Structure and method to optimize computational efficiency in low-power environments

• US 8,055,925 B2
• Filed: 07/18/2007
• Issued: 11/08/2011
• Est. Priority Date: 07/18/2007
• Status: Active Grant

• Current Assignee: Marvell Asia Pte Limited (Marvell Technology Group Limited)
• Original Assignee: International Business Machines Corporation
• Inventors: Williams, Keith, Ventrone, Sebastian T., Ogilvie, Clarence R., Goodnow, Kenneth J., Woodruff, Charles S.
• Primary Examiner(s): BUTLER, DENNIS
• Application Number:
  US11/779,432
• Publication Number:
  US 20090024862A1
• Time in Patent Office: 1,574 Days
• Field of Search: 713300-340
• US Class Current: 713/340
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode

Abstract

A method and structure to optimize computational efficiency in a low-power environment. The method includes determining an optimal point for maximizing computational efficiency in a low-power environment, and selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point. The structure includes a plurality of processing units, a load manager controlling selective parallel operation of at least one processing unit of the plurality of processing units, and an unregulated power source.

Claims

1. A structure, comprising:

a plurality of processing units; and

a load manager controlling selective parallel operation of at least one processing unit of the plurality of processing units based upon a determined optimal point for maximizing computational efficiency, wherein;

the optimal point is determined by sensing or measuring directly available sunlight, and knowing the available sunlight, or present operating condition, an applicable surface plot of operations per second as a function of frequency and number of operating units is accessed such that the optimal point for maximizing computational efficiency in low-power environments is determined from the surface plot for a current operating condition.

2. The structure of claim 1, wherein the load manager provides a clock signal to the at least one processing unit.

3. The structure of claim 2, wherein the load manager further comprises a frequency scaler providing the clock signal.

4. The structure of claim 2, wherein the load manager further comprises a ring oscillator frequency controller providing the clock signal.

5. The structure of claim 1, wherein the load manager comprises:

a plurality of switches through which power is selectively provided to the plurality of processing units; and

a load manager processor controlling the selective parallel operation of the at least one processing unit through selective operation of the plurality of switches.

6. The structure of claim 1, wherein the load manager comprises:

a current A/D converter; and

a voltage A/D converter.

7. The structure of claim 1, further comprising an intensity sensor for sensing an intensity of an environmental condition.

8. The structure of claim 1, wherein the applicable surface plot of operations per second as a function of frequency and number of operating units is accessed from a database of surface plots.

9. A method to optimize computational efficiency in a low-power environment, comprising:

determining a present power available;

selecting from data, obtained from an analysis applied voltage versus current, applicable data based upon the present power available;

determining an optimal point for maximizing computational efficiency in a low-power environment from the applicable data; and

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point.

10. The method of claim 9, further comprising controlling a frequency of a clock signal transmitted to the at least one processing unit in accordance with the determined optimal point.

11. A method to optimize computational efficiency in a low-power environment, comprising:

determining an optimal point for maximizing computational efficiency in a low-power environment;

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point; and

determining a present power available, wherein the determining the present power available comprises sensing the present power available with a sensing element, and the sensing element is a solar intensity sensor.

12. A method to optimize computational efficiency in a low-power environment, comprising:

determining an optimal point for maximizing computational efficiency in a low-power environment;

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point;

determining a present power available;

selecting from data, obtained from an analysis of operations/second versus frequency of operation and number of active processing units, applicable data based upon the determined present power available; and

determining an optimal point for maximizing computational efficiency from the applicable data.

13. The method of claim 12, wherein the data is pre-determined and stored in a memory.

14. A method to optimize computational efficiency in a low-power environment, comprising:

determining an optimal point for maximizing computational efficiency in a low-power environment;

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point;

determining a present power available;

selecting from data, obtained from an analysis applied voltage versus current, applicable data based upon the present power available; and

determining an optimal point for maximizing computational efficiency from the applicable data,

wherein the determining a present power available comprises;

measuring a present voltage and a present current; and

calculating the present power available from the present voltage and the present current.

