mostly related to performance engineering at the node-level, i.e., excluding MPI
topology, affinity, pinning
hardware performance counter, patterns
roofline model
case studies
This will include an hands-on on the cluster systems of PC².
Please note:
The couse assumes:
basic knowledge of the Linux shell. We provide a short online selftest. In case you have no experience with Linux, we recommend that you attend our next Linux Introduction Course
basic knowledge of workload managers (esp. Slurm) and jobscripts. We provide a short online selftest. In case you have no experience with workload manager on clusters, we recommend that you attend our next HPC Introduction Course.
knowledge of HPC system architecture, profiling and scaling analysis. Additionally, a little bit of programming knowledge (Fortran or C++) is required. We provide a short online selftest. In case you have no experience with the architechture of modern HPC compute nodes, we recommend that you attend our next course on Advanced Topics of HPC.
The language of the course is planned to be English. If you don't feel comfortable enough with English we can make this course also in German at a later date upon request.
For the hands-on part of the course the participants should bring their own laptops. No special operating system is required. You can participate with Windows, MacOS or some Linux-distribution. Please make sure that the wireless connection of your laptop works (for example with eduroam). In case you don't have access to eduroam, we provide a an additional wireless network during the course.
This course will be free of cost for members of German universities or publicly-funded research institutions in Germany.