Get started
Work

Tehokkaiden tiedonhallinta-algoritmien suunnittelu, toteutus ja evaluointi segmentoidussa muistimallissa

Year: 2014
Type: dissertation
Abstract: The data structures and algorithms used for storing and manipulating the ever growing data sets are the most crucial factor that affects the performance of a computer based system. During the past decade or so, most systems have begun to provide a standard set of high performance data structures either as system libraries or via direct support in the programming languages. Some legacy systems however have not quite kept up with this development and still require additional effort from the programmer to achieve good performance scalability. The objective of this work is to design, implement and evaluate a suitable set of selected data structures on top of one such legacy system, the DX 200 telecommunications platform. This task is complicated by the somewhat eccentric nature of the platform including the memory model it uses. While the modern IA-32 CPU hardware still fully supports the segmented memory model, the world has almost completely moved on and left it behind. In practice this means that it is almost impossible to get any modern high quality tools like e.g. compilers for it and the performance of segment operations hasn't been nearly as aggressively optimized as other CPU functions. All these things always need to be taken into consideration when trying to implement high performance software for the platform and it is almost impossible to implement anything without making at least some trade-offs due to them. The end result of this work is a tested standard library for the DX 200 platform that implements two types of lists and three types of binary trees and can be used as a basis for future expansion and further evaluation of implementation alternatives. The evaluation done on the implementation clearly shows that even in this kind of system, the balanced trees scale very nicely while the array that is the only data structure provided natively by the used programming languages doesn't scale all that well. This result is totally in line with all the theoretical background and so for its part acts as further proof of the validity of the produced implementation.
Author: Huttunen, Janne
Institution: Aalto University
Language: English
Open Access status: green
OpenAlex ID: W7113266170
Cites:
Cited by:
Related to: 10