Reducing the number of processors elements in systolic arrays for matrix multiplication

  • Authors:
  • Dragan Randjelovic

  • Affiliations:
  • Academy of Criminalistic and Police Studies, Belgrade, Serbia

  • Venue:
  • SEPADS'12/EDUCATION'12 Proceedings of the 11th WSEAS international conference on Software Engineering, Parallel and Distributed Systems, and proceedings of the 9th WSEAS international conference on Engineering Education
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Author is discussing a problems of determining parameters suitable systolic arrays for implementation regular 3-nested loop algorithms. Author shows that if the characteristics of so called adaptable algorithms to the projection direction are used we have the best results. This characteristics can be space (number of processor elements, chip area, input-output elements,...), time (flow period of processor, summary time of algorithm realization,...) or one combination of those and the subject of interesting in this paper is one from them - number of processor elements. In this paper is given transformation matrix, which maps the given index space in another index space and reduces the number of processing elements in the systolic array. Obtained results in this paper are illustrated on the example of matrix-matrix multiplication algorithm as one typical adaptable algorithm and that on the examples of both possibly synthesized systolic arrays of them i.e. planar and non-planar type systolic arrays.