US 11,704,545 B2
Integrated circuit chip apparatus
Shaoli Liu, Beijing (CN); Xinkai Song, Beijing (CN); Bingrui Wang, Beijing (CN); Yao Zhang, Beijing (CN); and Shuai Hu, Beijing (CN)
Assigned to CAMBRICON TECHNOLOGIES CORPORATION LIMITED, Beijing (CN)
Filed by CAMBRICON TECHNOLOGIES CORPORATION LIMITED, Beijing (CN)
Filed on Dec. 19, 2019, as Appl. No. 16/721,888.
Application 16/721,888 is a continuation of application No. PCT/CN2019/073453, filed on Jan. 28, 2019.
Claims priority of application No. 201711343642.1 (CN), filed on Dec. 14, 2017; application No. 201711346333.X (CN), filed on Dec. 14, 2017; application No. 201711347310.0 (CN), filed on Dec. 14, 2017; application No. 201711347406.7 (CN), filed on Dec. 14, 2017; application No. 201711347407.1 (CN), filed on Dec. 14, 2017; application No. 201711347408.6 (CN), filed on Dec. 14, 2017; and application No. 201711347767.1 (CN), filed on Dec. 14, 2017.
Prior Publication US 2020/0234107 A1, Jul. 23, 2020
Int. Cl. G06N 3/063 (2023.01); G06N 3/04 (2023.01); G06F 7/483 (2006.01); G06F 7/544 (2006.01); G06N 3/06 (2006.01); G06N 3/08 (2023.01); G06F 17/15 (2006.01); G06F 17/16 (2006.01); H01L 25/065 (2023.01)
CPC G06N 3/063 (2013.01) [G06F 7/483 (2013.01); G06F 7/5443 (2013.01); G06F 17/153 (2013.01); G06F 17/16 (2013.01); G06N 3/04 (2013.01); G06N 3/06 (2013.01); G06N 3/08 (2013.01); H01L 25/065 (2013.01); G06F 2207/4824 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for performing a neural network computation using an integrated circuit chip apparatus, wherein the method comprises:
receiving an operation instruction;
parsing the operation instruction to obtain input data and weight data for a first computation;
determining a first complexity of the first computation according to the input data, the weight data, and characteristics of the first computation, wherein the first complexity is a computation complexity of the first computation;
determining a first data type in which the first computation is to be performed, wherein the first data type is selected between a floating point data type and a fixed point data type, based on a comparison of the first complexity with a preset threshold; and
performing the first computation on the input data and the weight data in the first data type.