Wake up, Igor… The SoHPC has you
You are reading this text because either you know something about the SoHPC programme or you already know something about me. In both cases this article aims to give you a full puzzle of who I am and what is the programme I mentioned before.
Photo of me in a pre-virus era
My name is Igor Abramov and currently I am studying on a Computational Science and Engineering master program in École Polytechnique Fédérale de Lausanne. During my bachelor I had a lot of courses on numerical methods and projects with usage of finite difference/elements/volume methods and I was always curious about how to boost them and make implementations more efficient. In my bachelor project for the first time I tried to parallelize python code with some high level libraries, but it didn’t give me a deep understanding of how it works. That’s why I started to attend related courses about parallel and high performance computing in university and decided to apply for PRACE Summer of HPC.
What is Summer of HPC?
Summer of HPC is a PRACE programme that offers summer placements at HPC centres across Europe to late stage undergraduates and/or master students. Applicants from across Europe are selected to participate each year, during the registration each candidate should specify the three most interesting projects for him/her. With a usual timetable students are gathering for the first training week and then moving to the HPC centre to which they were accepted till the end of August to do project.
Now with the post-virus era it is still impossible to organize this in a usual format so that’s why the organizing committee decided not to cancel the programme (huge thanks for that!) but to move it online! I think that it was the best possible decision in such a complicated situation.
That’s why all of the participants virtually attended Vienna for a training week from 1 to 4 of July. We had lectures about computations and architecture of clusters in general, lectures about OpenMP, MPI and CUDA API and a lot of practical assignments. Also we had some online meetings with other participants and project mentors. It can’t replace live meetings and project collaboration but it is the safest possible way to spread knowledge and cooperate in current circumstances.
Me and my colleague Josip were accepted for a project about Fast Multipole Method implementation for GPUs in a Jülich Supercomputing Centre (JSC). This method is a numerical technique that was developed to speed up the calculation in the N-body problem. What are these N-bodies and how to juggle with them as fast as possible you will learn in a next post!