发明名称 GAUSSIAN MIXTURE MODEL ACCELERATOR WITH DIRECT MEMORY ACCESS ENGINES CORRESPONDING TO INDIVIDUAL DATA STREAMS
摘要 Systems, apparatus and methods are described including operations for memory access via direct memory access engines, of a Gaussian Mixture Model Accelerator, corresponding to individual data streams.
申请公布号 US2016351195(A1) 申请公布日期 2016.12.01
申请号 US201514722489 申请日期 2015.05.27
申请人 INTEL CORPORATION 发明人 FALIK OHAD;DEISHER MICHAEL E;SCHOINAS IOANNIS YANNIS;CHANG JENNY;KHOR HAI MING
分类号 G10L15/28;G10L15/14;G10L15/02 主分类号 G10L15/28
代理机构 代理人
主权项 1. A computer-implemented method for scoring operations, comprising: setting up, via an execution core, a scoring operation associated with a plurality of vector data types on a plurality of direct memory access engines on a stream-by-stream basis where a single vector data stream associated with a corresponding single vector data type is set up for processing via a corresponding single direct memory access engine; consuming, via the execution core, vector data stored in a plurality of data buffers during the scoring operation, wherein the vector data is consumed in a stream-by-stream basis; transferring, between the data buffers and the direct memory access engines, an indication that a given buffer associated with a given stream is has a certain fullness, wherein the transfer between the data buffers and the direct memory access engines is on a stream-by-stream basis; and throttling, via the direct memory access engines, data requests so as to stop and/or start such data requests in response to the indication from the data buffers that a given buffer associated with a given stream has a certain fullness, wherein the throttling is on a stream-by-stream basis.
地址 Santa Clara CA US