Method and apparatus for optimizing power consumption in a multiprocessor environment

• US 7,694,160 B2
• Filed: 08/31/2006
• Issued: 04/06/2010
• Est. Priority Date: 08/31/2006
• Status: Active Grant

• Current Assignee: ATI Technologies ULC (Advanced Micro Devices, Inc.)
• Original Assignee: ATI Technologies ULC (Advanced Micro Devices, Inc.)
• Inventors: Kwan, Wilson, Esliger, James
• Primary Examiner(s): Cao; Chun
• Application Number:
  US11/469,182
• Publication Number:
  US 20080059921A1
• Time in Patent Office: 1,314 Days
• Field of Search: 718/105, 718/100, 713/320, 713/324
• US Class Current: 713/320
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• CPC Class Codes: G06F 1/329 by task scheduling
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A method and apparatus for reducing net power consumption in a computer system includes identifying a plurality of processing states operable to execute a task. A processing state and current drain pattern is selected that is most power efficient. A selected processing state may include one or more processing elements of the computer system such as one or more processors or accelerators and indicates the manner in which one or more portions of the received task may be distributed among the processing elements of the computer system. The current drain pattern selected may be a constant current drain pattern or a pulsed current drain pattern and may be selected to optimize power consumption when executing the task among the one or more processing elements.

Claims

1. A method for reducing net power consumption in a computer system, the method comprising:

identifying a plurality of processing states each operable to execute a task, wherein each processing state indicates how the task is to be distributed among at least one processing element of the computer system during execution; and

selecting a processing state and a current drain pattern that is most power efficient for execution of the task and executing the task using the selected processing state and current drain pattern.

2. The method of claim 1, wherein selecting a processing state and a current drain pattern that is most power efficient comprises selecting a processing state and a current drain pattern that reduces net power consumption by:

for each of the plurality of processing states, calculating one or more of;

at least one constant current drain parameter and at least one pulsed current drain parameter; and

selecting the processing state and the current drain pattern based on the processing state having a desirable at least one constant current drain parameter or a desirable at least one pulsed current drain parameter.

3. The method of claim 2, wherein selecting the processing state and the current drain pattern is further based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computing system.

4. The method of claim 1, wherein selecting a processing state and a current drain pattern that is most power efficient comprises selecting a processing state and a current drain pattern that drains the least amount of power from a power source during execution of the task by:

selecting the processing state based on at least one of;

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computing system;

for the selected processing state, calculating one or more of;

the at least one constant current drain parameter and the at least one pulsed current drain parameter;

selecting a constant current drain pattern if the at least one constant current drain parameter is desirable; and

selecting a pulsed current drain pattern if the at least one pulsed current drain parameter is desirable.

5. A method for reducing net power consumption in a computer system, the method comprising:

selecting a processing state operable to execute a task, wherein the processing state indicates how the task is to be distributed among at least one processing element of the computer system during execution;

selecting a current drain pattern to draw current from a power source based on one or more of;

at least one constant current drain parameter and at least one pulsed current drain parameter; and

executing the task using the selected processing state and the selected current drain pattern.

6. The method of claim 5, wherein selecting the processing state operable to execute the task comprises:

identifying one or more available processing elements in the computer system each capable of executing at least one portion of the task, wherein each processing element comprises at least one of;

a processor and an accelerator;

identifying at least one combination of the one or more available processing elements wherein each combination is capable of executing the task; and

selecting the processing state from the identified at least one combination.

7. The method of claim 6, wherein identifying the one or more available processing elements in the computer system each capable of executing the at least one portion of the task comprises:

identifying one or more processing elements present in the computer system; and

associating each portion of the task with at least one of the identified one or more processing elements based on at least one of;

functional capabilities of each of the identified one or more processing elements; and

a number of instructions per second necessary for execution of the at least one portion of the task and a number of instructions per second each processing element is capable of executing.

8. The method of claim 5, wherein selecting the processing state operable to execute the task comprises is based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computer system.

9. The method of claim 5, wherein selecting the current drain pattern to draw current from the power source based on one or more of:

at least one constant current drain parameter and at least one pulsed current drain parameter comprises;

for the selected processing state, calculating one or more of;

the at least one constant current drain parameter and the at least one pulsed current drain parameter;

selecting a constant current drain pattern if the at least one constant current drain parameter is desirable; and

selecting a pulsed current drain pattern if the at least one pulsed current drain parameter is desirable.

