Optimized virtual memory management for dynamic data types

摘要

The present invention proposes effective solutions for the design of Virtual Memory Management for applications with dynamic data types in an embedded (HW or SW) processor context. A structured search space for VMM mechanisms with orthogonal decision trees is presented. Based on said representation a systematic power exploration methodology is proposed that takes into account characteristics of the applications to prune the search space and guide the choices of a VMM for data dominated applications. A parameterizable model, called Flexible Pools, is proposed. This model limits the exploration of the Virtual Memory organization considerably without limiting the optimization possibilities.