Qualcomm Incorporated

Eduardus Antonius Metz

Colin Christopher Sharp

Navid Farazmand

David Rigel Garcia Garcia

Song Zhao

Qiao Shen

Andrew Evan Gruber

Maurice Franklin Ribble

Nigel Terence Poole

Zachary Aaron Pfeffer

A method of graphics processing comprising receiving, at a graphics processing unit (GPU), a command stream, the command stream including one or more commands to be performed by the GPU and at least one command stream hint, the at least one command stream hint providing a characterization of a workload of the command stream, performing, by the GPU, a power management process based on the at least one command stream hint prior to executing the command stream, and executing, by the GPU, the command stream.

Characterizing GPU workloads and power management using command stream hinting

A device for processing data includes a processing unit configured to predict an execution time of a compute kernel on a secondary processing unit and, based on the predicted execution time, make a power management decision for the secondary processing unit.

Prediction-based power management strategy for GPU compute workloads

This disclosure is directed to deferred preemption techniques for scheduling graphics processing unit (GPU) command streams for execution on a GPU. A host CPU is described that is configured to control a GPU to perform deferred-preemption scheduling. For example, a host CPU may select one or more locations in a GPU command stream as being one or more locations at which preemption is allowed to occur in response to receiving a preemption notification, and may place one or more tokens in the GPU command stream based on the selected one or more locations. The tokens may indicate to the GPU that preemption is allowed to occur at the selected one or more locations. This disclosure further describes a GPU configured to preempt execution of a GPU command stream based on one or more tokens placed in a GPU command stream.

Deferred preemption techniques for scheduling graphics processing unit command streams

Techniques are described in which a processor determines how busy a graphics processing unit (GPU) was in rendering a current fame based on a time interval between when the GPU completed the rendering of the current frame and when the GPU completed the rendering of a previous frame that immediately precedes the current frame. The processor sets a clock rate of the GPU based on the determination of how busy the GPU was in rendering the current frame.

Frame based clock rate adjustment for processing unit

This disclosure proposes techniques for demand paging for an IO device (e.g., a GPU) that utilize pre-fetch and pre-back notification event signaling to reduce latency associated with demand paging. Page faults are limited by performing the demand paging operations prior to the IO device actually requesting unbacked memory.

GPU memory buffer pre-fetch and pre-back signaling to avoid page-fault

In general, aspects of this disclosure describe example techniques for efficient storage of data of various data types for graphics processing. In some examples, a processing unit may assign first and second contiguous range of addresses for a first and second data type, respectively. The processing unit may store at least one of graphics data of the first or second data type or addresses of the graphics data of the first or second data type within blocks whose addresses are within the first and second contiguous range of addresses, respectively. The processing unit may store, in cache lines of a cache, the graphics data of the first data type, and the graphics data of the second data type.

Growing community of inventors

Unionville, Canada

Eduardus Antonius Metz