Courses

1. Shared memory computer architectures and OpenMP

Lecturer: Dr. Thomas Bönisch, Stuttgart High Performance Computing Center (HLRS)

OpenMP was proposed as a unified API for programming shared memory supercomputers in FORTRAN and C. The latest version of the specification OpenMP 4.0 introduces support for computational accelerators such as graphics processors. Implementations supporting the latest specification will become available soon.

Familiarity with C or FORTRAN programming is a prerequisite.

2. CUDA Programming

Lecturer: Florian Seybold, Stuttgart High Performance Computing Center (HLRS)

CUDA became an established approach to program NVIDIA graphical processors. CUDA-accelerated codes can run on many GPU-clusters equipped with NVIDIA processors and ordinary workstations with NVIDIA graphics cards. Many open source codes for scientific applications already can be built with CUDA acceleration, other codes need considerable rework to support CUDA.

Familiarity with programming in C is required.

3. Message Passing Architectures and MPI Programming

Lecturer: Dr. Thomas Bönisch, Stuttgart High Performance Computing Center (HLRS)

MPI stands for Message Passing Interface and is the standard for message passing as defined by the MPI-Forum. MPI is available on most of the computational clusters available at research facilities around the world including clusters in Novosibirsk State Univeristy, Siberian Supercomputing Center, and the Institute of Computational Technologies. MPI applications run on top of different interconnection networks including Ethernet and InfiniBand. Within an MPI application multiple nodes communicate through message passing, thus the state of the computation can span the memory of all the nodes, permitting large scale computational models to reveal finer details of physical processes.

Familiarity with C or FORTRAN programming is a prerequisite.

4. Partitioned Global Address Space and GASPI standard

Lecturer: Dr. Thomas Bönisch, Stuttgart High Performance Computing Center (HLRS)

GASPI is an emerging standard for Global Address Space Programming Interface. The goal of the standardization effort is to define a basis for future PGAS tools development for the HPC Community. This course will be of interest to early adopters and those who are interested in PGAS programming tools.

Familiarity with programming in C is required.