15. The method of claim 14, wherein the data is pre-determined and stored in a memory.

16. A method to optimize computational efficiency in a low-power environment, comprising:

determining an optimal point for maximizing computational efficiency in a low-power environment, the determining comprising inferring a current power condition;

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point;

assuming a minimum power available;

activating all processing units;

outputting a minimum frequency as a present frequency to the active processing units;

saving the present frequency and a number which is representative of a quantity of active processing units;

designating an operations/second, determined according to the present frequency and the quantity of active processing units, as a maximum operations/second;

iteratively increasing the present frequency to a new present frequency where the voltage is at about a minimum for the new present frequency;

designating an operations/second, determined according to the new present frequency and the quantity of active processing units, as a current operations/second;

if the current operations/second is greater than the maximum operations/second, setting the current operations/second as a new maximum operations/second, saving the new present frequency and the number which is representative of the quantity of active processing units, and decrementing the quantity of active units; and

if the current operations/second is not greater than the maximum operations/second, restoring a last saved present frequency and a last saved number representative of the quantity of active units.

17. A method to optimize computational efficiency in a low-power environment, comprising:

determining an optimal point for maximizing computational efficiency in a low-power environment, the determining comprising inferring a current power condition;

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point;

assuming a minimum power available;

activating all processing units;

outputting a maximum frequency, based on a given power source voltage, as a present frequency to the active processing units;

saving the present frequency and a number representative of a quantity of active processing units;

designating an operations/second, determined according to the present frequency and the quantity of active processing units, as a maximum operations/second;

decrementing the quantity of active processing units;

designating an operations/second, determined according to the present frequency and a new quantity of active processing units, as a current operations/second;

if the current operations/second is greater than the maximum operations/second, setting the current operations/second as a new maximum operations/second and saving the new present frequency and a number representative of the new quantity of active processing units; and

if the current operations/second is not greater than the maximum operations/second, restoring a last saved present frequency and a last saved number representative of the quantity of active units.

18. A computer program product comprising a computer usable medium having readable program code embodied in the medium, the computer program product includes at least one component to:

determine an optimal point for maximizing computational efficiency in a low-power environment; and

selectively control operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point, wherein the determined optimal point is determined by sensing or measuring directly available power, and knowing the available power, an applicable surface plot of operations per second as a function of frequency and number of operating units is accessed such that the optimal point for maximizing computational efficiency in low-power environments is determined.

19. The computer program product of claim 18, further comprising selecting from data, obtained from an analysis applied voltage versus current, applicable data based upon the present power available, and determining an optimal point for maximizing computational efficiency from the applicable data.

20. The computer program product of claim 18, wherein the available power is determined by sensing or measuring directly available sunlight.

21. The computer program product of claim 18, wherein the applicable surface plot of operations per second as a function of frequency and number of operating units is accessed from a database of surface plots.

22. A method to optimize computational efficiency in a low-power environment, comprising:

determining an optimal point for maximizing computational efficiency in a low-power environment; and

selectively controlling operation of at least one processing unit of a plurality of processing units in accordance with the determined optimal point; and

one of;

(i) wherein the determining the optimal point for maximizing computational efficiency comprises;

assuming a minimum power available;

activating all processing units;

outputting a minimum frequency as a present frequency to the active processing units;

saving the present frequency and a number which is representative of a quantity of active processing units;

designating an operations/second, determined according to the present frequency and the quantity of active processing units, as a maximum operations/second;

iteratively increasing the present frequency to a new present frequency where the voltage is at about a minimum for the new present frequency;

designating an operations/second, determined according to the new present frequency and the quantity of active processing units, as a current operations/second;

if the current operations/second is greater than the maximum operations/second, setting the current operations/second as a new maximum operations/second, saving the new present frequency and the number which is representative of the quantity of active processing units, and decrementing the quantity of active units; and

if the current operations/second is not greater than the maximum operations/second, restoring a last saved present frequency and a last saved number representative of the quantity of active units; and

(ii) wherein the determining the optimal point for maximizing computational efficiency comprises;

assuming a minimum power available;

activating all processing units;

outputting a maximum frequency, based on a given power source voltage, as a present frequency to the active processing units;

saving the present frequency and a number representative of a quantity of active processing units;

designating an operations/second, determined according to the present frequency and the quantity of active processing units, as a maximum operations/second;

decrementing the quantity of active processing units;

designating an operations/second, determined according to the present frequency and a new quantity of active processing units, as a current operations/second;

if the current operations/second is greater than the maximum operations/second, setting the current operations/second as a new maximum operations/second and saving the new present frequency and a number representative of the new quantity of active processing units; and

if the current operations/second is not greater than the maximum operations/second, restoring a last saved present frequency and a last saved number representative of the quantity of active units.