Title :
Hardware/software codesign for video compression using the EXECUBE processing array
Author :
Sheliga, Michael ; Kogge, Peter M. ; Sha, Edwin Hsing-Mean
Author_Institution :
Dept. of Comput. Sci. & Eng., Notre Dame Univ., IN, USA
Abstract :
Real time image processing is an important area of study that requires the computational power of parallel computers. In this paper MPEG video compression is used as a benchmark application to study and improve the performance of the processor-in-memory (PIM) EXECUBE array which has been proposed as a solution for the PetaFlop project. Since parallel computers naturally lend themselves to the display of such pictures, the EXECUBE array of parallel processors is a logical choice. It is shown that the array is nor able to display video in real time due to computational intensive algorithms within MPEG, as well as an internal communication bottleneck. As such, hardware/software codesign is used to overcome these problems while minimizing design time. Through the use of specialized hardware, the timing constraints may be met. Results are shown which illustrate the efficiency of the system as well as the savings achieved
Keywords :
computational complexity; data compression; image processing; parallel algorithms; parallel architectures; video coding; EXECUBE processing array; MPEG video compression; PetaFlop project; benchmark; computational intensive algorithms; computational power; design time; hardware/software codesign; parallel computers; processor-in-memory EXECUBE array; real time image processing; timing constraints; video compression; Computer displays; Concurrent computing; Hardware; Image processing; Optical sensors; Power engineering computing; Power generation; Sensor systems; Transform coding; Video compression;
Conference_Titel :
Aerospace and Electronics Conference, 1997. NAECON 1997., Proceedings of the IEEE 1997 National
Conference_Location :
Dayton, OH
Print_ISBN :
0-7803-3725-5
DOI :
10.1109/NAECON.1997.618091