US 11,782,991 B2
Accelerated large-scale similarity calculation
Lin Ma, San Jose, CA (US); and Nathan Wiegand, San Jose, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Jun. 13, 2022, as Appl. No. 17/839,116.
Application 17/839,116 is a division of application No. 16/344,450, granted, now 11,379,535, previously published as PCT/US2018/030486, filed on May 1, 2018.
Prior Publication US 2022/0309101 A1, Sep. 29, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 7/00 (2006.01); G06F 16/906 (2019.01); G06F 12/0802 (2016.01)
CPC G06F 16/906 (2019.01) [G06F 12/0802 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving first data items at a first processing unit;
generating, by the first processing unit and from at least the first data items, data structures, each data structure representing a respective dataset corresponding to at least a portion of the first data items;
storing the data structures in a memory of a second processing unit;
receiving, at the first processing unit, a query from an injest component communicatively coupled to the first processing unit;
generating, by the first processing unit and from the received query, a query data structure including data representing a set of second data items;
processing concurrently, by the second processing unit, the query data structure against two or more of the data structures stored in the memory of the second processing unit, the processing comprising: for each of the two or more of the data structures, computing, by the second processing unit, a correlation score for the query data structure and the data structure according to a correlation algorithm; and
returning results corresponding to the query determined based on the computed correlation scores from the second processing unit to the first processing unit.