Apple Inc.

Eric O Sunalp

Gokhan Avkarogullari

Richard Warren Schreyer

Tatsuya Iwamoto

Patrick O Heynen

Kutty Banerjee

Andrew M Sowerby

Jason P Jane

Alexander K Kan

Domenico Troiano

Alp Yucebilgin

Maximilian Christ

Lionel Lemarie

Sebastian Schaefer

Ohad Frenkel

Daniel I Omachi

Dustin J Greene

A Resource Dependency Viewer for graphics processing unit (GPU) execution information is disclosed. The Resource Dependency Viewer provides profiling/debugging information concurrently with information about execution flow, resource utilization, execution statistics, and orphaned resources, among other things. A user-interactive graph ('dependency graph') may be provided via a graphical user interface to allow interactive analysis of code executed on a GPU (e.g., graphics or compute code). Resource utilization and execution flow of encoders may be identified by analyzing contents of a GPU workload representative of a GPU execution trace to generate the dependency graph. Information about dependencies and execution statistics may be further analyzed using heuristics to identify potential problem areas. The dependency graph may include visual indicators of these problem areas. Results oriented navigation from the dependency graph and other visual panes provide debugging navigation techniques different than a debugging interface configured to 'step' through code or provide static analysis.

Graphics processing unit resource dependency viewer

Described is a system that processes texture assets for an application to be distributed across a range of devices while still leveraging device-specific hardware capabilities. In one embodiment, the system processes various criteria and determines specialized texture asset variants for a set of devices. These criteria may include various selected options including an interpretation selection for interpreting a source image for a texture, and device trait options such as device type, memory, and software support. In addition, the system may create mipmaps based on attributes of a set of target devices. Accordingly, a set of specialized texture assets may be created and distributed across a suite of devices.

Providing variants of texture assets for an asset catalog

Embodiments are directed toward systems and methods for scheduling resources of a graphics processing unit that determine, for a number of applications having commands to be issued to the GPU, a static priority level and a dynamic priority level of each application, work iteratively across static priority levels until a resource budget of the GPU is consumed, and starting with a highest static priority identify the applications in a present static priority level, assign a processing budget of the GPU to each of the applications in the present static priority level according to their dynamic priority levels, and admit to a queue commands from the applications in the present static priority level according to their processing budgets, and release the queue to the GPU.

Starvation free scheduling of prioritized workloads on the GPU

Techniques are provided for managing the power consumption and performance of a processing device. Power consumption and utilization ratios for a processing device may be continuously measured. The measured power consumption and utilization ratios may be compared to target power consumption and utilization ratios to adjust an operating frequency of the processing device. In one implementation a power controller may take the target and measured power consumption as inputs to generate a power output and a utilization controller may take the target and measured utilization ratios as inputs to generate a utilization output. The lower of the power output and the utilization output may be selected and used to adjust the operating frequency of the processing device. The power and utilization controllers may implement a proportional-integral control scheme.

Processor power and performance manager

Growing community of inventors

Los Gatos, CA, United States of America

Eric O Sunalp