US 12,141,606 B2
Cascading of graph streaming processors
Venkata Ganapathi Puppala, Hyderabad (IN); Sarvendra Govindammagari, Hyderabad (IN); Lokesh Agarwal, Hyderabad (IN); and Satyaki Koneru, Folsom, CA (US)
Assigned to Blaize, Inc., El Dorado Hills, CA (US)
Filed by Blaize, Inc., El Dorado Hills, CA (US)
Filed on Oct. 16, 2023, as Appl. No. 18/380,299.
Application 17/833,981 is a division of application No. 16/282,464, filed on Feb. 22, 2019, granted, now 11,379,262, issued on Jul. 5, 2022.
Application 18/380,299 is a continuation of application No. 17/833,981, filed on Jun. 7, 2022, granted, now 11,822,960.
Application 16/282,464 is a continuation in part of application No. 16/270,766, filed on Feb. 8, 2019, granted, now 11,150,961, issued on Oct. 19, 2021.
Application 16/270,766 is a continuation in part of application No. 15/164,848, filed on May 25, 2016, granted, now 10,437,637, issued on Oct. 8, 2019.
Claims priority of provisional application 62/166,507, filed on May 26, 2015.
Prior Publication US 2024/0036921 A1, Feb. 1, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 1/20 (2006.01); G06F 8/41 (2018.01); G06F 9/38 (2018.01); G06F 9/48 (2006.01)
CPC G06F 9/4881 (2013.01) [G06F 8/45 (2013.01); G06F 9/3851 (2013.01); G06F 9/3885 (2013.01); G06T 1/20 (2013.01)] 10 Claims
OG exemplary drawing
 
1. A method of graph stream processing, comprising:
scheduling, by a thread manager of each of a plurality of graph streaming processors, a plurality of threads, wherein
the scheduling includes;
interpreting, by an input command buffer parser of each stage of a plurality of stages of the thread manager, commands within a corresponding input command buffer located between the stage and a previous stage; and
generating a plurality of threads based on the interpreted commands;
wherein one or more shared command buffers are located between each of the plurality of graph streaming processors, wherein the one or more shared command buffers each includes a buffer address, a write pointer, and a read pointer;
further comprising;
writing, by a graph streaming processor of the plurality of graph streaming processors, commands to a shared command buffer of the one or more shared command buffers as indicated by the write pointer;
reading, by the graph streaming processor of the plurality of graph streaming processors, commands from the shared command buffer and indicated by the read pointer, comprising interpreting, by an input command buffer parser of a first stage of a thread manager of the graph streaming processor, commands within the shared command buffer;
managing the write pointer and the read pointer to avoid overwriting unused commands of the one or more shared command buffers.