Dynamically controlling power based on work-loop performance

Patent metadata

Field	Value
Publication number	US-9632566-B2
Application number	US-201213629365-A
Country	US
Kind code	B2
Filing date	Sep 27, 2012
Priority date	Sep 27, 2012
Publication date	Apr 25, 2017
Grant date	Apr 25, 2017

Abstract

The present embodiments provide a system that dynamically controls power consumption in a computing device. During operation, the system measures the performance of the computing device while executing a work-loop. Next, the system determines a derived completion time for the work-loop based on the measured performance. (For example, the derived completion time can be an expected completion time, a maximum completion time, or more generally a completion time distribution.) The system then determines a deadline-proximity for the work-loop based on a comparison between the derived completion time and a deadline for the work-loop. (For example, the deadline-proximity can be an expected deadline-proximity, a minimum deadline-proximity, or more generally a deadline-proximity distribution.) Finally, the system controls the power consumption of the computing device based on the determined deadline-proximity for the work-loop.

First claim

What is claimed is:

1. A method for dynamically controlling power consumption of a computing system, comprising:
obtaining a first set of indications, each indication acquired through an application program interface (API) and signaling a start of a work-loop being performed by a computing system;
obtaining a second set of indications, each indication acquired through the API and signaling an end of the corresponding work-loop performed by the computing system;
determining a statistical distribution for completion times for the work-loop based on the first set and second set of indications;
determining a derived completion time for the work-loop based on the statistical distribution;
determining a deadline-proximity for the work-loop based on a comparison between the derived completion time and a deadline for completion of the work-loop; and
controlling a power consumption of the computing system based on the determined deadline-proximity for the work-loop.

2. The method of claim 1 , wherein the computing system includes a plurality of processors, and wherein the work-loop is performed by two or more processors of the plurality of processors.

3. The method of claim 1 , further comprising:
determining the deadline for completion of the work-loop based on obtaining the deadline from an application associated with the work-loop.

4. The method of claim 1 , further comprising:
determining the deadline for completion of the work-loop based on a comparison between a frequency at which the work-loop is started and the determined completion.

5. The method of claim 1 , wherein:
acquiring each of the first set of indications comprises making a call to one or more performance-measurement functions through the API during performance of the work-loop, and
acquiring each of the second set of indications comprises making a call to the one or more performance-measurement functions through the API during the corresponding performance of the work-loop.

6. The method of claim 5 , wherein each of the calls for the first set and second set of indications is made using an identifier associated with the work-loop.

7. The method of claim 5 , wherein determining the statistical distribution for the completion times for the work-loop comprises:
determining multiple measurements of the completion time for the work-loop by the one or more performance-measurement functions based on the first set and second set of indications, and
using the multiple measurements to determine the statistical distribution for the completion times for the work-loop.

8. The method of claim 1 , wherein the derived completion time is:
an expected completion time;
a maximum completion time; or
a completion time distribution.

9. The method of claim 1 , wherein the deadline-proximity is:
an expected deadline-proximity;
a minimum deadline-proximity; or
a deadline-proximity distribution.

10. The method of claim, 1 wherein controlling the power consumption of the computing system based on the determined deadline-proximity for the work-loop includes:
adjusting a power-performance state (P state) of a processor of the computing system, wherein the P state determines at least one of a clock speed of the processor and an operating voltage of the processor.

11. The method of claim 1 , wherein controlling the power consumption of the computing system based on the determined deadline-proximity for the work-loop includes:
adjusting a CPU-power state (C state) for a processor of the computing system, wherein the C state determines whether the processor is in: an operating state, a halted state; or a sleep state.

12. The method of claim 1 , wherein controlling the power consumption of the computing system based on the determined deadline-proximity for the work-loop includes:
scheduling multiple threads to run on one or more processors of the computing system.

13. The method of claim 10 , wherein scheduling the multiple threads involves:
scheduling the multiple threads to run in parallel on the same processor;
scheduling the multiple threads to run serially on the same processor; or
scheduling the multiple threads to run in parallel on different processors.

14. The method of claim 1 , wherein the work-loop is executed in sequence on multiple processors of the computing system.

15. The method of claim 14 , wherein the work-loop includes at least one of:
a process executed on at least one of the multiple processors; and
a thread executed on at least one of the multiple processors.

16. The method of claim 1 , wherein the computing system includes multiple hardware devices, wherein controlling the power consumption of the computing system based on the determined deadline-proximity for the work-loop includes selectively adjusting power consumption of one of the multiple hardware devices, and wherein the one of the multiple hardware devices is at least one of the following:
a hardware device that is capable of performing smaller performance adjustments than all other ones of the multiple hardware devices;
a hardware device that is more idle than all other ones of the multiple hardware devices as measured with respect to the deadline-proximity; and
a hardware device that provides a larger power savings for a given amount of performance loss than all other ones of the multiple hardware devices.

17. An apparatus including one or more integrated circuits that dynamically controls power consumption of a computing system, comprising:
a measurement mechanism embodied in the one or more integrated circuits, the measurement mechanism configured to:
obtain a first set of indications, each indication acquired through an application programming interface (API) and signaling a start of a work-loop being performed by a computing system, and
obtain a second set of indications, each indication acquired through the API and signaling an end of the corresponding work-loop performed by the computing system;

a computation mechanism embodied in the one or more integrated circuits, the computation mechanism configured to:
determine a statistical distribution for completion times for the work-loop based on the first set and second set of indications;
determine a derived completion time for the work-loop based on the statistical distribution; and
determine a deadline-proximity for the work-loop based on a comparison between the derived completion time and a deadline for completion of the work-loop; and

a power-controlling mechanism embodied in the one or more integrated circuits, the power-controlling mechanism configured to:
control a power consumption of the computing system based the determined deadline-proximity for the work-loop.

18. The apparatus of claim 17 , wherein the computing system includes a plurality of processors, and wherein the work-loop is performed by two or more processors of the plurality of processors.

19. The apparatus of claim 17 , wherein the apparatus is further configured to determine the deadline for completion of the work-loop based on obtaining the deadline from an application associated with the work-loop.

20. The apparatus of claim 17 , wherein the apparatus is further configured to determine the deadline for completion of the work-loop based on a comparison between a frequency at which the work-loop is started and the determined completion time.

21. The apparatus of claim 17 , wherein
acquiring each of the first set of indications comprises making a call to one or more performance-measurement func

Assignees

• Apple Inc

Inventors

• Culbert Daniel J

Classifications

• Y02B60/144

  Cross-Sectional Technologies · mapped topic

• G06F1/329 Primary

  by task scheduling · CPC title

• G06F9/4893

  taking into account power or heat criteria (power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206) · CPC title

• Y02D10/00

  Energy efficient computing, e.g. low power processors, power management or thermal management · CPC title