Memory Power Management via Memgroup Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory power management systems are inefficient as they often render memory unusable in low-power modes and require full power to load or store data, leading to challenges in gracefully powering down memory regions without data loss, and lack granular control over power allocation.
Innovation Solution
A system and method that divides memory into memgroups, allowing for separate power management of physical pages, where a resource controller marks a memgroup as stopped, remaps virtual pages, and reduces power to unused physical pages, enabling efficient power reduction without data loss by relocating active pages to other memgroups.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by stationary object
If memory is placed in low-power mode, then power consumption is reduced, but memory becomes unusable and data cannot be loaded or stored
Solution Approach 1:
The memory system is divided into multiple memgroups (first memgroup, second memgroup), each capable of independent power management. This allows one memgroup to be powered down while another remains active and usable, resolving the contradiction between power reduction and memory usability.
Solution Approach 2:
Power management is applied locally to individual memgroups rather than uniformly to the entire memory system. The resource controller can selectively reduce power to specific memgroups that are not currently in use, while maintaining full power to active memgroups, thus achieving power savings without compromising overall memory functionality.
2Use of energy by stationary object
If memory regions are powered down to reduce power consumption, then power usage decreases, but data loss occurs and memory cannot be gracefully powered down
Solution Approach 1:
The resource controller performs preliminary actions by tracking which virtual pages are mapped to which physical pages before attempting to power down a memgroup. This advance knowledge allows the system to safely power down memgroups that contain only unmapped or unused physical pages, preventing data loss while achieving power reduction.
Solution Approach 2:
The system implements feedback mechanisms where the resource controller continuously monitors memory allocation states and uses this information to make informed decisions about which memgroups can be safely powered down. This feedback ensures that power management actions do not compromise data integrity.
3Productivity
If granular control over power allocation is implemented, then power management efficiency improves, but system complexity increases
Solution Approach 1:
The resource controller is designed to perform multiple functions: it manages virtual-to-physical page mappings, tracks memgroup allocation states, and controls power delivery to different memgroups. By consolidating these functions in a single controller, the system achieves granular power management without proportionally increasing overall system complexity.
Data Source
AI summary
A system including memory and a resource controller. The memory includes a first memgroup and a second memgroup, wherein the first memgroup comprises a first physical page mapped to a virtual page, and wherein the second memgroup comprises a second physical page. The resource controller is configured to receive a request to stop the first memgroup, instruct a memory power management subsystem to mark the first memgroup as stopped in response to receiving the request to stop the first memgroup, wherein no free pages are allocated from the first memgroup after the first memgroup is marked as stopped, remap the virtual page to the second physical page in response to the marking the first memgroup as stopped, and reduce power to the first memgroup in response to a determination that the first physical page is not mapped to the virtual page.