Patents Assigned to FuriosaAI Co.
  • Publication number: 20220245436
    Abstract: A device for processing convolution operations includes: a processor that executes, in a neural network, a convolution operation on input data in a form of width×height×input channel and on a filter in a form of K×K×input channel or K×K to correspond to a form of the input data, K being an integer greater than or equal to one, and that generates output data in a form of width×height×output channel; and a reader that sequentially reads, from a memory storing the input data, a data group having more pieces of data than unit data throughput of an operator, and provides the data group to the operator to reuse at least one piece of data constituting the data group in the convolution operation. The processor executes, by using one or more operators identical to the operator, the convolution operation multiple times based on the unit data throughput.
    Type: Application
    Filed: June 2, 2020
    Publication date: August 4, 2022
    Applicant: FuriosaAI Co.
    Inventors: Han Joon Kim, Young Geun Choi, Byung Chul Hong, Min Jae Kim, Bon Cheol Gu
  • Patent number: 10698730
    Abstract: A processing unit for neural network processing includes: an instruction memory that stores tasks including one or more instructions; a data memory that stores data related to the tasks; a data flow processor that determines whether the data has been prepared for the tasks and notifies a control flow processor that preparations for the tasks have been finished in order of finished data preparation; the control flow processor that controls execution of the tasks in order of notification from the data flow processor; and a functional processor that performs computations resulting from the one or more instructions of the tasks controlled for execution by the control flow processor.
    Type: Grant
    Filed: September 7, 2018
    Date of Patent: June 30, 2020
    Assignee: FuriosaAI Co.
    Inventors: Hanjoon Kim, Boncheol Gu, Jeehoon Kang, Changman Lee
  • Publication number: 20190303198
    Abstract: A processing unit for neural network processing includes: an instruction memory that stores tasks including one or more instructions; a data memory that stores data related to the tasks; a data flow processor that determines whether the data has been prepared for the tasks and notifies a control flow processor that preparations for the tasks have been finished in order of finished data preparation; the control flow processor that controls execution of the tasks in order of notification from the data flow processor; and a functional processor that performs computations resulting from the one or more instructions of the tasks controlled for execution by the control flow processor.
    Type: Application
    Filed: September 7, 2018
    Publication date: October 3, 2019
    Applicant: FuriosaAI Co.
    Inventors: Hanjoon Kim, Boncheol Gu, Jeehoon Kang, Changman Lee