Higher-order finite difference stencils for improving efficiency of seismic simulation

Recruiter
Location
Cstjf, France
Salary
Competitive
Posted
17 Nov 2016
Closes
05 Jan 2017
Ref
5566BR
Sector
Oil and Gas
Contract Type
Permanent
Hours
Full Time
Job Description
The aim of the project is to investigate the impact on the HPC efficiency of the use of higher-order stencils, in space and/or time, in finite-difference solutions of the wave-equation for seismic imaging and inversion. The efficiency of present implementations is limited by the bandwidth of data transmission between the CPU and memory; it is hoped that higher-order stencils will allow bigger (and therefore fewer) time-steps with little or no increase in the time to calculate each time step despite an apparent increase in the number of operations.
This project is a second step following the work of a previous intern. The intern will initially perform a bibliographic review of the theory of finite-difference simulation and its application to definition of different stencils and the relevant coefficients. He will then familiarize himself with the instances presently implemented in Total's inhouse code platform, DIVA and with the work of the previous intern. He will test the methodology of the previous work and apply the knowledge of the coding methodology to write new codes for one or more finite-difference kernels implementing higher-order (initially in time) stencils. He will initially work on the isotropic acoustic wave-equation; if time permits he will extend his investigation to anisotropic acoustic and/or elastic kernels as directed by his supervisors. He will test and validate these in order to compare the efficiency of the new codes with the existing stencils (potentially on different HPC architecture). He will complete his internship by documenting his work and conclusions in a report and presentation.

Profile sought
Education : Engineering school i n Geophysics and/or Computing Sciences
Level of study : 2nd or 3 rd year of engineering school
Competencies: Rigor, communication, motivation, autonomy,..
Technical and linguistic skills: efficient communication in English
Background in geophysics with some basic knowledge on depth imaging tools. Skills in coding (mainly Fortran) and in Finite Differences approximations could be useful..