The Community for Technology Leaders

Corrections to "Time Dependent Processing in a Parallel Pipeline Architecture"

John
Berk
Ken
Kenneth
David

Pages: p. 241


Introduction

In the article cited above, which appeared in the November/December 2007 issue (vol. 13, no. 6) of this journal, a portion of the text that appeared on p. 1379 was missing. The following is the text in its entirety:

shows that the algorithm objects do not directly communicate with each other, this allows two key operations:

  • The executives can insert additional time related metadata into the information stream. This data is used to provide the control mechanism that satisfies our time requests without affecting the algorithms themselves.
  • All traffic whether it be data or metadata, passing to or from an algorithm, is in fact passed from executive to executive via information objects before entering the algorithm. This allows an executive to change the data that is produced by one algorithm before passing it on to another. The executive uses this feature to collect data from multiple time steps before passing it on to a filter requiring temporal data.

4.3 Temporal Pipeline Extensions

We have enhanced the capabilities of the executives to recognize and act upon new information keys which have been introduced to support temporal data. The following key/value pairs have been defined in our pipeline.

REFERENCES



FULL ARTICLE
52 ms
(Ver 3.x)