BIOLOGICAL-LIKE MEMORY ALLOCATION SCHEME SIMULATION

Abstract

When Gordon Moore observed the number of transistor increasing pattern while memory bandwidth could not catch up with processing unit performance, this diverging rate kept stretching out to create what eventually transpired to be “Memory Wall.” This consequence becomes a major performance bottleneck. Many bottleneck elimination approaches have been attempted. They incorporate considerable overhead and high complexity. This research proposes a novel memory allocation scheme that employs biological behavioral principles of the living creatures. At the principal construct of their life form lives the cells having limited resources, yet passively operates with little overhead. The proposed method imitates this unicellular characterization that operates on one task at a time, thereby memory occupation is reduced to First-In-First-Out activation discipline. Processing can thus be regulated by a global clock that permits one active task at any given time to reside in memory. Consequently, low overhead memory allocation scheme can be achieved without the need for elaborate algorithms. The most anticipatory benefit is technological transfer of the proposed scheme to hardware that will eventually alleviate the Memory Wall problem.