As a guest user you are not logged in or recognized by your IP address. You have
access to the Front Matter, Abstracts, Author Index, Subject Index and the full
text of Open Access publications.
We report on how we are preparing a seismic imaging code for the Intel Knight's Landing (KNL) Xeon Phi processor. The seismic imaging code in question is a Reverse Time Migration (RTM) code based on an unconventional rotated staggered grid (RSG) method and has undergone a significant amount of Xeon/Xeon Phi-focused re-engineering at ICHEC. The code uses an explicit finite difference (FD) scheme of variable spatial order, and second order in time, to model wave propagation in three isotropy cases. The code is parallelized with MPI and OpenMP, for wide hardware compatibility. Vectorization and efficient cache utilization were carefully considered in the kernel design, while attempting to maintain portability and maintainability. The stencil-based kernels of the code have low arithmetic intensity and are bound by the main memory bandwidth of the Xeon/Xeon Phi, which we have successfully alleviated by making highly efficient use of shared last level cache (LLC) on Xeon. The latter optimization is one that throws up some interesting challenges for achieving optimal performance on Xeon Phi, which we discuss further in this report.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.