USRE36954EExpiredUtility

SIMD system having logic units arranged in stages of tree structure and operation of stages controlled through respective control registers

29
Assignee: FUJITSU LTDPriority: Sep 19, 1988Filed: Jul 19, 1995Granted: Nov 14, 2000
Est. expirySep 19, 2008(expired)· nominal 20-yr term from priority
G06F 15/8023G06F 15/8007
29
PatentIndex Score
4
Cited by
21
References
8
Claims

Abstract

In a parallel computer system using a SIMD method constituted by a controller and a plurality of processor elements, each of the processor elements has a storage unit to store data to be processed, the controller controls operation of the processor elements, and the parallel computer system performs processing of the data based on a calculation control signal transmitted from the controller. The parallel computer system further a data collection unit connected between the processor elements and the controller for receiving output data from the processor elements, performing a predetermined calculation, and outputting calculated data to the controller; and a calculation control unit connected between the data collection unit and the controller for transmitting the calculation control signal from the controller to the data calculation unit to make it possible to perform the predetermined calculation in the data collection circuit.

Claims

exact text as granted — not AI-modified
What we claim is: 
     
       1. A parallel computer system using a single instruction stream multiple data stream (SIMD) method, said parallel computer system having a controller and a plurality of processor elements, each of the processor elements having storage means for storing data to be processed and outputting output data processed by the processor elements, the controller controls operation of the processor elements, and said parallel computer system performing processing of the data based on a calculation control signal and a synchronization signal transmitted from the controller, said parallel computer system comprising: data collection means, connected between the processor elements and the controller and arranged in a .[.binary.]. tree configuration having stages, for receiving the output data from the processor elements responsive to the synchronization signal generated by and received from the controller, for performing a predetermined calculation based on the stages in the .[.binary.]. tree configuration, and for outputting calculated data to the controller; and   calculation control means, connected between said data collection means and the controller and arranged in series corresponding to each of the stages, for transmitting the calculation control signal from the controller to said data collection means based upon a pipe-line method to perform the predetermined calculation in said data collection means,   wherein said data collection means comprises a plurality of gathering logic units connected to each other in the .[.binary.]. tree configuration having the stages, first gathering logic units of a first stage of the stages receives the output data from each of the processor elements and outputs first calculation data, second gathering logic units of a second stage of the stages receives the first calculation data obtained from the first stage, and the first calculation data obtained from the second stage is output to a final gathering logic unit of a first stage of the stages as second calculation data, the final calculation data obtained from the final gathering logic unit of the final stage responsive to said second calculation data is output to the controller, and   wherein said calculation control means comprises a plurality of control registers each corresponding to one of the stages, each of the plurality of control registers connected in series to each other by the pipe-line method, and each of the plurality of control registers sequentially outputting the calculation control signal to each of the gathering logic units in the corresponding stage.   
     
     
       2. A parallel computer system as claimed in claim 1, wherein the calculation control signal is one of an AND calculation signal, an OR calculation signal, and a MAX/MIN calculation signal. 
     
     
       3. A parallel computer system as claimed in claim 2, wherein each of said gathering logic units comprises: an OR calculation means for performing a logic OR calculation on the output data from each of the processor element;   an AND calculation means for performing a logic AND calculation on the output data from each of the processor elements;   a MAX/MIN/ADD calculation means for obtaining one of a maximum value, a minimum value and an added value of the output data from each of the processor elements; and   selector means for selecting one of said OR calculation means, AND calculation means and MAX/MIN/ADD calculation means, and for performing calculations in response to the calculation control signal.   
     
     
       4. A parallel computer system as claimed in claim 1, wherein each of said gathering logic units comprises: an OR calculation means for performing a logic OR calculation on the output data from each of the processor elements;   an AND calculation means for performing a logic AND calculation on the output data from each of the processor elements;   a MAX/MIN/ADD calculation means for obtaining one of a maximum value, a minimum value, and an added value of the output data from each of the processor elements; and   selector means for selecting one of said OR calculation means, AND calculation means and MAX/MIN/ADD calculation means, and for performing calculations in response to the calculation control signal.   
     
     
       5. A parallel computer system using a single instruction stream multiple data stream (SIMD) method, including processor elements and a controller generating control and synchronization signals, said parallel computer system comprising: data collection means, connected between the processor elements and the controller and arranged in a .[.binary.]. tree structure having stages, for receiving output data from the processor elements responsive to the synchronization signal received by the processor elements from the controller, for performing a predetermined calculation responsive to the control signal based on the stages in the .[.binary.]. tree structure, and for outputting calculated data to the controller; and   calculation control means, connected between said data collection means and the controller and arranged in series corresponding to each of the stages, for transmitting the control signal received from the controller to said data collection means based upon a pipe-line process,   wherein said data collection means comprises a plurality of gathering logic units connected to each other in the .[.binary.]. tree structure having the stages, first gathering logic units of a first stage of the stages receives the output data from each of the processor elements and outputs first calculation data, second gathering logic units of a second stage of the stages receives the first calculation data obtained from the first stage, and the first calculation data obtained from the second stage is output to a final gathering logic unit of a final stage of the stages as second calculation data, and final calculation data obtained from the final gathering logic unit of the final stage responsive to said second calculation data is output to the controller, and   wherein said calculation control means comprises a plurality of control registers each corresponding to one of the stages, each of the plurality of control registers connected in series to each other by the pipe-line method, and each of the plurality of control register sequentially outputting the control signal to each of the gather logic units in the corresponding stage.   
     
     
       6. A parallel computer system as claimed in claim 5, wherein the calculation control signal is one of an AND calculation signal, an OR calculation signal, and a MAX/MIN calculation signal. 
     
     
       7. A parallel computer system as claimed in claim 5, wherein each of said gathering logic units comprises: an OR calculation means for performing a logic OR calculation on the output data from each of the processor elements;   an AND calculation means for performing a logic AND calculation on the output data from each of the processor elements;   a MAX/MIN/ADD calculation means for obtaining one of a maximum value, a minimum value and an added value of the output data from each of the processor elements; and   selector means for selecting one of said OR calculation means, AND calculations in response to the calculation and for performing calculations in response to the calculation control signal. .Iadd.   
     
     
       8.  A method of calculation processing, comprising: processing input data in parallel in plural processing units to produce parallel output data;   combining the parallel output data using a tree combination of tree stages to produce calculated output data; and   controlling said processing and controlling said tree stage combining at each stage to produce a predetermined calculated output data result..Iaddend..Iadd.9. A method of calculation processing as recited in claim 8, wherein said combining of the parallel output data is performed logically..Iaddend..Iadd.10. A method of calculation processing as recited in claim 8, wherein said controlling of said processing and said combining   
     
     
        is performed in pipeline stages..Iaddend..Iadd.11.  An apparatus, comprising: parallel processing units processing input data in parallel to produce parallel output data;   data combination units connected to said parallel processing units and combining the parallel output data using a tree combination having combining control at each stage to produce calculated output data; and   a controller connected to said data combination units and said parallel processing units and controlling said processing and said combining to produce a predetermined calculated output data result..Iaddend.

Cited by (0)

No later patents cite this yet.

References (0)

No backward citations on record.