Project Team Trio

Members
Overall Objectives
Scientific Foundations
Application Domains
Software
New Results
Dissemination
Bibliography
PDF e-pub XML


Section: Software

VITRAIL

Participants : Damien Bodenes, Pierre Caserta, Olivier Zendra.

The aim of the VITRAIL operation is to provide tools for the advanced and immersive visualization of programs. It partners with the University of Montréal, University of Montpellier and Pareo team of INRIA Nancy Grand Est.

Last years, in VITRAIL, we had developed software to instrument and trace Java programs at the bytecode level. We then had developed an analysis tool able to exploit these traces to compute relevant software metrics. We had hired Damien Bodenes as software developer, and had begun the work on a prototype able to render a 3D world, symbolizing software, onto various visualization hardware, with the possibility to change the display metaphor. The main part of our development work had been in 2009 the choice and validation of the technology, and a first architecture. In 2010, the development had go on at a good pace, building on chosen technologies and architecture. This had brought new experience, and with the first actual runs of our platform, we had realized that with the Irrlicht platform we had chosen, we could reach unforeseeable problem when scaling up. We had thus decided to reverse our choice to the Ogre3D 3D engine at the beginning of 2010. Our development had then progressed steadily.

We had released in 2010 a first prototype of our platform, with all the underlying architecture, able to provide navigation features and interaction capacities limited to the driving of the navigation, as per our plans. This had included dual screen management.

Our first prototype, using 2 large 2D screens, with a city metaphor, had been demonstrated during the "FĂȘte de la Science" in November 2010 and had received a lot of attention and enthusiasm from the general public. About 55 persons per day had visited our booth and got demonstrations.

We had also progressed significantly in our Java bytecode tracer, by improving its granularity, the completeness of the traced information, and its performance as well. We have a unique tool which is able to trace both program classes and JDK classes, at basic block level. In addition, it does so with a dynamic instrumentation of classes, which means there is no need to have an instrumented version of the class files on disk. This is very convenient, especially when changing machine of JVM, or when upgrading either the JDK or the program itself. In addition, the performance is good enough that the instrumented programs are still fully usable in an interactive way, without bothering the user. To the best of our knowledge, this is the only Java bytecode tracer that offers these features nowadays.

Our software development had lead to several registrations with APP:

In 2011, we acquired a workstation and three 30 inches computer screens, to be able to set up a "boxed 3D workstation", that would provide display in front and on both sides of the operator. This would constitute the next step in our experiments, by improving immersion with a larger field of vision (on the sides). The software developments to do this are ongoing. We also integrated a WiiMote interaction device to our system, but our experiments found that its spacial resolution was too poor for our needs.

We finally improved significantly our VITRAIL prototype in 2011, especially by designing and implementing a new representation for the relations between software (hence visual) elements, with limited clutter and the possibility to regroup links and see their direction.