Lunch & Learn: Optimizing fusion particle codes for massively parallel computers

The last decade has wit­nessed a rapid emer­gence of larger and faster com­put­ing sys­tems in the US. Mas­sively par­al­lel machines have gone main­stream and are now the tool of choice for large sci­en­tific sim­u­la­tions. Keep­ing up with the con­tin­u­ously evolv­ing tech­nol­ogy is quite a chal­lenge though. Sci­en­tific appli­ca­tions need to be mod­i­fied, adapted, and opti­mized for each new sys­tem being intro­duced. In this talk, the evo­lu­tion of a gyro­ki­netic particle-in-cell code devel­oped at Prince­ton University’s Plasma Physics Lab­o­ra­tory is pre­sented as it was adapted and improved to run on suc­ces­sively larger com­put­ing platforms.

About the speaker: 

Dr. Stephane Ethier is a Com­pu­ta­tional Physi­cist in the Com­pu­ta­tional Plasma Physics Group at the Prince­ton Plasma Physics Lab­o­ra­tory (PPPL). He received a Ph.D. from the Depart­ment of Energy and Mate­ri­als of the Insti­tut National de la Recherche Sci­en­tifique (INRS) in Mon­treal, Canada. His cur­rent research involves large-scale gyro­ki­netic particle-in-cell sim­u­la­tions of micro­tur­bu­lence in mag­netic con­fine­ment fusion devices as well as all aspects of high-performance com­put­ing on mas­sively par­al­lel systems.

Down­load the pre­sen­ta­tion slides (.pdf)
Lis­ten to the pod­cast (.mp3)