Finally, the statistic analyses are used to compare the relationship between these different sources of reflectivity. The increased correlation between GV and bogus PR reflectivity data demonstrated the potential use of GAs in merging multi-platform remote sensing data.

In remote sensing, image registration is to find the best transform between a reference and input images that may be different due to changes in position or altitude of or noise in the sensors. Image registration is one of the first steps in the analysis of remotely sensed images and requires high computational resources. The computation time is affected by two factors: search data size and search space. This paper describes an efficient image registration algorithm that uses multi-resolution wavelet decomposed images to reduce the search data size, and Genetic Algorithms to optimize the search solution space. Experimental results have shown subpixel accuracy and high efficiency over conventional methods.