10. The method of claim 5, wherein selecting the processing state operable to execute the task and wherein selecting the current drain pattern to draw current from the power source based on one or more of:

at least one constant current drain parameter and at least one pulsed current drain parameter comprises;

identifying at least one combination of one or more available processing elements in the computer system wherein each combination is capable of executing the task;

for each combination, calculating one or more of;

at least one combination constant current drain parameter and at least one combination pulsed current drain parameter; and

selecting the processing state and the current drain pattern based on the combination having a desirable at least one combination constant current drain parameter or a desirable at least one combination pulsed drain parameter.

11. The method of claim 10, wherein selecting the processing state and the current drain pattern is further based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computer system.

12. The method of claim 5, wherein selecting a current drain pattern to draw current from a power source based one or more of:

the at least one constant current drain parameter and the at least one pulsed current drain parameter comprises at least one of;

determining, for at least one processing state, an operating constant current value representing an amount of current the computer system would drain if running using a constant current drain pattern;

determining, for at least one processing state, an operating pulsed current value representing an amount of current the computer system would drain if running using a pulsed current drain pattern; and

determining, for at least one processing state, an operating relaxation width value representing a length of time between each pulse if the computer system was running using the pulsed current drain pattern.

13. The method of claim 12, wherein the amount of current the computer system would drain if running using a pulsed current drain pattern includes the amount of current drained for each pulse in the pulsed current drain pattern and the amount of background current drained in between each pulse in the pulsed current drain pattern.

14. The method of claim 12, farther comprising making at least one of:

a comparison between the operating constant current value and a determined threshold constant current value representing the current level at which the pulsed current drain pattern is more efficient than the constant current drain pattern for execution of the task;

a comparison between the operating pulsed current value and a determined maximum pulsed current value representing the maximum amount of current the power source delivers if the computer system uses a pulsed current drain pattern;

a comparison between the operating relaxation width value and a determined minimum relaxation width value representing a minimum length of time between pulses if the computer system uses a pulsed current drain pattern; and

a comparison between the operating relaxation width value and a determined previous processing state operating relaxation width value representing the operating relaxation width value for a preceding processing state.

15. The method of claim 14, wherein:

a constant current drain pattern is selected when at least one of;

the operating constant current value is less than the determined threshold constant current value;

the operating pulsed current value is greater than the determined maximum pulsed constant current value; and

the operating relaxation width value is less than or equal to the determined minimum relaxation width value and greater than or equal to the determined previous processing state operating relaxation width value; and

a pulsed current drain pattern is selected when;

the operating constant current value is greater than or equal to the determined threshold constant current value;

the operating pulsed current value is less than or equal to the determined maximum pulsed current value; and

at least one of;

the operating relaxation width value is greater than the determined minimum relaxation width value; and

the operating relaxation width value is less than the determined previous processing state operating relaxation width value.

16. The method of claim 5, wherein the power source is at least one battery and wherein an operating time is maximized on a single battery charge of the at least one battery.

17. The method of claim 5, wherein when the selected processing state includes at least two processing elements, the method further comprises at least one of:

supplying at least a portion of a multimedia accelerator with an overall slower clock than when the selected processing state includes a single processing element; and

supplying at least a portion of the multimedia accelerator with a reduced voltage than when the selected processing state includes a single processing element.

18. A computer system comprising one or more processors, a power source, at least one processing element, and at least one memory having executable instructions stored therein, such that when executed by the one or more processors, the executable instructions cause the one or more processors to:

select a processing state operable to execute a task, wherein the processing state indicates how the task is to be distributed among the at least one processing element during execution;

select a current drain pattern to draw current from the power source based on one or more of;

at least one constant current drain parameter and at least one pulsed current drain parameter; and

wherein the computer system is configured to execute the task using the selected processing state and the selected current drain pattern, thereby reducing power consumption in the computer system.

19. The computer system of claim 18, wherein the executable instructions that cause the one or more processors to select the processing state operable to execute the task further cause the one or more processors to:

identify one or more available processing elements in the computer system each capable of executing at least one portion of the task, wherein each processing element comprises at least one of;

a processor and an accelerator;

identify at least one combination of the one or more available processing elements wherein each combination is capable of executing the task; and

select the processing state from the identified at least one combination.

20. The computer system of claim 18, wherein the executable instructions that cause the one or more processors to select the processing state operable to execute the task further cause the one or more processors to select the processing state based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

design constraints of the computer system.

