COMPUTER ARCHITECTURE AND SYSTEM THROUGH MACHINE LEARNING- A COMPREHENSIVE EXPLORATION OF APPLICATIONS

Abstract

In traditional computer architecture, there are six primary components: the controller, arithmetic unit, memory, storage, input, and output. Both memory and storage are crucial as data can be lost when power is lost, and while storage may retain data, it has slow access speeds. Recently, non-volatile random access memory has been developed, which allows for the combination of two different types of memory. Based on this development, a new computer architecture using dual-space memory is proposed, which can be implemented using the hardware shift latch to provide random access to terabytes of storage space. However, it's now necessary to reconsider how machine learning (ML) and computer architecture/systems interact and how ML can influence the creation of these systems. By applying ML to system design, we can increase designers' output and create a positive feedback loop between ML and system design, leading to improvements in both. In this paper, we focus on two main categories of ML-based system design: ML-based modeling, which predicts performance metrics or other criteria, and ML-based design methodology, which directly uses ML as a design tool. We review current studies focused on ML-based modeling in this paper