Garbage Collection in a gbeta Virtual Machine with the Train Algorithm

Ilsøe, Peer Møller ; Pedersen, Simon Hem

10. term

Master of Science in Engineering in Computer Engineering (Aalborg)

2001

Dette speciale udvikler en virtuel maskine (software, der kører programmer) til programmeringssproget gbeta, kaldet gbvm. Dens særkende er en garbage collector baseret på tog-algoritmen (train algorithm), en metode til automatisk oprydning i hukommelsen. Specialet omfatter tre dele: (1) en introduktion til gbeta, virtuelle maskiners arkitektur samt centrale begreber og algoritmer inden for garbage collection; (2) en gennemgang af design og implementering af gbvm; og (3) eksperimenter, der undersøger, hvordan ydelsen påvirkes, når man ændrer både parametre og variationer af tog-algoritmen. Det tilhørende eksperimentelle testmiljø gør det muligt at udføre flere forsøg med gbvm. Konklusionen er, at der ikke findes én fast generel indstilling, der er både tids- og hukommelseseffektiv for alle de testede programmer.

This thesis develops a virtual machine (software that runs programs) for the programming language gbeta, called gbvm. Its distinctive feature is a garbage collector based on the train algorithm, a method for automatically reclaiming memory. The thesis has three parts: (1) background on gbeta, virtual machine architectures, and key concepts and algorithms in garbage collection; (2) the design and implementation of gbvm; and (3) experiments that examine how performance changes when tuning parameters and altering aspects of the train algorithm. The experimental framework built for this work enables further studies with gbvm. The main conclusion is that no single fixed configuration is both time- and space-efficient across all tested programs.

[This abstract was generated with the help of AI]

Download
View record in AAU Student Projects