Abstract
A physical page allocator in a mobile operating system places pages without regard to how long the requesting application stays resident, so live pages spread across the whole address space. Partial array self-refresh (PASR) acts at the boundary of a rank or a bank, so one live page can hold a whole region in refresh. This document sets out a method that classifies an application at launch through an on-device model that predicts launch probability and residency. The pages of that application then go to a refresh-persistent zone or to a collapsible zone on separate ranks. Once a collapsible zone runs empty, a refresh mask controller writes a mode register and drops refresh for the vacant region. Refresh power can fall without the reactive page migration a prior approach performs during idle windows.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Mohammad, Haris and Baidya, Susmita, "Machine Learning Guided Physical Page Segregation into Refresh-Persistent and Collapsible Zones for Partial Array Self-Refresh", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11707