| 摘要 |
PROBLEM TO BE SOLVED: To improve utilization efficiency of a processor to reduce device scale by increasing granularity to be processed in parallel as much as possible to reduce the number of required processors, and further reducing idle processors within a range that processing is completed within a time limit. SOLUTION: A parallel processor successively receives data divided by processing unit, and processes an operation to the data using a plurality of processors. The parallel processor is provided with: a scheduler 4 which calculates the time required for receiving, calculating, and transmitting an operation result, respectively, using information shown in data input, to assign data to the processors for operation according to the operation result; and a data transfer means 6 for transferring the data to the assigned processors. The scheduler 4 assigns the data to one processor when the time from the start of data reception to the end of transmission of the operation result does not exceed a predetermined time, and assigns subsequent data to other processors. COPYRIGHT: (C)2011,JPO&INPIT |
