Systems and methods for measuring the effectiveness of a workload predictor on a mobile device
AssigneeQUALCOMM Incorporated
InventorsBrian J. Salsbery, Norman S. Gargash
Life of this patent
Estimating the term…
Abstract
Systems and methods for measuring the effectiveness of a workload predictor operative on a mobile device are disclosed. A load manager includes a workload predictor, a sensor, an error generator and a controller. The workload predictor generates an estimate of the workload on a processor core operative on the mobile device. The sensor generates a measure of the actual workload on the processor core. The error generator receives the estimate of the workload and the measure of the actual workload on the processor core and generates an error signal. The controller receives the error signal and determines the effectiveness of the workload predictor as a function of the error signal over time.
Claims16
A method for measuring an effectiveness of a workload predictor on a mobile device having a processor and a controller, the method comprising:
- generating by the processor a real-time measure of a workload on the mobile device;
- generating by the processor an estimate of the workload on the mobile device using a workload predictor;
- comparing by the processor the real-time measure of the workload to the estimate of the workload on the mobile device to generate an error signal;
- recording by the processor the error signal over time;
- generating by the processor a function of the error signal as a measure of the effectiveness of the workload predictor;
- applying by the processor the error signal to the controller, wherein the controller adjusts at least one operating parameter on the mobile device in response to the error signal;
- and
- using by the controller a measure of cross-correlation of the error signal to determine if a present operating state of the mobile device is similar to a known use case to select a previously optimized combination of a performance scaling algorithm and a parameter set associated with the known use case to manage the workload across multiple processing cores available on the mobile device.
A load manager for execution on a mobile device, comprising:
- a workload predictor configured to generate an estimate of the workload on a processor core operative on the mobile device;
- a sensor configured to generate a measure of the workload on the processor core operative on the mobile device;
- an error generator configured to receive the estimate of the workload and the measure of the workload and generate an error signal;
- and
- a controller arranged to receive the error signal and adjust operation of the processor core,
- wherein the controller adjusts at least one operating parameter on the mobile device in response to the error signal, and
- wherein the controller uses a measure of cross-correlation of the error signal to determine if a present operating state of the mobile device is similar to a known use case to select a previously optimized combination of a performance scaling algorithm and a parameter set associated with the known use case to manage the workload across multiple processing cores available on the mobile device.
A system for managing a load on a mobile device, the system comprising:
- means for measuring a workload of a processor core operative on the mobile device;
- means for generating an estimate of the workload of the processor core operative on the mobile device;
- means for generating an error signal responsive to the workload and the estimate of the workload;
- means for storing the error signal over time;
- means for associating a stored error signal over time with a present state of use of the mobile device;
- means for selectively adjusting at least one operating parameter on the mobile device in response to the stored error signal over time;
- and
- means for using a measure of cross-correlation of the error signal to determine if a present operating state of the mobile device is similar to a known use case to select a previously optimized combination of a performance scaling algorithm and a parameter set associated with the known use case to manage the workload across multiple processing cores available on the mobile device.
A non-transitory computer-readable medium stored thereon processor-executable instructions configured to cause a processor of a mobile device to perform operations comprising:
- generating a real-time measure of a workload on the mobile device;
- generating an estimate of the workload on the mobile device;
- comparing the real-time measure of the workload to the estimate of the workload on the mobile device to generate an error signal;
- recording the error signal over time;
- calculating a function of the error signal;
- determining an effectiveness of the workload predictor in response to the function of the error signal;
- applying the error signal to a mobile device controller that adjusts at least one operating parameter on the mobile device in response to the error signal;
- and
- using a measure of cross-correlation of the error signal to determine if a present operating state of the mobile device is similar to a known use case to select a previously optimized combination of a performance scaling algorithm and a parameter set associated with the known use case to manage the workload across multiple processing cores available on the mobile device.
Dependencies are inferred from the claim text.