Supercomputing Research Group
In order to manage complex and big data and perform high performance computing we have proposed a Supercomputing Research Group (SRG). The Super Computing Group (SRG) can work on research, industry and defence projects, including big data processing, signal processing, and low power computing. The SRGtargest different high performance architectures for super computing. A generic block diagram of target design is shown in Figure.
The proposed architecture uses FPGAs based accelerators, DSP, GPU and RISC multi-core system. A High level synthesis tool is used to program FPGA accelerators for specific applications. These dedicated accelerators have low footprint and low power consumption and provide high performance. SIMD architecture uses DSPs and the CUDA Tesla, Fermi, etc. architectures," the GPU Computing Modules. Low power ARM and high performance x86 multi-core processor systems are also used in the Super-Cluster system which allows system to execute most of the generic C/C++ applications. The proposed system is ideal for seismic processing, biochemistry simulations, weather and climate modeling, signal processing, computational finance, CAE, CFD and data analytics.
The main objectives of the SRGare:
Provide a low-power low-cost super computing resources to work out of academic, research and defence project.
Generate a team of collaborations for faculty, post-graduate students, specialists, scientists and technical resources working on high performance computing projects and initiatives.
Propose and Develop low-power low-cost embedded and heterogeneous system architectures for compute intensive applications.
The SCRG targets three types of Super-Cluster architectures which are the Low Power Low-Cost Super-Cluster, the Transitional Performance Super-Cluster and the High-Performance Super-Cluster. The SRGis categorized into the following subsections: Target Applications, Programming Models and Toolkits, Supported Operating System, and Performacne Cost and Power.
UCERD Gathering Intellectuals