Adaptive optimization of low power strategies
Methods and apparatus are provided for adaptive optimization of low-power strategies. In one novel aspect, the device monitors one or more thermal-performance parameters and determines a plurality of operation scenarios for a plurality of corresponding low-power policies. Based on corresponding operation scenarios, the device selects corresponding low-power policy. The device applies different low-power strategy for temperature control based on low-power policies. Different low-power policy is applied to different low-power techniques, such as the DVFS, the CPU hot-plug, and the task migration. In another novel aspect, the device obtains one or more user-defined policy for each corresponding low-power technique. The selection of each low-power policy is further based on its corresponding user-defined policy. In one embodiment, the user-defined DVFS policy includes power policy, performance policy, and DVFS-balanced policy. The user-defined CPU hot-plug policy includes conservative policy, aggressive policy, and hot-plug-balanced policy. The user-defined task-migration policy includes performance policy, and task-migration-balanced policy.
1. A method, comprising:
monitoring one or more thermal-performance parameters by an apparatus;
determining a plurality of operation scenarios for a plurality of corresponding low-power policies based on the one or more thermal-performance parameters;
generating one or more long-term average values of the one or more thermal-performance parameters, and one or more short-term average values of the one or more thermal-performance parameters;
selecting each low-power policy based on the corresponding operation scenario, wherein each low-power policy applies to a corresponding low-power strategy, wherein the selecting of the plurality of low-power policies further comprising: obtaining a user-defined policy for each low-power strategy and selecting each low-power policy further based on corresponding user-defined policy; and
dynamically applying each low-power policy to its corresponding low-power strategy, wherein the user-defined policy for a Dynamic Voltage and Frequency Scaling (DVFS) comprising: power policy, performance policy, and DVFS-balanced policy, and wherein the user-defined policy for a CPU hot-plug comprising: conservative policy, aggressive policy, and hot-plug-balanced policy, and wherein the user-defined policy for a task migration comprising: performance policy, and task-migration-balanced policy.
2. The method of claim 1 , wherein the low-power strategies comprising: a Dynamic Voltage and Frequency Scaling (DVFS), a CPU hot-plug, and a task migration.
3. The method of claim 1 , wherein the thermal-performance parameters comprising: a loading, a task-level parallelism (TLP), a number of CPU cores, and an operating frequency.
4. The method of claim 1 , wherein the operation scenario is based on one or more long-term average values of the one or more thermal-performance parameters.
5. The method of claim 1 , wherein the selection of each low-power policy is further based on the long-term average values, and the short-term average values.
6. The method of claim 1 , wherein the long-term average loading value is high while one or more short term loading values are low, and wherein a fast-reaction policy is adopted by one or more low-power strategies.
7. The method of claim 1 , wherein the long-term average loading value is low while one or more short term loading values are high, and wherein a slow-reaction policy is adopted by one or more low-power strategies.
8. An apparatus, comprising:
one or more sensors that monitor one or more thermal-performance parameters; and
a controller connected to the one or more sensors, the controller configured to
determine a plurality of operation scenarios for a plurality of low-power policies based on the one or more thermal-performance parameters, generate one or more long-term average values of the one or more thermal-performance parameters, and one or more short-term average values of the one or more thermal-performance parameters, select each low-power policy based on the corresponding operation scenario, wherein each low-power policy applies to a corresponding low-power strategy, and dynamically apply each low-power policy to its corresponding low-power strategy, wherein the selecting of the plurality of low-power policies further comprising: obtaining a user-defined policy for each low-power strategy and selecting each low-power policy further based on corresponding user-defined policy, and wherein the user-defined policy for a Dynamic Voltage and Frequency Scaling (DVFS) comprising: power policy, performance policy, and DVFS-balanced policy, and wherein the user-defined policy for a CPU hot-plug comprising: conservative policy, aggressive policy, and hot-plug-balanced policy, and wherein the user-defined policy for a task migration comprising: performance policy, and task-migration-balanced policy.
9. The apparatus of claim 8 , wherein the low-power strategies comprising: a Dynamic Voltage and Frequency Scaling (DVFS), a CPU hot-plug, and a task migration.
10. The apparatus of claim 8 , wherein the thermal-performance parameters comprising: a loading, a task-level parallelism (TLP), a number of CPU cores, and an operating frequency.
11. The apparatus of claim 8 , wherein the operation scenario is based on one or more long-term average values of the one or more thermal-performance parameters.
12. The apparatus of claim 8 , wherein the selection of each low-power policy is further based on the long-term average values, and the short-term average values.
13. The apparatus of claim 8 , wherein the long-term average loading value is high while one or more short term loading values are low, and wherein a fast re-action policy is adopted by one or more low-power strategies.
14. The apparatus of claim 8 , wherein the long-term average loading value is low while one or more short term loading values are high, and wherein a slow re-action policy is adopted by one or more low-power strategies.
15. An apparatus, comprising:
one or more sensors that monitor one or more thermal-performance parameters; and
a controller connected to the one or more sensors, the controller configured to determine a plurality of operation scenarios for a plurality of low-power policies based on the one or more thermal-performance parameters, generate a user-defined policy for each low-power strategy, select each low-power policy based on the corresponding operation scenario and corresponding user-defined policy, wherein each low-power policy applies to a corresponding low-power strategy, and wherein the user-defined policy for a Dynamic Voltage and Frequency Scaling (DVFS) comprising: power policy, performance policy, and DVFS-balanced policy, and wherein the user-defined policy for a CPU hot-plug comprising: conservative policy, aggressive policy, and hot-plug-balanced policy, and wherein the user-defined policy for a task migration comprising: performance policy, and task-migration-balanced policy, and
dynamically apply each low-power policy to its corresponding low-power strategy.