21. The computer system of claim 18, wherein the executable instructions that cause the one or more processors to select the current drain pattern to draw current from the power source based on one or more of:

at least one constant current drain parameter and at least one pulsed current drain parameter further cause the one or more processors to;

for the selected processing state, calculate one or more of;

the at least one constant current drain parameter and the at least one pulsed current drain parameter;

select a constant current drain pattern if the at least one constant current drain parameter is desirable; and

select a pulsed current drain pattern if the at least one pulsed current drain parameter is desirable.

22. The computer system of claim 18, wherein the executable instructions that cause the one or more processors to select the processing state operable to execute the task and to select the current drain pattern to draw current from the power source based on one or more of:

at least one constant current drain parameter and at least one pulsed current drain parameter further cause the one or more processors to;

identify at least one combination of one or more available processing elements in the computer system wherein each combination is capable of executing the task;

for each combination, calculate one or more of;

at least one combination constant current drain parameter and at least one combination pulsed current drain parameter; and

select the processing state and the current drain pattern based on the combination having a desirable at least one combination constant current drain parameter or a desirable at least one combination pulsed drain parameter.

23. The computer system of claim 22, wherein the executable instructions that cause the one or more processors to select the processing state and current drain pattern further cause the one or more processors to select the processing state and the current drain pattern based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computer system.

24. The computer system of claim 18, wherein the executable instructions that cause the one or more processors to select a current drain pattern to draw current from a power source based one or more of:

the at least one constant current drain parameter and the at least one pulsed current drain parameter further cause the one or more processors to determine, for at least one processing state, at least one of;

an operating constant current value representing an amount of current the computer system would drain using a constant current drain pattern;

an operating pulsed current value representing an amount of current the computer system would drain using a pulsed current drain pattern; and

an operating relaxation width value representing a length of time between each pulse if the computer system used the pulsed current drain pattern.

25. The computer system of claim 24, wherein the executable instructions that cause the one or more processors to select a current drain pattern to draw current from a power source based one or more of:

at least one constant current drain parameter and at least one pulsed current drain parameter farther cause the one or more processors to make at least one of;

a comparison between the operating constant current value and a determined threshold constant current value representing the current level at which the pulsed current drain pattern is more power efficient than the constant current drain pattern for execution of the task;

a comparison between the operating pulsed current value and a determined maximum pulsed current value representing the maximum amount of current the power source delivers if the computer system uses a pulsed current drain pattern;

a comparison between the operating relaxation width value and a determined minimum relaxation width value representing a minimum length of time between pulses if the computer system uses a pulsed current drain pattern; and

a comparison between the operating relaxation width value and a determined previous processing state operating relaxation width value representing the operating relaxation width value for a preceding processing state.

26. The computer system of claim 25, wherein the executable instructions that cause the one or more processors to select a current drain pattern to draw current from a power source based one or more of:

at least one constant current drain parameter and at least one pulsed current drain parameter further cause the one or more processors to select;

a constant current drain pattern when at least one of;

the operating constant current value is less than the determined threshold constant current value;

the operating pulsed current value is greater than the determined maximum pulsed constant current value; and

the operating relaxation width value is less than or equal to the determined minimum relaxation width value and greater than or equal to the determined previous processing state operating relaxation width value; and

a pulsed current drain pattern when;

the operating constant current value is greater than or equal to the determined threshold constant current value;

the operating pulsed current value is less than or equal to the determined maximum pulsed current value; and

at least one of;

the operating relaxation width value is greater than the determined minimum relaxation width value; and

the operating relaxation width value is less than the determined previous processing state operating relaxation width value.

27. The computer system of claim 18, wherein when the selected processing state includes at least two processing elements, the computer system is operative to supply at least one portion of a multimedia accelerator with at least one of:

an overall slower clock than when the selected processing state includes a single processing element; and

a reduced voltage than when the selected processing state includes a single processing element.

28. A computer system comprising:

a load balancer operative to identify a plurality of processing states each operable to execute a task, wherein each processing state indicates how the task is to be distributed among at least one processing element of the computer system during execution; and

wherein the load balancer and a current drain selector are operative to select a processing state and a current drain pattern that is most power efficient for execution of the task.

29. The computer system of claim 28, wherein:

the most power efficient processing state and current drain pattern reduces net power consumption in the computer system;

the current drain selector is further operative to calculate;

for each of the plurality of processing states, one or more of;

at least one constant current drain parameter and at least one pulsed current drain parameter; and

the load balancer is further operative to select the processing state and the current drain selector is further operative to select the current drain pattern based on the processing state having a desirable at least one constant current drain parameter or a desirable at least one pulsed current drain parameter.

30. The computer system of claim 29, wherein the load balancer is further operative to select the processing state based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computing system.

