DCSE Summerschool: Numerical Linear Algebra on High Performance Computers

The Delft High Performance Computing Center is organizing an intensive week of lectures and hands-on training given by local and international experts in HPC and numerical linear algebra. The content covers current topics in HPC such as task-based and GPU programming, performance engineering and communication-avoiding algorithms. Participants will get a chance to interact with teachers and peers in an informal setting, and experiment with the content taught in extensive lab sessions on the DelftBlue supercomputer. In order to facilitate this in-depth training, the number of seats is limited.

Topics covered

• Dense Linear Algebra and Applications (Prof. Paolo Bientinesi, University of Umeå, Sweden)
• Performance of Sparse Matrix Operations (Prof. Gerhard Wellein, University of Erlangen-Nuremberg, Germany)
• Reducing Communication (Prof. Laura Grigori, EPFL, Switzerland)
• Randomized Algorithms in Linear Algebra (Laura Grigori)
• Introduction to GPU programming (Kees Lemmens, TU Delft)
• Software Abstractions for Linear Algebra (Jonas Thies, TU Delft)
• Recent Developments in Krylov Subspace Methods (Prof. Martin van Gijzen, TU Delft)
• Advanced Domain Decomposition Methods (Alexander Heinlein, TU Delft)
• Introduction to Quantum Computing (Matthias Möller, TU Delft)

Prerequisites

The summerschool is open to PhD students associated with the Delft Institute of Computational Science and Engineering (DCSE). Participants should be proficient in at least one of the programming languages C, C++ or Fortran, and have experience in working on a remote system using the command line. Some experience with OpenMP and MPI is recommended.