Embedded Systems

Implementation and Training of a Machine Learning Model to Predict Cache Hit/Miss Rates

Bachelor’s Thesis / Student Research Project

Abstract

Abstract modelling of HW/SW systems is a fairly new research topic. The goal of this technique is to capture only the essential parameters of software and hardware which influence their timing behavior. One crucial aspect of any HW/SW system’s timing behavior is how long a memory instruction will take. Unfortunately a complete cache simulation can be very expensive.

Goal of this student project is to implement and train a machine learning model to predict the cache hit/miss rate of a trace.

Requirements

  • Python, Scikit, PyTorch
  • Machine Learning Knowledge
  • Successfully atteded the lecture “Grundlagen der Rechnerarchitektur” and/or “Parallele Rechnerarchitekturen” (optional)
  • Linux (optional)

Contact

Jung, Alexander

Lübeck, Konstantin

Bringmann, Oliver