- 专利标题: Microthreading for accelerated deep learning
-
申请号: US16603950申请日: 2018-04-17
-
公开(公告)号: US11475282B2公开(公告)日: 2022-10-18
- 发明人: Sean Lie , Michael Morrison , Michael Edwin James , Gary R. Lauterbach , Srikanth Arekapudi
- 申请人: Cerebras Systems Inc.
- 申请人地址: US CA Los Altos
- 专利权人: Cerebras Systems Inc.
- 当前专利权人: Cerebras Systems Inc.
- 当前专利权人地址: US CA Los Altos
- 代理机构: PatentVentures
- 代理商 Bennett Smith
- 国际申请: PCT/IB2018/052640 WO 20180417
- 国际公布: WO2018/193361 WO 20181025
- 主分类号: G06N3/063
- IPC分类号: G06N3/063 ; G06N3/08 ; G06N3/04 ; G06F13/00 ; G06F9/30 ; H04L49/00 ; G06F9/38 ; H04L12/54 ; G06F5/06 ; G06F13/40 ; H04L49/506 ; G06F30/392
摘要:
Techniques in advanced deep learning provide improvements in one or more of accuracy, performance, and energy efficiency. An array of compute elements and routers performs flow-based computations on wavelets of data. Some instructions are performed in iterations, such as one iteration per element of a fabric vector or FIFO. When sources for an iteration of an instruction are unavailable, and/or there is insufficient space to store results of the iteration, indicators associated with operands of the instruction are checked to determine whether other work can be performed. In some scenarios, other work cannot be performed and processing stalls. Alternatively, information about the instruction is saved, the other work is performed, and sometime after the sources become available and/or sufficient space to store the results becomes available, the iteration is performed using the saved information.