Browse Prior Art Database

Run-Time Algorithms for Distributed Processing Synchronization and Address Calculations

IP.com Disclosure Number: IPCOM000038967D
Original Publication Date: 1987-Mar-01
Included in the Prior Art Database: 2005-Feb-01
Document File: 2 page(s) / 14K

Publishing Venue

IBM

Related People

DeMuth, GL: AUTHOR

Abstract

A program generation technique was described for distributed signal processing architectures in the article by G. L. DeMuth, "High Order Notation and Automated Program Generation for Real Time Signal Processing," Proceedings of the IEEE Conference on Acoustics, Speech and Signal Processing, pp. 683-686, May 1982. The following is a disclosure of run time algorithms which are resident in each distributed processing element, to implement scheduling and address calculation for each of the distributed processing elements, in a coordinated manner with the other distributed processing elements in the network. The automated program generator, described in the above-cited article, builds a load file for initial program load of a signal processor.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 1 of 2

Run-Time Algorithms for Distributed Processing Synchronization and Address Calculations

A program generation technique was described for distributed signal processing architectures in the article by G. L. DeMuth, "High Order Notation and Automated Program Generation for Real Time Signal Processing," Proceedings of the IEEE Conference on Acoustics, Speech and Signal Processing, pp. 683-686, May 1982. The following is a disclosure of run time algorithms which are resident in each distributed processing element, to implement scheduling and address calculation for each of the distributed processing elements, in a coordinated manner with the other distributed processing elements in the network. The automated program generator, described in the above-cited article, builds a load file for initial program load of a signal processor. The load file permits each processing element to operate in a distributed processing environment with a minimum of external control. Synchronization of multiple processing elements in the distributed processing network and the scheduling of tasks within individual processing elements are accomplished using an interval count where the interval count is the input data batch number. Resident in each distributed processing element is a run-time algorithm which extracts the required cyclic counts from the interval counter to control the dispatching of tasks and the advancing of address pointers within subdivided data sets at each distributed processing element. Three parameters control the timing of the execution of each program statement. The three parameters are the number of channel subgroups, the segment processing periodicity, and a reference time slot interval number. Scheduling Rules In accordance with the invention, the schedules are generated subject to scheduling rules. The total number of channels are subdivided into subgroups of equal size, where one subgroup is processed every 2I intervals or (2I-1) subgroups are processed every 2I intervals; I is a positive integer (with zero permissible in the first case). In other words, one subgroup from a given segment can be processed every 1, 2, 4, 8, 16,... intervals or three subgroups can be processed every four intervals, seven subgroups e...