Research

 



SILKAN participates in various collaborative Research and Development projects that serve a number of complementary purposes:

  • Seize the opportunity of highly demanding contexts to further challenge its research teams and improve its technologies and products.
  • Stay at the forefront of innovation in its strategic market domains
  • Develop synergies and partnerships with other projects members to create new business opportunities

In order to continuously improve its technology platforms, SILKAN is strongly committed to R&D efforts in the following areas:

  1. Automatic HPC code generation from DSL
  2. Distributed Real time Synchronisation
  3. Technologies for Acurate and fast simulation of Physical Models

 

1. Automatic HPC code generation from DSL:

Models of complex industrial systems, whether for design or behavioral simulation purposes, are often described using “Domain Specific Languages” (DSL): typical examples include Scilab / Matlab scripts, and Simulink / XCOS schemes. These expressive modeling languages are highly convenient for engineers to describe their systems, but may often lead to poor run-time performance which makes then unsuitable for real time or embedded usage.

SILKAN has developed a strong competence in the automatic transformation of such models into high performance code (typically C++ code, with OpenMP or OpenCL pragmas) suited for specific target architectures (new generation many-core, GPU, ARM processors, etc.). SILKAN achieves this goal using software packages such as COLD (Scilab/Matlab to C++), FABRIC (XCOS/Simulink to C) and Par4All (formal code rewriting), which are continuously improved to perform at the highest level thanks to collaborative Research and Development projects. SILKAN also tackles the question of optimized handling of specialized data representations and related algorithms in its Research and Development efforts.

MACH

logo MACH Déc. 2013 – Nov. 2016 ; ITEA3

  • The Project : Aims at providing architecture-aware automatic code generation for Domain Specific data structures and algorithms embedded in generic programming languages (DSeL).
  • Partners : Infineon plus 18 partners from Germany (FZI, Fraunhofer SCAI, TESOBE, TWT GmbH), Belgium (Nobel Biocare, NOESIS, Vrije Universiteit Brussel), France (AS+, Thalès CS, CEA LIST, INRA, Ter@tec), Korea (ENSOLTEK, TN), the Netherlands (Vector Fabrics BV, ViNotion BV)
  • SILKAN : Contributes by providing its expertize and tools for automatic high performance code generation based on the semantic analysis capacities embedded in the Par4All technology and the CruncherTM product

SMECY

smecy 70x66

 

Feb. 2010 – Jan. 2013 ; Artemis 9

  • The Project : Aimed at developing programming models and methods suitable for an effective use of emerging multi-core technologies in the implementation of embedded applications, both at software and architecture levels.
  • Partners : CEA LIST , Thomson Grass Valley, Skylab Industries, ST Microelectronics, Thalès R&T, Univ. Joseph Fourier.
  • SILKAN : Contributed by developing advanced code-to-code compilation to transform sequential C code into parallel end code optimized at low level for different types of target processors.

SIMILAN

Dec. 2010 – Jun. 2013 ; FUI 10

  • The Project : Aims at providing a framework for professional-friendly parallel implementation of radar signal processing algorithms on multi-core and GPU architectures.
  • Partners : Thalès Air Systems , Dassault Aviation, Thalès R&T, DXO Labs, IS2T, Kalray, Scilab, ONERA/DPRS, Supélec/SDR4all, Univ. Paris Descartes.
  • SILKAN : Contributed by introducing COLD for automatic generation of parallel code within the frame of SPEAR descriptions of algorithms dedicated to radar signal processing

RICHELIEU

richelieu 120x60 Nov. 2012 – Nov. 2014 ; FUI 14

 

  • The Project : Aims at providing SCILAB with top rank run time performances through the development and integration of a “Just In Time” compiler.
  • Partners : Scilab Enterprises , OCamlPro, Dassault Aviation, ArcelorMittal, CNES, UPMC, INRIA
  • SILKAN : Contributes by implementing a fully Scilab-compliant COLD compiler that allows to generate on the fly an optimized run-time implementation of the Scilab models

Media GPU

ANR 108x47 Oct. 2009 – Dec. 2012 ; ANR

 

  • The Project : Aimed at developing a software architecture and adapting a number of classical multimedia algorithms (focusing on real time geometric operations), considering the latest advances offered by the new hardware architectures, such as Hybrid CPU+GPUand GPGPU.
  • Partners : PLAY ALL , ATEME, Labri, Institut Telecom.
  • SILKAN : Contributed , as a Hybrid Computing specialist, by taking the responsibility of the Platform Architecture, and the Prototyping environment and Tools.

Open GPU

open gpu 120x41  Dec. 2009 – Sep. 2012 ; FUI 8

 

  • The Project : Aimed at building both an open integrated platform gathering tools facilitating parallel GPU porting of existing code, and a set of S/W and H/W architectures convenient for an efficient usage of the GPU-based computing.
  • Partners : Wallix , AS+, BULL, Numtech, Ter@tec
  • SILKAN : Contributed by providing automatic parallelization tools based on the semantic analysis capacities embedded in the Par4All technology.

