Team moais

Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: Other Grants and Activities

Hardware Platforms

The GRIMAGE platform

The GrImage platform ( ) gathers a network of cameras and a PC cluster. It is dedicated to interactive applications. GrImage is co-leaded by the Moais and Perception projects . It is the milestone of a strong and fruitful collaboration between Moais and Perception (common publications, software and application development).

GrImage (Grid and Image) aggregates commodity components for high performance video acquisition, computation and graphics rendering. Computing power is provided by a PC cluster, with some PCs dedicated to video acquisition and others to graphics rendering. A set of digital cameras enables real time video acquisition. The main goal is to rebuild in real time a 3D model of a scene shot from different points of view. Visualization can be performed using a head mounted display for first-person interactions or on a multi-projector display-wall for high resolution rendering.

Since July 2009, the computing cluster was upgraded through grants from INRIA and CNRS-LIG. Grimage uses some specific nodes from the Digitalis machine capable of hosting several daughter boards (mainly video acquisition and graphics cards). It relies on Intel Nehalem processors and a high speed Infiniband network. This integrated approach will enable to test interactive applications using a very high number of processing resources as other nodes from the Digitalis machine can be reserved if needed.

The Digitalis machine

Digitalis is a 780 cores cluster based on Intel Nehalem processors and Infiniband network located at INRIA Rhône-Alpes. Digitalis has been designed to suit both the needs for batch computations and interactive applications. As mentioned before, one rack is dedicated to nodes hosting video acquisition boards and graphics cards. These nodes are mainly used for the Grimage platform, but can also be used for batch computing. Additional nodes with Nvidia Tesla GPUs will be installed during the first quarter of 2010.

By having a single unified machine for batch and interactive computing we expect to better use the available resources, favor the emergence of high performance applications integrating interactive steering and vice versa enable the development of a new generation of interactive 3D applications using a significantly larger number of CPUs and GPUs that what has been done so far on the Grimage platform.

SMP Machines

MOAIS invested in 2006 on two SMP architectures:

These machines enables us to keep-up with the evolution of parallel architectures and in particular today's availability of large multi-core machines. They are used to develop and test parallel adaptive algorithms taking advantage of the processing power provided by the multiple CPUs and GPUs available. We expect to have a new large SMP architecture by the end of 2010 or the beginning of 2011 to replace these machines.


ST Microelectronics provided us a STM8010 machine for experimenting parallel adaptive algorithms on MPSoC.


Logo Inria