US 12,229,581 B2
Virtualization and multi-tenancy support in graphics processors
Rajesh Sankaran, Portland, OR (US); Bret Toll, Hillsboro, OR (US); William Rash, Saratoga, CA (US); Subramaniam Maiyuran, Gold River, CA (US); Gang Chen, Milpitas, CA (US); and Varghese George, Folsom, CA (US)
Assigned to Intel Corporation, Santa Clara, CA (US)
Filed by INTEL CORPORATION, Santa Clara, CA (US)
Filed on Aug. 29, 2023, as Appl. No. 18/239,489.
Application 18/239,489 is a continuation of application No. 16/456,300, filed on Jun. 28, 2019, granted, now 11,748,130.
Prior Publication US 2024/0061700 A1, Feb. 22, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 9/455 (2018.01); G06F 12/1009 (2016.01); G06T 1/20 (2006.01)
CPC G06F 9/45558 (2013.01) [G06F 12/1009 (2013.01); G06T 1/20 (2013.01); G06F 2009/4557 (2013.01); G06F 2009/45583 (2013.01); G06F 2009/45591 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a plurality of graphics processing apparatuses; and
an interconnect to couple the plurality of graphics processing apparatuses,
wherein a first graphics processing apparatus within the plurality of graphics processing apparatuses is to:
receive a first request from a virtual machine monitor (VMM) to provision a first subcluster of graphics processing apparatuses within the plurality of graphics processing apparatuses, the first subcluster including a plurality of graphics processing engines,
provision the interconnect to route data within the first subcluster of graphics processing apparatuses,
provision a subset of resources on the first graphics processing apparatus for the first subcluster of graphics processing apparatuses responsive to the first request from the VMM, and
responsive to a second request to migrate a workload from the first subcluster of graphics processing apparatuses to a second subcluster of graphics processing apparatuses within the plurality of graphics processing apparatuses,
identify a portion of memory of the first graphics processing apparatus that was written to while another portion of memory was transferred to the second subcluster,
determine the portion of the memory is below a threshold, and
send a message to the VMM indicating the portion of the memory and further indicating that the first subcluster is ready to be migrated.