Adaptive power management of dynamic random access memory

Patent metadata

Field	Value
Publication number	US-10956057-B2
Application number	US-201816115845-A
Country	US
Kind code	B2
Filing date	Aug 29, 2018
Priority date	Aug 29, 2018
Publication date	Mar 23, 2021
Grant date	Mar 23, 2021

Abstract

Various additional and alternative aspects are described herein. In some aspects, the present disclosure provides a method of controlling a memory of a computing device by an adaptive memory controller. The method includes collecting usage data from the computing device over a first bin, wherein the first bin is associated with a first weight, wherein the first weight is indicative of one or more of a first partial array self-refresh (PASR) setting a first partial array auto refresh (PAAR) setting and a first deep power down (DPD) setting. The method further includes associating the collected data with a second weight, adapting the first bin based on the second weight, wherein the second weight is indicative of one or more of a second PASR, PAAR, and DPD setting. The method further includes controlling the memory during the next first bin based on the second weight.

First claim

What is claimed is:

1. A method of controlling a memory of a computing device by an adaptive memory controller, the method comprising:
collecting usage data from the computing device over an instance of a first bin corresponding to a repeating time period, wherein the first bin is associated with a first weight based on previously collected usage data of the computing device over a previous instance of the first bin, wherein the first weight maps to one or more of a first partial array self-refresh (PASR) setting, a first partial array automatic refresh (PAAR) setting, and a first deep power down (DPD) setting, wherein the first weight is a numerical value that controls a number of portions of the memory to which to apply power management;
associating the collected usage data over the instance of the first bin with a second weight;
associating the first bin with the second weight based on the second weight being different than the first weight, wherein the second weight maps to one or more of a second PASR setting, a second PAAR setting, and a second DPD setting, wherein the second weight is a numerical value that controls a number of portions of the memory to which to apply power management, wherein the memory is configured to be refreshed during the next instance of the first bin based on the second weight prior to the start of the next instance of the first bin; and
controlling a refresh of the memory during the next instance of the first bin based on the second weight by utilizing the one or more of the second PASR setting, the second PAAR setting, and the second DPD setting to determine a refresh rate.

2. The method of claim 1 , wherein associating the collected usage data over the instance of the first bin with the second weight is based in part on memory usage of the computing device during the first bin.

3. The method of claim 1 , wherein associating the collected usage data over the instance of the first bin with the second weight is based in part on information from one or more sensors of the computing device during the first bin.

4. The method of claim 1 , wherein the collected usage data includes data indicative a location, a movement, or a battery level.

5. The method of claim 1 , wherein the collected usage data is integrated, and wherein integrating the data comprises one or more of averaging or filtering the collected usage data.

6. The method of claim 1 , further comprising:
collecting usage data from the computing device over a second bin, wherein the second bin is associated with a third weight, wherein the third weight is indicative of one or more of a third PASR setting, a third PAAR setting, and a third deep power down DPD setting;
associating the collected usage data with a fourth weight, wherein the fourth weight is indicative of one or more of a fourth PASR setting, a fourth PAAR setting, and a fourth DPD setting;
adapting the second bin based on the fourth weight; and
controlling the memory during the next second bin based on the fourth weight.

7. A method of controlling a memory of a computing device by an adaptive memory controller, the method comprising:
collecting a first data set from the computing device over at least an instance of a first bin, wherein the first data set comprises usage data;
associating the collected first data set with a first weight, wherein the first weight is indicative of one or more of a first partial array self-refresh (PASR), a first partial array auto refresh (PAAR) setting, and a first deep power down (DPD) setting, wherein the first weight is a numerical value that controls a number of portions of the memory to which to apply power management, wherein the memory is configured to be refreshed during the next instance of the first bin based on the first weight prior to the start of the next instance of the first bin; and
controlling a refresh of the memory during the next instance of the first bin based on the first weight by utilizing the one or more of the first partial array self-refresh (PASR), the first partial array auto refresh (PAAR) setting, and the first deep power down (DPD) setting to determine a refresh rate.

8. The method of claim 7 , wherein the first bin length is predetermined.

9. The method of claim 8 , where the first bin length is one hour.

10. The method of claim 7 , wherein associating the collected first data set with the first weight is based in part on memory usage of the computing device during the instance of the first bin.

11. The method of claim 7 , wherein associating the collected first data set with the first weight is based in part on information from one or more sensors of the computing device during the instance of the first bin.

12. The method of claim 7 , wherein the collected first data set includes data indicative a location, a movement, or a battery level.

13. The method of claim 7 , comprising:
collecting a second data set from the computing device over the next instance of the first bin;
determining if the second data set is indicative of the first weight; and
confirming the first bin when the second data set is indicative of the first weight, and
deleting the first bin when the second data set is indicative of a weight other than the first weight.

14. The method of claim 13 , wherein the next instance of the first bin is on a different day than the instance of the first bin.

15. The method of claim 7 , further comprising:
processing the collected first data set with a neural network and generating the at least one instance of the first bin based on the first data set.

16. The method of claim 15 , comprising:
processing the collected first data set with a neural network and generating a second bin associated with a second weight, wherein the second bin is associated with a second weight, and wherein the second weight is indicative of one or more of a second PASR setting, a second PAAR setting, and a second DPD setting; and
controlling the memory in part based on the second bin.

17. The method of claim 15 , wherein generating at the least one instance of the first bin based on the first data set is verified by feedback received at the computing device; and
when the feedback is indicative of a verification, controlling the power to the memory based on the at least one instance of the first bin; and
when the memory power setting is not verified, removing the at least one instance of the first bin.

18. A non-transitory medium comprising instructions that when executed by a processor of a computing system cause the computing system to execute a method of controlling a memory of a computing device by an adaptive memory controller, the method comprising:
collecting usage data from the computing device over an instance of a first bin corresponding to a repeating time period, wherein the first bin is associated with a first weight based on previously collected usage data of the computing device over a previous instance of the first bin, wherein the first weight maps to one or more of a first partial array self-refresh (PASR) setting, a first partial array automatic refresh (PAAR) setting, and a first deep power down (DPD) setting, wherein the first weight is a numerical value that controls a number of portions of the memory to which to apply power management;
associating the collected usage data over the instance of the first bin with a second weight;
associating the first bin with the second weight based on the second weight being different than the first weight, wherein the second weight maps to one or more of a second PASR setting, a second

Assignees

• Qualcomm Inc

Inventors

• Li Yanru
• Chun Dexter Tamio

Classifications

• G11C11/40622 Primary

  Partial refresh of memory arrays · CPC title

• G11C11/4074

  Power supply or voltage generation circuits, e.g. bias voltage generators, substrate voltage generators, back-up power, power control circuits · CPC title

• G06F3/0659

  Command handling arrangements, e.g. command buffers, queues, command scheduling · CPC title

• G06F3/0625 Primary

  Power saving in storage systems · CPC title

• G11C2211/4067

  Refresh in standby or low power modes · CPC title