An Object-Oriented Software Framework for Execution of Real-Time, Parallel Algorithms

  • Authors:
  • J. Brent Spears;Brett Gossage

  • Affiliations:
  • -;-

  • Venue:
  • ICCS '01 Proceedings of the International Conference on Computational Sciences-Part I
  • Year:
  • 2001

Quantified Score

Hi-index 0.00

Visualization

Abstract

Infrared (IR) scene projection systems that employ microresisters can have significant, non-uniform spatial variation in their output. This non-uniformity causes unwanted artifacts in the projected scene sufficient to prevent accurate sensor testing. To compensate for this nonuniformity, high-speed digital hardware is used to apply nonuniformity correction (NUC) to the images. However, the hardware is very costly and must be custom-built for each projector. With high performance computers, NUC can be implemented in software a t a fraction of the hardware cost while still meeting the real-time requirements. The purpose of this paper is to present object-oriented frameworks, implemented in C++, for executing NUC algorithms in parallel in real-time. The frameworks provide abstractions for multi-threading, parallel processing, shared memory, frame-based scheduling, variable interrupt sources, and scheduling disciplines. Results for NUC algorithms on an 8-processor SGI Onyx-2 will also be presented.