421 views
# 2021 DarkEra PhD <div align="center"><table><tr><td><img src="https://i.imgur.com/FjbvarH.png" alt="Context" height=50px/></td><td><img src="https://i.imgur.com/YNJmkva.jpg" alt="Context" height=80px></td><td><img src="https://i.imgur.com/04DvYvv.png" height=70px></td></tr></table></div> ## *Design and Programming of Low-Power High Performance Computing Systems in Astronomy* Keywords: `High Performance Computing`, `Heterogeneous Architectures`, `System Design`, `Dataflow`, `Models of Computation`. ###### tags: `VAADER` `PhD Thesis` `DARK-ERA` ## Context <p align="justify"> The exascale radio telescope Square Kilometer Array (SKA) will have to generate in real-time huge scientific products like hyperspectral images of the sky from a 7.2 Tb/s data stream. The SKA supercomputer will also have to be power-efficient with only 1 MWatt for 250 Petaflops. Such energy and computation requirements imply the computing system to be an innovative architecture based on a standard HPC system combined with Field Programmable Gate Arrays (FPGAs) or application-specific architectures like Graphical Processing Units (GPUs) or the manycore Kalray Massively Parallel Processor Array (MPPA). The challenge of the DARK-ERA ANR project is to develop efficient co-design methods and rapid prototyping tools to assess the performance both in time and energy of new complex scientific dataflow algorithms on not-yet-existing complex computing infrastructures. </p> <!--p align="center"><img src="https://i.imgur.com/vwP32UO.png"></p--> <p align="center"><img src="https://i.imgur.com/N5mtiJ4.png"></p> <p align="justify"> During this PhD, you will take part in the DARK-ERA project. You will collaborate with embedded system experts at Centrale/Supelec Paris, HPC experts at INRIA and Atos-Bull, and astronomers at the Observatoire de Paris and the Côte d'Azur University. The goal is to create a development flow aiming at easing the deployment of SKA applications on massively parallel architectures by astronomers. The proposed development flow is based on dataflow Models of Computation (MoC). Dataflow Models of Computation enable the specification of application as graphs, which make specification of application parallelism explicit and intuitive, with a complete abstraction from architectural concerns. </p> ## Objective <p align="justify"> The goal of this PhD is to update and upgrade PREESM, an open source rapid prototyping tool. PREESM simulates signal processing applications and generates code for heterogeneous multi/many-core embedded systems. Its dataflow language eases the description of parallel signal processing applications. </p> <p align="justify"> PREESM will be associated to SimGrid to create SimSDP. SimSDP will provide reliable simulations of large scale heterogeneous HPC systems to explore algorithm and architecture spaces in the Exascale SKA context. SimSDP will be used at design time when the SKA computing system is sized; but also all along its lifetime, for the early estimation of time and energy requirements of new algorithms to execute on the SKA computing system. SimSDP will be used by astronomers to deploy new algorithms at scale avoiding complex and error-prone hardware-dependent manual optimisations. </p> ## Skills * C/C++, Java * Embedded system programming or HPC programming * Parallel programming * English ## Characteristics * Duration: 3 years * Start: October 2021 * Salary : 1827€ net, 2272€ gross ## Location * Research team: VAADER team, IETR laboratory. * Address: INSA Rennes, 20 Avenue des Buttes de Coësmes CS 70839 35708 Rennes Cedex 7 France ## Supervisors * Jean-François Nezan (IETR, Equipe VAADER, Rennes) - [jnezan@insa-rennes.fr](jnezan@insa-rennes.fr) * Karol Desnos (IETR, Equipe VAADER, Rennes) - [kdesnos@insa-rennes.fr](kdesnos@insa-rennes.fr) ## Applications Send resume and application letter to [kdesnos@insa-rennes.fr](mailto:kdesnos@insa-rennes.fr?subject=Candidature%20Stage%20MPPA) ## Websites [1] PREESM : https://preesm.github.io [2] SimGrid : https://simgrid.org/doc/latest/ [3] Dark-Era : https://www.cnrs.fr/mi/IMG/pdf/astro-informatique2019_jr_presentation_skallas_gac_v_light.pdf