Merge branch 'job_19a7c686cfe81822fb762690116fb994' into 'master'

New job offer job_19a7c686cfe81822fb762690116fb994

See merge request !471

content/job_offers/job_19a7c686cfe81822fb762690116fb994.md

+Title: Parallelization and Tests of Computational Fluid Dynamics (CFD) code METAS
+Date: 2020-01-06 10:10
+Slug: job_19a7c686cfe81822fb762690116fb994
+Category: job
+Authors: BOBELIN Laurent
+Email: laurent.bobelin@insa-cvl.fr
+Job_Type: Stage
+Tags: stage
+Template: job_offer
+Job_Location: Bourges
+Job_Duration: 5 months
+Job_Website: 
+Job_Employer: PRISME laboratory
+Expiration_Date: 2020-03-10
+Attachment: job_19a7c686cfe81822fb762690116fb994_attachment.pdf
+
+**Context**
+Many scientific and industrial applications require much more processing power than a workstation. Among the most popular applications, numerical simulation of fluid mechanics (Computational Fluid Dynamics or CFD) is very demanding in terms of memory and computations. PRISME Laboratory has developed an application called METAS that simulates explosion waves and their interaction with structures. METAS has been coded in Fortran 90 and has been designed to use multiple computing resources but up to now uses only one CPU. Simulation duration of nowadays realistic use cases becomes very long (weeks or months). PRISME laboratory wants to refactor the code, based on the skills of LIFO lab, so as to let the code make full use of two kind of computing resources: 1) GPU 2) computing resources available at the regional computing center CaSciModOT.
+
+**Internship Description**
+The code has been designed for parallelization. A mesh discretizes the field of computations in cells, then uniform computations of flows are carried out for each face of the cell. The type of processing and the mesh used should allow a rather direct transposition on GPU or parallel multi-core architectures. The purpose of the internship is to explore the different possibilities of parallelization, and test the performances on large size practical applications. The internship tasks includes:
+- GPU parallelization using CUDA Fortran on a Titan Xp GPU.
+- Multi-core oriented parallelization using Coarray Fortran.
+- Performance tests and deployment on the Cascimodot cluster.
+- Case studies on real applications.
+**Candidate Profile**
+The candidate should be a Master 2 student or a 5th year engineering student in Computer Science or in Physics oriented toward numerical simulation and has the following skills:
+• knowledge in parallelism
+• parallel programming skills (CUDA)
+• Fortran is a plus, as is a basic knowledge of CFD.
+Speaking French is recommended for living in Bourges but not mandatory at lab.
+**How to Apply**
+If you are interested, please send us a short message detailing your motivation and a resume by
+mail. To apply and / or for any additional information request, please send an email to
+laurent.bobelin@insa-cvl.fr and mame.william-louis@univ-orleans.fr
\ No newline at end of file