About the LBT laboratory
Numerical simulations on biological macromolecules are carried out in the Laboratoire de Biochimie Théorique (denoted LBT). This group has more than ten years of experience in developing and applying modeling methods to study the deformation of biological macromolecules [19-23]. Besides a strong publication activity, LBT participates in National and European projects like the ANR project FonFlon and the EU project IMMUNOPRION. LBT has considerable experience in advanced simulation methods based on molecular dynamics simulations of challenging biological systems like membrane proteins with up to 340 000 atoms. As such systems require long simulation timescales, LBT is an important user of the French national IDRIS computing center and familiar with high performance computing and parallelized codes on PC clusters or supercomputers. Furthermore simplified protein models (going from a full atomic description to one or two points per residue only) are being developed and evaluated in order to use modern multi-scale approaches for efficient simulation of complex systems, even on commodity hardware.
Current research topics like the investigation of mechanical properties of proteins or macromolecular docking are particularly apt to virtual reality extensions. An interactive graphics system with a haptic interface, an early FVNano prototype named SHAMAN, was already developed and aims at interfacing the numerous in-house software packages [58,59]. In particular, LBT has developed software based on internal coordinates reducing the number of variables taken into account during modeling and allowing the use of energy minimization for the study of large conformational changes [20, 21].
Partner laboratories
CEA/DIF/DSSI
CEA/DIF/DSSI at Bruyères-le-Châtel operates the CEA supercomputing complex (TERA and CCRT computing centers, resp. for defense and research/industrial applications). Many different large- scale scientific simulations are run at CEA/DIF (dealing with laser plasma interaction, CFD, molecular dynamics for materials science, micro-electronics...) for applications ranging from life and earth sciences to defense, energy and aeronautics. CEA/DIF has also an open research facility in the TER@TEC area at Bruyères-le-Châtel, equipped with PC clusters for parallel computing and visualization, a tiled display and various VR peripherals.
DSSI (Département des Sciences de la Simulation et de l'Information) has teams involved in the administration and operation of complex clusters, development and exploitation of massively parallel simulation codes and high performance 3D meshing and visualization [52-55].
Laboratoire d'Informatique Fondamentale d'Orléans
The Laboratoire d'Informatique Fondamentale d'Orléans (www.univ-orleans.fr/lifo/, denoted LIFO) is a laboratory of the University of Orléans. Research projects conducted at LIFO are concerned with Software Science and Information Technology.
One of the current research topics consists in applying our background in semantics, parallel processing and distributed systems to virtual reality and high-end visualization environments. The results of this research orientation can be divided in three parts. The first one is an experimental software platform NetJuggler [32] built on the freeware VRJuggler from Iowa State University's Virtual Reality Applications Center. We add to this solution swaplock and genlock software (SoftGenlock) to provide active stereo on display walls driven by PC clusters [33].The second one is the middleware FlowVR (flowvr.sf.net), a common development with Moais (INRIA project) [34] initiated in the RNTL Geobench project in order to offer a tool for heterogeneous code coupling. LIFO continues its contribution to FlowVR with the design of cost models and efficient application mappings according to the targeted distributed architecture. The last one concern s level of details algorithms to optimize scientific visualization in case of massive data rendering. Adapted to our cluster-based environment we applied our research in developing a distributed framework for generalized LOD rendering which was applied to large terrains [50, 51]. This research axis is exploited in the ANR Dalia project and in the context of collaboration with a local PME GeoHyd dealing with environmental problems and GIS (Geographical Information System). Another current research topic is the design of semantics and implementations for highly modular programming languages. Modularity is in this case mainly studied as various kinds of compositionality in functional languages. This research has been supported by the Caraml project (caraml.free.fr, ACI Grid Program) and the Propac project (propac.free.fr, Young Researchers Program).The LIFO set up a Virtual Reality platform MIReV based on a high performance PC cluster composed of 17 dual-processor machines interconnected through a dual-gigabit Ethernet and Myrinet network and on a 1.5x1.1m display wall with a 2048x768 resolution display for stereoscopic projection from 4 video projectors.
Moais team, INRIA Rhône-Alpes
The Moais team from INRIA Rhône-Alpes.
