Energy optimization through intentional errors
First Claim
1. A method for intentionally allowing errors in a computational system to optimize energy consumption of the computational system, the method comprising:
- performing a cost-benefit analysis to identify one or more allowable errors and one or more non-allowable errors in the computational system, the allowable errors being identified by the cost-benefit analysis as being acceptable errors for optimizing energy consumption with respect to accuracy of the computational system, the non-allowable errors being identified by the cost-benefit analysis as being unacceptable errors for optimizing energy consumption with respect to accuracy of the computational system; and
transforming the computational system from a first state in which the computational system corrects or prevents the allowable errors and the non-allowable errors into a second state in which the computational system allows the allowable errors and corrects or prevents the non-allowable errors.
6 Assignments
0 Petitions
Accused Products
Abstract
Technologies are described herein for intentionally allowing errors in a computational system to optimize energy consumption of the computational system. A cost-benefit analysis is performed to identify one or more allowable errors and one or more non-allowable errors in the computational system. The allowable errors may be identified by the cost-benefit analysis as being acceptable errors for optimizing energy consumption with respect to accuracy of the computational system. The non-allowable errors may be identified by the cost-benefit analysis as being unacceptable errors for optimizing energy consumption with respect to accuracy of the computational system. The computational system is transformed from a first state in which the computational system corrects or prevents the allowable errors and the non-allowable errors into a second state in which the computational system allows the allowable errors and corrects or prevents the non-allowable errors.
158 Citations
20 Claims
-
1. A method for intentionally allowing errors in a computational system to optimize energy consumption of the computational system, the method comprising:
-
performing a cost-benefit analysis to identify one or more allowable errors and one or more non-allowable errors in the computational system, the allowable errors being identified by the cost-benefit analysis as being acceptable errors for optimizing energy consumption with respect to accuracy of the computational system, the non-allowable errors being identified by the cost-benefit analysis as being unacceptable errors for optimizing energy consumption with respect to accuracy of the computational system; and transforming the computational system from a first state in which the computational system corrects or prevents the allowable errors and the non-allowable errors into a second state in which the computational system allows the allowable errors and corrects or prevents the non-allowable errors. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. A computer-readable medium having instructions stored thereon for execution by a processor to provide a method for intentionally allowing errors in a computational system to optimize energy consumption of the computational system, the computer-readable medium comprising instructions configured for:
-
determining a quantitative cost associated with allowing a given error in the computational system, the quantitative cost comprising a measure of a loss of accuracy of the computational system resulting from allowing the given error; determining a quantitative benefit associated with allowing the given error in the computational system, the quantitative benefit comprising a measure of energy savings of the computational system resulting from allowing the given error; determining whether the quantitative benefit exceeds the quantitative cost by a threshold; responsive to determining that the quantitative benefit exceeds the quantitative cost by the threshold, identifying the given error as an allowable error; responsive to determining that the quantitative benefit does not exceed the quantitative cost by the threshold, identifying the given error as a non-allowable error; and selecting and configuring design operations for allowing the allowable error in the computational system and correcting or preventing the non-allowable error with respect to designing the computational system. - View Dependent Claims (12, 13, 14, 15)
-
-
16. A system operative to intentionally allow errors in a computational system to optimize energy consumption of the computational system, the system comprising:
-
a design module operative to select and configure design operations to facilitate allowing an allowable error in the computational system; a manufacturing module operative to select and configure manufacturing operations to facilitate allowing the allowable error in the computational system; an implementation module operative to select and configure implementation operations to facilitate allowing the allowable error in the computational system; an error correction and prevention module operative to correct or prevent the allowable error while the computational system is executing a computer program; a system analysis module operative to select the allowable error and to analyze the operations of the design module, the manufacturing module, the implementation module to estimate an accuracy and an energy consumption of the computational system; and a performance validation module operative to validate the estimated accuracy and the estimated energy consumption during actual execution of the computer program on the computational system. - View Dependent Claims (17, 18, 19, 20)
-
Specification