SCALOPES

scalopes 70x73  Jan. 2009 – Mar. 2011 ; ARTEMIS

  • The Project : Worked on efficient programming models suited for cross-domain technology (communication infrastructure, surveillance systems, smart mobile terminals and stationary video & entertainment systems), targeting tool developments for multi-core architectures
  • Partners : NXP Semiconductors Netherland B.V., plus 39 other partners.
  • SILKAN : Contributed by developing a software suite for an automatic OpenMP based parallelization for multi-core processors, taking into account the specificities of the SCMP system of CEA.

ACM

logo ACM Aug. 2015 – Sept. 2017 ; ARTEMIS

 

  • The Project :Aims at acccelerating the TRL assessment of advanced embedded signal processing applications through an automatic generation of a real-time implementation of their MATLAB algorithmic specification.
  • Partners :Thales Research & Technology, with involvement of the Air Systems and Optronics Divisions.
  • SILKAN : Contributes by coordinating the project, and developing an optimized performance-driven code generation capability to be implemented in the COLD compiler.

Back to top

2. Distributed Real Time Synchronization

Modeling large and complex real time systems on distributed architectures requires mechanisms to synchronize various software models and/or hardware devices under the most challenging conditions. To achieve a reliable and high-performing data transfer and synchronization in such systems, SILKAN has developed a software bus suitable for “man in the loop” simulators, as well as a hardware bus based on ARION’s technology for more bandwidth demanding applications (such as test and integration benches, etc.). Based on those technologies, SILKAN can deliver the safest and mots reliable systems.

AGREGATION

Feb. 2011 – Aug. 2013 ; FUI 10

  • The Project : Aims at delivering a framework where both virtual and real devices can be aggregated as a coherent mecatronic system, to either develop command laws using numerical models as virtual devices, or qualify them with actual devices plugged in.
  • Partners : SDI , ENSEA, EISTI, Scilab
  • SILKAN : Contributed by providing ARION real time technology, automatic code generation from XCOS descriptions, and numerical solvers for multiphysics models.

CETRAC

feder 70x67 Nov. 2011 – Nov. 2013 ; FEDER 12

 

  • The Project : Aims at delivering an infrastructure to design, simulate and certify safe and dependable Ethernet-based network architectures, ensuring deterministic data transfer and native segregation of data fluxes.
  • Partners : E NSTA , RTaW
  • SILKAN : was leader of this project, where Arion technology and know-how gave birth to a highly innovative, fully hardware, deterministic real-time Ethernet switch.

DEPARTS

Logo DEPARTS Jan. 2012 – Jan. 2017 ; FUI 14 – BGLE

  • The Project : Aims at building a development platform for the design of safe real time applications, certification compliant in distributed multi-core context. Aims at acccelerating the TRL assessment of advanced embedded signal processing applications through an automatic generation of a real-time implementation of their MATLAB algorithmic specification
  • Partners : CS SI , Consortium DDASCA, SNCF, CLEARSY, Critical Systems, RTaW, Nexter, Scilab Enterprises, Tronico, ENSAM, ENSTA, ESEO, INRIA, CNES Lanceurs, EDF, Wind River.
  • SILKAN : will deliver the key software elements the platform will be built upon.

EMC2

Logo EMC2 Apr. 2013 – Mar. 2017 ; ARTEMIS – AIPP5

  • The Project : Aims at delivering a sustainable service-oriented architecture for mixed criticality applications in dynamic and changeable real-time environments. EMC2 is the largest ARTEMIS initiative to date on this topic
  • Partners : Infineon (D) and 98 other partners from 16 European Contries. The French sub-consortium includes Thales, CEA, Freescale, INRIA, Kalray, Magillem, Rockwell Collins, ENS, CNRS and the Systematic competitivity cluster
  • SILKAN : will address the model-based co-design of muticore SoCs and software applications under dependability constraints.

NADAE

Jun. 2011 – Jun. 2013 ; RAPID-DGA

  • The Project : Aims at building VHDL/Verilog IP with protection mechanisms at architectural level against atmospheric neutron particles (SEU)
  • Partners : ENSTA
  • SILKAN : Contributed by managing the project and designing and architectural protection mechanisms against SEU.

SCARLETT

logo scarlett May. 2008 – May. 2012 ; FP7

  • The Project : Aims at building new avionics platform (IMA2G concepts) which is scalable, reconfigurable fault-tolerant driven and secure new avionics platform, namely DME: Distributed Modular Electronics
  • Partners : 40 partners including all the major European aeronautical companies.
  • SILKAN : Defined a simulator architecture and middleware components supporting ARINC 653 Operating Systems. Provided specific module to allow implementation of physical I/O (ARINC 429, CAN, Wired I/O …).

