US 11,080,213 B2 · Apparatus and method for dynamic provisioning, quality of service, and scheduling in a graphics processor

Abstract

An apparatus and method for dynamic provisioning and traffic control on a memory fabric. For example, one embodiment of an apparatus comprises: a graphics processing unit (GPU) comprising a plurality of graphics processing resources; slice configuration hardware logic to logically subdivide the graphics processing resources into a plurality of slices; and slice allocation hardware logic to allocate a designated set of slices to each virtual machine (VM) of a plurality of VMs running in a virtualized execution environment; and a plurality of queues associated with each VM at different levels of a memory interconnection fabric, the queues for a first VM to store memory traffic for that VM at the different levels of the memory interconnection fabric; arbitration hardware logic coupled to the plurality of queues and distributed across the different levels of the memory interconnection fabric, the arbitration hardware logic to cause memory traffic to be blocked from one or more upstream queues of the first VM upon detecting that a downstream queue associated with the first VM is full or at a specified threshold.

Claims17

Dependent claims

An apparatus comprising:

a graphics processing unit (GPU) comprising a plurality of graphics processing resources;
and
scheduler hardware logic coupled to the plurality of graphics processing resources, wherein the scheduler hardware logic is to identify a priority of each task within a plurality of tasks provided by a plurality of virtual machines (VMs), and allocate the plurality of graphics processing resources to execute the plurality of tasks based on priorities of the plurality of tasks, wherein the plurality of tasks provided by the plurality of VMs shares a cache system, wherein a first task that has a first priority higher than a second priority of a second task is to be allocated with more cache lines than the second task, wherein the allocation of the plurality of graphics processing resources based on the priorities of the plurality of tasks comprises submitting the tasks to a plurality of priority-based task queues, and wherein traffic to be blocked from queueing into one or more upstream queues of a first VM upon detecting that a downstream queue associated with the first VM is full or at a specified threshold.

A non-transitory machine-readable storage medium that provides instructions that, when executed by a machine, are capable of causing the machine to perform operations comprising:

identifying a priority of each task within a plurality of tasks provided by a plurality of virtual machines (VMs);
and
allocating a plurality of graphics processing resources to execute the plurality of tasks based on priorities of the plurality of tasks, wherein the plurality of tasks provided by the plurality of VMs shares a cache system, wherein a first task that has a first priority higher than a second priority of a second task is to be allocated with more cache lines than the second task, wherein the allocation of the plurality of graphics processing resources based on the priorities of the plurality of tasks comprises submitting the tasks to a plurality of priority-based task queues, and wherein traffic to be blocked from queueing into one or more upstream queues of a first VM upon detecting that a downstream queue associated with the first VM is full or at a specified threshold.

Dependencies are inferred from the claim text.

Apparatus and method for dynamic provisioning, quality of service, and scheduling in a graphics processor

Life of this patent

Apparatus and method for dynamic provisioning, quality of service, and scheduling in a graphics processor

Life of this patent

Abstract

Claims17