Principle Investigator Stuart Ramsden Project s78

Australian Centre for the Arts and Technology Machine CM

Institute of the Arts

An Algorithm Visualiser for the CM-5

The complimentary nature of animation and computation have lead to design philosophies for visual programming languages in which algorithms are not only defined visually but are examined and debugged using the tools of realtime animation. Most work to date has been concerned with 2D representation and animation . Despite the ability of parallel computers to deal naturally with multi dimensional data, and the fact that parallel algorithms have an intrinsic geometric flavour, current debugging tools (for instance PRISM) rely only on 2D representations. The work to be undertaken here involves the development of an environment for algorithm visualisation, which will utilise the ANUSF's new Onyx graphics workstation to calculate and render complex geometric representations of parallel algorithms running concurrently in realtime on the CM-5. The work will build on and extend the realtime networking and graphics software developed by the principle researcher during his last project Visualising Recursive Function Spaces in Time. The work has application in both the teaching and understanding of the ideas of parallel programming, and in the analysis of parallel algorithms themselves.

What are the basic questions addressed?

The researcher asks how realtime interactive computer 3D graphics can itself be a tool in representing the function (abstract or otherwise) of a computer in operation, and how this can be used in iterative program development.

What are the results to date and the future of the work?

Time spent thus far has been largely the development of the front end graphical interface for the project. Some early results using interactive networking have been obtained and the system is still in its infancy.

This is a long term project by nature which promises to lead to the development of new paradigms for interacting with supercomputers.

What computational techniques are used and why is a supercomputer required?

This project will look at 3D animated models for data and process representation, as well as program specification and interactive control itself. Questions such as remote/local computation versus bandwidth trade offs will be addressed, with the aim of developing some heuristics for balancing the computation/visualisation load across the networked computers.

This project has stemmed in part from the difficulties associated with analysing the computational behaviour of a supercomputer using traditional methods. Another driving force has been the desire to speed up the exploration of large parameter spaces via fast interactive numerical experiment.


Exploring Recursive Function Systems, S Ramsden, paper presented at the symposium Synaesthetica 94 in July 1994.