Back to top

3. Technologies for Accurate and Fast Simulation of Physical models

High fidelity simulations of systems require the modeling of the real world physics involved in order to provide an accurate interaction with 3D dynamic virtual reality. Furthermore, when embedded in sophisticated modern simulators (flight simulators, for instance), these simulations require the capability to manipulate simultaneous flows of large amounts of data for complex scene creation and management. To this end, SILKAN takes opportunity of Collaborative Research Projects to support the development and validation of innovative technologies in the fields of high performance simulation, versatile remote rendering, order reduction for non-linear physical models, etc.

ACDC

ACDC 70x66 Jun. 2010 – Nov. 2012 , ITEA 2

 

  • The Project : Aimed at exploring the feasibility of a new generation of on demand augmented video services delivered from the cloud.
  • Partners : Thomson Video Networks , plus 21 others (Finland, France, Luxembourg and Turkey).
  • SILKAN : Contributed by implementing a parallel distributed service for on the flight transcoding, based on the distributed file system and MapReduce approach provided by Hadoop.

Culture3D Clouds

logo C3DC Oct. 2012 – Mar. 2016 ; FUI 14

  • The Project : Aims at building a cloud-based platform dedicated to 3D scanning, documentation, preservation and diffusion of cultural heritages (sculpture, architecture, etc.).
  • Partners : RMN , CMN, CNRS, IGN, INRIA, CEA, Institut Telecom, ESITI, BeIngenious, Reciproque.
  • SILKAN : Will contribute by implementing high performance geometric processing algorithms.

CSDL

Oct. 2009 – Oct. 2012 ; FUI 7 csdl 108x61

 

  • The Project : Aimed at building a general framework architecture to support the whole of highly complex design processes such as those developed in the aeronautics.
  • Partners : Dassault Aviation , MBDA, EADS, Thalès, Distene, ESI Group, Oxalya, plus 18 other partners.
  • SILKAN : Contributed by developing a complete hosting infrastructure for a generic purpose pre- and post-processing station that can be seamlessly connected to the CSDL distributed framework to provide real time rendering of the simulations.

ICARE

logo ICARE Oct. 2012 – Mar. 2015 ; ITEA 2

  • The Project : Aims at developing a high performance infrastructure to technically support an innovative cloud based business model dedicated to on demand delivery of enriched and customized TV services. ICARE can be seen as an extension of ACDC.
  • Partners : Thomson Video Networks , plus 21 others (Finland, France, Luxembourg, Spain and Turkey).
  • SILKAN : Developped erOCCI, a framework to generate OCCI-compliant web applications, and extended OCCI for the specification of image processing API, with direct application of such cloud-oriented technologies to Silkan’s training simulators’ design.

HPAC

 

ANR 108x47 Jan. 2012 – Jan. 2016 ; ANR

 

  • The Project : Aims at building reference libraries for intensive algebraic computations (including exact arithmetics), that are efficiently suited to many-core architectures, for the sake of enhanced cryptographic systems.
  • Partners : CNRS, Université Joseph Fourier, UPMC/LIP6
  • SILKAN : Contributes by providing technologies for the automatic analysis and optimized implementation of codes based on so-called Domain Specific Languages (DSL) on graphical accelerators.

MECASIF

logo MECASIF-300x120 Sept. 2013 – Sept. 2016 ; FUI 15

  • The Project : aims at providing the industry with advanced, high fidelity Reduced Order Models (ROMs) fitted to real life engineers’ concerns. This will be achieved by developing innovative methodologies, beyond the limitations evidenced by CSDL for most of the common existing approaches. The bulk effort will address fundamental but complex transient phenomena in solid and fluid mechanics, including those driven by strong nonlinearities.
  • Partners : ArcelorMittal, Dassault Aviation, Renault, Snecma, Bertin Technologies, ESI Group, CADLM, DPS, FW4SEA, Scilab Enterprises, Structure Computation, ARMINES, INRIA, ECL/LTDS, ENS/LMT, I2M/TREFLE, UPMC/LJLL
  • SILKAN : Contributes by coordinating the project, developing model reduction for impact dynamics and supporting an optimized implementation in Scilab.

So HuSim

 

ANR 108x47 Oct. 2010 – Oct. 2014; ANR

 

  • The Project : Aims at developing numerical models for the accurate simulation of human / objects that interact through soft surfaces (like the skin / bed contact, for instance, which can lead to bedsores), taking into account the real physiological data
  • Partners : INRIA , Fatronik France, CHU Montpellier.
  • SILKAN : Contributed to the high performance implementation of the complex contacts algorithms.

XLcloud

 

xl_cloud 120x34 Jan. 2012 – Dec. 2015 ; FSN Cloud 1

  • The Project : Aims at designing, developing and integrating the software elements of a High Performance Cloud Computing (HPCC) system.
  • Partners : Bull , Institut Telecom, Artemis, R2SM, ESITI, ATEME, OW2, INRIA, CEA/LIST
  • SILKAN : Contributed by setting up cloud-based, resilient technologies applicable to remote rendering of scenes generated by HPC simulation of complex physical phenomena

Back to top