MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641085430 A) filed by Vellore Institute Of Technology on July 13, 2026, for System And Method For Adaptive Memory Scheduling Based On Predictive Access Prioritization And Retention-Aware Refresh Control.
Inventors include Jayakumar K; Abhilaksh Chauhan; Maheep Singh; and Ocean Verma.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: An adaptive memory access scheduling system and method for improving memory performance, fairness, and energy efficiency in a dynamic random-access memory (DRAM) subsystem are disclosed. The system comprises a memory controller incorporating a stride- based prediction engine, a thread clustering scheduler, a parallelism-aware batch scheduling engine, a rank-aware workload reshaping module, and a retention-aware refresh management module. The prediction engine identifies recurring memory access patterns and proactively activates predicted memory rows to reduce access latency. The scheduler classifies threads according to memory intensity and dynamically prioritizes memory requests to improve throughput and fairness. The workload reshaping module consolidates frequently accessed data into selected memory ranks while relocating infrequently accessed data to create extended idle periods suitable for low-power operation. The refresh management module profiles memory retention characteristics and selectively schedules refresh operations while utilizing read-induced refresh effects to eliminate redundant refresh commands. The integrated architecture reduces latency, refresh overhead, power consumption, and hardware complexity without relying on machine learning-based control mechanisms.
Disclaimer: Curated by HT Syndication.
