High Performance Algorithm and Architecture for Hyper Spectral Data Processing

Sponsor: NASA
Prinicpal Investigator: Professor Tarek El-Ghazawi

Hyperspectral remote sensing sensors are capable of collecting remote sensing imagery at several hundred bands over the spectrum. Under these advanced observation tools, the observed phenomena can be identified with unique signatures, however, the resulting data volumes are quite massive. Processing hyperspectral data using new efficient techniques and architectures is therefore very critical.

Our advanced hyperspectral processing falls into three different areas;

  1. Investigating high-performance algorithms for dimension reduction of hypersepctral data based on methods such as Principal Component Analysis (PCA), Projection Pursuit, and Wavelet analysis.
  2. Investigating high-performance algorithms for remote sending data fusion.
  3. Investigating reconfigurable computer architectures for satellite onboard processing of hyperspectral imagery for dimension reduction and cloud detection

UPC Project

SPONSOR: DoD

Principal Investigator: Tarek El-Ghazawi

UPC, or Unified Parallel C, is a parallel extension of ANSI C. UPC follows a distributed shared memory programming model aimed at leveraging the ease of programming of the shared memory paradigm, while enabling the exploitation of data locality. The HPC Lab coordinates the UPC consortium activities and actively participates in the development of the language specifications. The HPCL is also leading the performance evaluation and validation research efforts, as well as the development of the UPC I/O specifications in collaboration with Argonne National Lab and UC Berkeley.

Ultraviolet Project

SPONSOR: DARPA

Principal Investigator: Tarek El-Ghazawi

In an effort to boost high-performance computing users productivity and cut time-to-solution, which is a function of development time and execution time, Silicon Graphics, along with the George Washington University, the Massachusetts Institute of Technology, University of Minnesota, and the University of Utah; have been conceiving novel computing architectures and programming models for a next generation advanced computer system. The final outcome will be the first commercial peta-scale super computer, the Ultraviolet, which can be deployed by 2010. Funding for this project is provided through DARPA’s high-productivity computing systems(HPCS) program, Phase I.

Reconfigurable computing

Sponsor: DoD

Principal Investigators: Tarek El-Ghazawi and Nikitas Alexandridis

The synergistic advances in high-performance computing systems and in reconfigurable computing, based on field programmable gate arrays (FPGAs), form the basis for a new paradigm shift in supercomputing, namely reconfigurable supercomputing. This can be achieved through hybrid systems of microprocessors as well as FPGA modules. Such systems inherently support both fine-grain and coarse-grain parallelism, and can tune their architectures to fit the needs of applications.

While our objective is to pursue the aforementioned concepts in general, the objectives of our current project are to:

  1. Accelerate the development of parallel reconfigurable computers based on the Starbridge and SRC architectures,
  2. Validate the concepts through useful application developments, testing and benchmarking with focus on the security area,
  3. Iintegrate the experience of the research community and views of industry, academia and government research scientists. Our application areas currently focus on cryptography applications.

IP Tools

Sponsor: DoD
Principal Investigators: Nikitas Alexandridis and Tarek El-Ghazawi

The emergence of intellectual property components has caused a redefinition of the field of embedded systems. Many modern embedded systems are defined as multi-chip modules (MCMs) or System on a Chip (SoC) designs composed of IP components from a variety of vendors. Engineers often spend a significant portion of time reviewing IP components for suitability to their respective embedded system design. This process considers a myriad of permutations before the final components are selected. This task seeks to answer the following questions:

Effective Use of Distributed Reconfigurable Computing Resources

While the number of reconfigurable computing resources available on computer networks has been growing rapidly in recent years, within an organization or across federated organizations, these systems are still expensive compared to commodity workstations. Therefore, it is important to try to maximize their utilization. This project establishes the middleware needed for monitoring, aggregating, and scheduling reconfigurable resources for shared use in a grid-computing style. In doing so, the team has investigated and extended Job Management Systems (JMSs) to recognize, monitor, and schedule reconfigurable computing resources over the network. A prototype using LSF has been established and successfully used.