HPC Application Project

The aim is to study the different levels of parallelization of an HPC application: vectorization, multi-threading, distributed memory within a single application. The chosen application is the implementation of linear algebra kernels, themselves frequently at the heart of large-scale simulation and learning applications.

Knowledge of OpenMP and MPI

The project will take place over the course of the semester and will involve parallelizing each layer of a linear algebra kernel to achieve the best possible performance, and understanding what can impact the performance of a code, its energy consumption, its communication volume, ...