After successful completion of the course, students are able to use modern parallel processor architectures efficiently and understand their strengths and weaknesses. Students will get a deep understanding of the main workloads in computational science and how many-core architectures can be leveraged to run larger problems in less time than with conventional approaches.

Topics covered in this course include:

• Ahmdal's Law
• FLOPs, Bandwidth, and Latency
• Performance Modeling
• Graphics Processing Units (SIMT processing, thread block synchronization)
• Programming Models (Annotation-driven such as OpenMP, native such as CUDA)
• Field Programmable Gate Arrays
• Emerging Many-Core Architectures

Presentation of new material via prepared videos for self-study each week. Each week there is an interactive one-hour-lecture (hybrid format) to answer questions and assert clarity. There are hands-on exercises for students to familiarize themselves with the new material in between lectures. Submission of short reports by the students. Review of student submissions, discussion of problems.

Virtual oral exam after positive evaluation of the practical part. Registration in TISS.

Please register via TISS in order to participate in the hands-on exercises.

Slides from the lecture will be made available for download.

Familiarity with at least one programming language (e.g. C or Python)