Moais research focuses on adaptive parallel algorithms, parallel code coupling, distributed programming environments and interactive parallel computing (moais.imag.fr). Besides a strong publication activity, Moais participates in various National and European projects like, RNTL Clic and GeoBench, ANR projects Dalia and Numasis, the Sceptre project of the «pole de compétitivité» Minalogic, or the European project CoreGrid. It has various contracts with industrial partners like Bull, ST Microelectronics or DCN. Moais software developments are mainly focused on two libraries, FlowVR (flowvr.sf.net), developed in collaboration with the LIFO, and Kaapi (kaapi.gforge.inria.fr). Moais has also contributed to several significant FlowVR extensions, distributed today in the FlowVR Suite (FlowVR is detailed in section 7.2). Moais is strongly involved in the management of 2 experimental platforms: the itanium cluster i-cluster 2 (i-cluster2.inrialpes.fr/) and the GrImage platform (www.inrialpes.fr/grimage). GrImage gathers a cluster of 26 dual processor PCs, a 16 cores PC, a 16 projector display wall and 15 cameras. One goal that Moais pursues through the GrImage platform, in collaboration with the Evasion, Perception and I3D teams from INRIA Rhône-Alpes, is to study how parallel computing, virtual reality, scientific visualization, computer graphics and computer vision can be combined to develop novel high performance interactive applications. Based on its experience on parallel algorithms and GPU programming Moais is also working on the new generations of programmable GPUs (using the Nvidia Cuda library). One goal is to develop parallel algorithms that can efficiently and seamlessly take advantage of the CPUs and GPUs available on the target machine.
References
R19. Cluzel, P., et al., DNA: An extensible molecule. Science, 1996. 271: p. 792-794.
R20. Lafontaine, I. and R. Lavery, Collective Variable Modelling of Nucleic Acids. Curr. Opin. Struct. Biol., 1999. 76: p. 2760-2768.
R21. Lavery, R., Modeling Nucleic Acids: Fine Structure, Flexibility and Conformational Transitions. Adv. Comput. Biol., 1994. 1: p. 69-145.
R22. Lavery, R., et al., Structure and mechanics of single biomolecules: experiment and simulation. J. Phys.: Condens. Matter, 2002. 14: p. R383-R414.
R23. Lebrun, A., R. Lavery, and H. Weinstein, Modeling multi-component protein-DNA complexes: the role of bending and dimerization in the complex of p53 dimers with DNA. Protein Engineering, 2001. 14: p. 233-243.
R32. Allard, J., et al., Net Juggler: Running VR Juggler with Multiple Displays on a Commodity Component Cluster. IEEE VR, 2002: p. 275- 276.
R33. Allard, J., et al. Softgenlock: Active Stereo and Genlock for PC Cluster. in Proceedings of the Joint IPT/EGVE'03 Workshop. 2003. Zurich, Switzerland.
R34. Allard, J., et al. FlowVR: a Middleware for Large Scale Virtual Reality Applications. in EuroPar 2004. 2004.
R50. Gouranton, V.a.M., S. and Melin, E. and Nortet, C. Interactive Rendering of Massive Terrains Using PC Cluster. in EuroVis 2005: Eurographics/IEEE-VGTC Symposium on Visualisation. 2005.
R51. Madougou, S. and J. Vairon, Fast deployment of a commodity VR cluster. VR-Cluster'03 Review, 2003.
R52. J.E. Coste, Ph. Guerville, J.-Ph. Nominé, Visualisation haute performance et haute résolution à base de clusters de PC sous Linux Rapport CEA-R-6001, 2002
R53. J.-Ph. Nominé, Visualization for High Performance Computing : which Part for Virtual Reality Approaches and Techniques? 2nd International INTUITION Workshop, Senlis, 24-25 Novembre 2005
R54. O. Bressand, G. Colin de Verdière, J.-C. Lafoucrière, J.-P. Nominé, I. Surin, Exploitation des résultats de calcul : du stockage à la visualisation des données Revue CEA/DAM CHOCS No 28 « TERA », Octobre 2003
R55. D. Aguilera, T. Carrard, G. Colin de Verdière, J.-Ph. Nominé, Visualizing Large Scale Laser-Plasma Interaction 3D Simulations Using Parallel VTK and extensions Poster IEEE Visualization 2005, Minneapolis, Octobre 2005
R58. M. Baaden, E. Gasser, C. Prévost, R. Lavery, SHAMAN – Système HAptique de MAnipulation Nanoscopique. http://www.ibpc.fr/UPR9080/axes_recherches1.html
R59. E. Gasser, SHAMAN – Manipulation sensitive des macromolécules, Rapport DESS, Aout 2003
Support
This project has been selected by the ANR funding agency in the CIS (Calcul Intensif, High performance Computing) program 2007.