31. The computer system of claim 28, wherein:

the load balancer is further operative to select the processing state based on at least one of;

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computing system;

the current drain selector is further operative, for the selected processing state, to calculate one or more of;

the at least one constant current drain parameter and the at least one pulsed current drain parameter, and

the current drain selector if further operative to select a pulsed current drain pattern if the at least one pulsed current drain parameter is desirable.

32. A computer system comprising:

a load balancer operative to select a processing state operable to execute a task, wherein the processing state indicates how the task is to be distributed among the at least one processing element during execution;

a current drain pattern selector operative to select a current drain pattern to draw current from the power source based on one or more of;

at least one constant current drain parameter and at least one pulsed current drain parameter; and

wherein the computer system is configured to execute the task using the selected processing state and the selected current drain pattern, thereby reducing power consumption in the computer system.

33. The computer system of claim 32, wherein the load balancer is further operative to:

identify one or more available processing elements in the computer system each capable of executing at least one portion of the task, wherein each processing element comprises at least one of;

a processor and an accelerator;

identify at least one combination of the one or more available processing elements wherein each combination is capable of executing the task; and

select the processing state from the identified at least one combination.

34. The computer system of claim 32, wherein the load balancer is further operative to select the processing state based on at least one of:

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computer system.

35. The computer system of claim 32, wherein the current drain pattern selector is further operative to:

for the selected processing state, calculate one or more of;

the at least one constant current drain parameter and the at least one pulsed current drain parameter;

select a constant current drain pattern if the at least one constant current drain parameter is desirable; and

select a pulsed current drain pattern if the at least one pulsed current drain parameter is desirable.

36. The computer system of claim 32, wherein:

the load balancer is further operative to identify at least one combination of one or more available processing elements in the computer system wherein each combination is capable of executing the task;

the current drain pattern selector is further operative to, for each combination, calculate one or more of;

at least one combination constant current drain parameter and at least one combination pulsed current drain parameter;

and wherein at least one of the load balancer and the current drain pattern selector is operative to select the processing state and the current drain pattern based on the combination having a desirable at least one combination constant current drain parameter or a desirable at least one combination pulsed drain parameter.

37. The computer system of claim 36, wherein the at least one of:

the load balancer and the current drain pattern selector is further operative to select the processing state based on at least one of;

an amount of shared memory available on the computer system;

one or more coupling requirements of the task; and

one or more design constraints of the computer system.

38. The computer system of claim 32, wherein the current drain pattern selector is further operative to determine, for at least one processing state, at least one of:

an operating constant current value representing an amount of current the computer system would drain using a constant current drain pattern;

an operating pulsed current value representing an amount of current the computer system would drain using a pulsed current drain pattern; and

an operating relaxation width value representing a length of time between each pulse if the computer system used the pulsed current drain pattern.

39. The computer system of claim 38, wherein the current drain pattern selector is further operative to make at least one of:

a comparison between the operating constant current value and a determined threshold constant current value representing the current level at which the pulsed current drain pattern is more power efficient than the constant current drain pattern for execution of the task;

a comparison between the operating pulsed current value and a determined maximum pulsed current value representing the maximum amount of current the power source delivers if the computer system used a pulsed current drain pattern;

a comparison between the operating relaxation width value and a determined minimum relaxation width value representing a minimum length of time between pulses if the computer system uses a pulsed current drain pattern; and

a comparison between the operating relaxation width value and a determined previous processing state operating relaxation width value representing the operating relaxation width value for a preceding processing state.

40. The computer system of claim 39, wherein the current drain pattern selector is further operative to select:

a constant current drain pattern when at least one of;

the operating constant current value is less than the determined threshold constant current value;

the operating pulsed current value is greater than the determined maximum pulsed constant current value; and

the operating relaxation width value is less than or equal to the determined minimum relaxation width value and greater than or equal to the determined previous processing state operating relaxation width value; and

a pulsed current drain pattern when;

the operating constant current value is greater than or equal to the determined threshold constant current value;

the operating pulsed current value is less than or equal to the determined maximum pulsed current value; and

at least one of;

the operating relaxation width value is greater than the determined minimum relaxation width value; and

the operating relaxation width value is less than the determined previous processing state operating relaxation width value.

41. The computer system of claim 32, wherein when the selected processing state includes at least two processing elements, the computer system is operative to supply at least one portion of a multimedia accelerator with at least one of:

an overall slower clock than when the selected processing state includes a single processing element; and

a reduced voltage than when the selected processing state includes a single processing element.