Technical Report CSI-0010

Automatic High Performance Structural Optimisation for Agent-based Models

A. V. Husselmann and K. A. Hawick

Archived: 2014

Abstract

The problem of structural optimisation for agent-based models is one which holds great promise. Being able to optimise a set of behaviours has potential to improve productivity greatly, and at the very least, generate inspiration. This problem consists of three smaller problems which must be mitigated: ease of use, performance, and also the use of combinatorial optimisation. In this article, these three problems are managed by introducing a domain-specific language (DSL) operating over graphical processing units (and also single-threading) for performance, and a suitable optimiser for this architecture. We carry out a number of experiments to demonstrate and evaluate the performance and effectiveness of this approach. We conclude that such a methodology is indeed useful and performs adequately but is currently limited by the lack of debugging support and visual programming tools.

Keywords: CUDA, parallel, optimisation, domain-specific languages, agent-based models, karva

Full Document Text: PDF version.

Citation Information: BiBTeX database for CSI Notes.

BiBTeX reference:

@InProceedings{CSI-0010,
        Title = {Automatic High Performance Structural Optimisation for Agent-based Models},
        Author = {A. V. Husselmann and K. A. Hawick},
        Booktitle = {Proc. 14th Int. Conf. on Software Engineering Research and Practice (SERP'14)},
        Year = {2014},
        Address = {Las Vegas, USA},
        Month = {21-24 July},
        Note = {ISBN: 1-60132-286-0},
        Pages = {346-352},
        Publisher = {WorldComp},
        Keywords = {CUDA, parallel, optimisation, domain-specific languages, agent-based models, karva},
        Owner = {kahawick},
        Timestamp = {2014.08.17},
        Url = {http://digital.hull.ac.uk/reports/0010/csi-0010.html}
}