US 11,676,322 B2
Graphics architecture including a neural network pipeline
Hugues Labbe, Granite Bay, CA (US); Darrel Palke, Portland, OR (US); Sherine Abdelhak, Beaverton, OR (US); Jill Boyce, Portland, OR (US); Varghese George, Folsom, CA (US); Scott Janus, Loomis, CA (US); Adam Lake, Portland, OR (US); Zhijun Lei, Hillsboro, OR (US); Zhengmin Li, Hillsboro, OR (US); Mike Macpherson, Portland, OR (US); Carl Marshall, Portland, OR (US); Selvakumar Panneer, Portland, OR (US); Prasoonkumar Surti, Folsom, CA (US); Karthik Veeramani, Hillsboro, OR (US); Deepak Vembar, Portland, OR (US); and Vallabhajosyula Srinivasa Somayazulu, Portland, OR (US)
Assigned to Intel Corporation, Santa Clara, CA (US)
Filed by Intel Corporation, Santa Clara, CA (US)
Filed on Oct. 13, 2021, as Appl. No. 17/500,631.
Application 17/500,631 is a continuation of application No. 16/537,140, filed on Aug. 9, 2019, granted, now 11,151,769.
Claims priority of provisional application 62/717,603, filed on Aug. 10, 2018.
Claims priority of provisional application 62/717,685, filed on Aug. 10, 2018.
Claims priority of provisional application 62/717,593, filed on Aug. 10, 2018.
Prior Publication US 2022/0058853 A1, Feb. 24, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 17/20 (2006.01); G06N 3/08 (2023.01); G06T 1/60 (2006.01); G06T 15/40 (2011.01); G06T 1/20 (2006.01); G06T 15/00 (2011.01)
CPC G06T 15/005 (2013.01) [G06N 3/08 (2013.01); G06T 1/20 (2013.01); G06T 1/60 (2013.01); G06T 15/40 (2013.01); G06T 17/20 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A graphics processor comprising:
a block of graphics cores; and
circuitry including a programmable neural network unit, the programmable neural network unit including one or more neural network hardware blocks, wherein a neural network hardware block includes circuitry to perform neural network operations and activation operations for a layer of a neural network, the programmable neural network unit addressable by cores within the block of graphics cores, the programmable neural network unit configured to determine visibility for a geometry culling operation via the neural network hardware block, at least one core within the block of graphics cores configured to:
provide geometry data for a scene to the neural network hardware block, the neural network block to perform neural network operations to detect an obstructed geometric object within the scene based on the geometry data;
receive a set of obstructed geometric objects for the scene from the neural network block; and
cull the set of obstructed geometric objects from the scene.