US 11,941,512 B2
Neural network engine utilizing a serial bus
Dmitry Obukhov, San Jose, CA (US); Anshuman Singh, San Jose, CA (US); and Anuj Awasthi, Fremont, CA (US)
Assigned to Western Digital Technologies, Inc., San Jose, CA (US)
Filed by Western Digital Technologies, Inc., San Jose, CA (US)
Filed on Jun. 26, 2019, as Appl. No. 16/453,735.
Prior Publication US 2020/0410326 A1, Dec. 31, 2020
Int. Cl. G06N 3/04 (2023.01); G06F 13/38 (2006.01); G06N 3/063 (2023.01); G06N 3/08 (2023.01)
CPC G06N 3/063 (2013.01) [G06F 13/385 (2013.01); G06N 3/04 (2013.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system of performing neural network computations, the system comprising:
an input interface configured to receive input data for neural network processing; and
a common serial bus configured to serially transmit portions of the input data and a first source identification data to a plurality of nodes of a neural network, wherein in response to each portion of the input data being serially transmitted from the common serial bus, a count is incremented;
wherein each node of the plurality of nodes is configured to:
in response to the first source identification data matching one or more addresses stored at the node, store each portion of the input data in a respective register of the node; and
in response to the count meeting a threshold, for each node that stored the input data:
transmit the stored input data to a transfer function of the node;
receive an output of the transfer function;
add a second source identification data to the output of the transfer function; and
transmit the output of the transfer function with the second source identification data to each node of the plurality of nodes via the common serial bus.