A Mobile High Performance Computing System for Deployment in the Field
Abstract
Major Goals: The project specifically addresses the problem of porting High-Performance Computing (HPC) applications directly to FPGA-based architectures, and the desire to automate the process. Accordingly, the overall goals of the proposed research are to: 1. Develop a comprehensive floating-point library of essential functions for military applications. 2. Demonstrate order-of-magnitude speedup of reconfigurable computing applications over executing them on a traditional microprocessor-based system. 3. Train students in the use of Remote And Reconfigurable Environment (RARE) tools for rapid military application development and to stimulate their interests in careers in DoD. Thrust 1: RARE Architecture Enhancement and Module Development - Major Goals. In previous research, implementing several diverse applications on Field Programmable Gate Arrays (FPGAs) we achieved 2 orders-of-magnitude speedup over software implementation of the same applications executing on a conventional processor. The reason we achieved 2 orders-of-magnitude speedup is because we implemented a simple customized General Purpose Processor (GPP) on an FPGA. This customized processor is simply a traditional sequential GPP with a custom floating point ALU designed for each application. Additionally, the processor can simultaneously read/write several single precision floating point values each cycle from multiple memories while producing multiple results each cycle. The increase in demand for computing power has led to the development of heterogeneous systems which are typically composed of conventional processors coupled with a high-speed coprocessor, i.e. the Graphics Processing Unit (GPU). The GPU is gaining popularity as a parallel computing platform for accelerating compute intensive applications. A typical GPU architecture consists of multiple controllers, multiple cache memories, DRAM memory and large number of ALUs to perform complex computations in parallel.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 09, 2022
- Accession Number
- AD1196992
Entities
People
- Christopher Doss
- Clay Jr Gloster
Organizations
- North Carolina Agricultural and Technical State University