Dynamic activity-creating data-driven computer architecture

• US 4,644,461 A
• Filed: 04/29/1983
• Issued: 02/17/1987
• Est. Priority Date: 04/29/1983
• Status: Expired due to Fees

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Jennings, Glenn A.
• Primary Examiner(s): Shaw, Gareth D.
• Assistant Examiner(s): Dorsey, Daniel K.
• Application Number:
  US06/490,154
• Time in Patent Office: 1,390 Days
• Field of Search: 364/130, 364/131, 364/200 MS File, 364/900 MS File, 364/716
• US Class Current: 718/106
• CPC Class Codes: G06F 15/17375 One dimensional, e.g. linear array, ring
• CPC Class Codes: G06F 9/4494 data driven

Abstract

A computer architecture wherein data inputs causes the dynamic creation of appropriate activities employing stored functions as necessary to accomplish the desired end result for the data. The architecture employs a large scale multi-processing environment for parallel computation. A fast forward propagating queue structure and improved interfacing crossbar are employed.

Claims

1. In a computer architecture of the Petri Net type wherein a plurality of pre-established sub-functions, defining a total function, process portions of the common input data asynchronously to generate a single combined output reflecting the total function as applied to input data, an improvement comprising:

(a) a plurality of data processors;

(b) means for subdividing said total function when said total function is beyond the capability of an individual one of said data processors.

(c) function library means coupled to said data processors for storing each unique sub-function in a storage area as a replicator which can be copied and used by said data processors as needed on a dynamic basis;

(d) data queue means for recirculating data to be processed by said data processors continuously past said data processors until data is removed from said data queue means by one of said data processors, and for supplying said data to selected data processors;

(e) activity queue means for storing a plurality of sets of instructions for operating on selected corresponding data by said data processors, and for recirculating said instructions past said data processors until one of said instructions is taken from said queue by one of said data processors, said activity queue means including means for supplying to all of said data processors, instruction codes, for causing said data processors, (1) to remove particular labeled data from the data queue means, (2) to perform a computational function on said data (3) to introduce output labelled data into said data queue means, (4) to retrieve from the function library stored generalized instruction sequences, and (5) to introduce additional instruction codes into the activity queue means;

(f) means for associating one of said sets of instructions from said activity queue with an idle one of said processors;

(g) means for applying selected data from said data queue to said one processor, said selected data corresponding to said instructions selected from said activity queue; and

(h) said means for subdividing including means for applying subdivided data to the data queue, means for applying sub-function instructions to the activity queue such that further processing by a plurality of idle ones of said data processors can be accomplished and means for recombining the separately processed sub-functions whereby the processing of said total function is accomplished.

2. A computer architecture as defined in claim 1 wherein said queue means includes a plurality stages of latch means for holding the data.

3. A computer architecture as defined in claim 1 including crossbar circuit means for accessing the function library.

4. A computer architecture as defined in claim 1 wherein auxiliary storage nodes are provided, and wherein said function library resides therein.

5. A computer architecture as defined in claim 1 wherein the data in said data queue is labelled data and forms part of the means for applying selected data corresponding to instructions selected by said data processor from said activity queue.