DARPA Software Defined Hardware

Abstract

To benchmark the capabilities of novel electronic processing devices being developed by DARPA, six common algorithms, or kernels, were selected to be used for comparison purposes. The kernels include Fast Fourier Transforms, Dense and Sparse Matrix Multiplication, Two Dimensional Convolution, Dijkstra's Algorithm, and the Auction Algorithm. The processing performance of these kernels on new devices, in terms of speed and power requirements, are to be compared against the performance on modern Application Specific Integrated Circuits (ASICs). Since ASICs for these specific kernels are not equally available, Field Programmable Gate Arrays (FPGAs) are used as proxies. For each kernel, an operations per watt value is derived when implemented on FPGAs, from which the metric for ASIC performance is estimated.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 2021
Accession Number
AD1145204

Entities

People

  • Jonathan P. Skeans

Organizations

  • University of Dayton

Tags

Communities of Interest

  • Advanced Electronics
  • Energy and Power Technologies
  • Space

DTIC Thesaurus Topics

  • Air Force
  • Air Force Facilities
  • Air Force Research Laboratories
  • Algorithms
  • Application-Specific Integrated Circuits
  • Central Processing Units
  • Circuits
  • Computers
  • Digital Signal Processing
  • Field Effect Transistors
  • Field Programmable Gate Arrays
  • Governments
  • Graphics Processing Unit
  • Integrated Circuits
  • Networks
  • Parallel Processing
  • Software-Defined Hardware
  • Sparse Matrix
  • Standards
  • Two Dimensional
  • Very Large Scale Integration

Readers

  • Integrated Circuit Design and Technology.
  • Parallel and Distributed Computing.

Technology Areas

  • Microelectronics