Hardware configuration selection using machine learning model

1. A method for identifying a hardware configuration, the method comprising:

applying feature measurements to a trained model, wherein the trained model includes a plurality of sets of trees, each different set of trees corresponding to a different hardware configuration;

for each set of trees, combining output values of each member tree of the set of trees to form a combined output value for the set, wherein the combined output value is generated from a plurality of certainty scores, wherein each certainty score is specified by a different tree of the set of trees;

selecting a hardware configuration based on the combined output values of the plurality of sets of trees, by selecting a hardware configuration of a set of trees that has the highest combined output value out of a plurality of combined output values corresponding to the plurality of sets of trees; and

operating a computer processor according to the selected hardware configuration,

wherein the output values comprise certainty scores, rankings, or regression values.

2. The method of claim 1 , wherein applying the feature measurements to the trained model comprises:

traversing one or more trees of the trained model, comparing the feature measurements to thresholds of the one or more trees to arrive at a leaf node.

3. The method of claim 2 , wherein obtaining the output values includes obtaining the output values from leaf nodes of the one or more trees of the trained model.

4. The method of claim 2 , wherein the leaf nodes of the trees include regression values.

5. The method of claim 1 , wherein the output values comprise ranking, and the rankings comprise bins of feature measurements.

6. The method of claim 1 , wherein each set of trees comprises a trained forest.

7. The method of claim 1 , wherein combining the output values comprises taking a mean or a mode of the output values.

8. A system for identifying a hardware configuration, the system comprising:

a processor; and

a processor reconfiguration unit configured to:

apply feature measurements to a trained model, wherein the trained model includes a plurality of sets of trees, each different set of trees corresponding to a different hardware configuration;

for each set of trees, combine output values of each member tree of the set of trees to form a combined output value for the set, wherein the combined output value is generated from a plurality of certainty scores, wherein each certainty score is specified by a different tree of the set of trees;

select a hardware configuration based on the combined output values of the plurality of sets of trees, by selecting a hardware configuration of a set of trees that has the highest combined output value out of a plurality of combined output values corresponding to the plurality of sets of trees; and

configure the processor to operate according to the selected hardware configuration,

wherein the output values comprise certainty scores, rankings, or regression values.

9. The system of claim 8 , wherein applying the feature measurements to the trained model comprises:

traversing one or more trees of the trained model, comparing the feature measurements to thresholds of the one or more trees to arrive at a leaf node.

10. The system of claim 9 , wherein obtaining the output values includes obtaining the output values from leaf nodes of the one or more trees of the trained model.

11. The system of claim 9 , wherein the leaf nodes of the trees include regression values.

12. The system of claim 8 , wherein the output values comprise ranking, and the rankings comprise bins of feature measurements.

13. The system of claim 8 , wherein each set of trees comprises a trained forest.

14. The system of claim 8 , wherein combining the output values comprises taking a mean or mode of the output values.

15. A processor comprising:

one or more caches;

a hybrid prefetcher configured to perform prefetching operations on the one or more caches; and

a prefetch reconfiguration unit, configured to:

apply feature measurements to a trained model, wherein the trained model includes a plurality of sets of trees, each different set of trees corresponding to a different hardware configuration;

for each set of trees, combine output values of each member tree of the set of trees to form a combined output value for the set, wherein the combined output value is generated from a plurality of certainty scores, wherein each certainty score is specified by a different tree of the set of trees;

select a hardware configuration based on the combined output values of the plurality of sets of trees, by selecting a hardware configuration of a set of trees that has the highest combined output value out of a plurality of combined output values corresponding to the plurality of sets of trees; and

configure the hybrid prefetcher to operate according to the selected hardware configuration,

wherein the output values comprise certainty scores, rankings, or regression values.

16. The processor of claim 15 , wherein applying the feature measurements to the trained model comprises:

traversing one or more trees of the trained model, comparing the feature measurements to thresholds of the one or more trees to arrive at a leaf node.

17. The processor of claim 16 , wherein obtaining the output values includes obtaining the output values from leaf nodes of the one or more trees of the trained model.

18. The processor of claim 16 , wherein the leaf nodes of the trees include regression values.

19. The processor of claim 15 , wherein the output values comprise ranking, and the rankings comprise bins of feature measurements.

20. The processor of claim 15 , wherein each set of trees comprises a trained forest.