Mündliche Prüfung (bei großer Teilnehmerzahl Klausur), durch erfolgreiche Übungen kann ein Notenbonus erarbeitet werden.

The module deals with the topic of parallel data processing from the perspective of computer architecture. Computer architecture concepts are presented, with the help of which parallelism can be exploited at various levels to increase performance. The module covers the following topics, among others: Machine instruction-level parallelism: superscalar technique, speculative execution, jump prediction, VLIW principle, multi-threaded instruction execution. Modern parallel computing concepts, memory-coupled parallel computers, symmetric multiprocessors, distributed shared memory multiprocessors, message-oriented parallel computers, multicore architectures, cache coherence protocols, performance evaluation of parallel computing systems, parallel programming models, interconnection networks (topology, routing), heterogeneous system architectures and GPGPUs.

The students have extended technical competences in the area of modern computer architectures with a focus on parallel architectures, interconnection networks and heterogeneous systems. They know the advantages and disadvantages of the various parallel architectures as well as the difficulties that arise when programming such systems. This enables the students to apply appropriate programming concepts for parallel architectures in a situation-appropriate manner. In the exercises, the participants acquire a further understanding of the complexity of parallel processes and the resulting difficulties. Due to the independent work in small groups, the ability to work in a team and leadership qualities are particularly promoted.

• J. L. Hennessy, D. A. Patterson: Computer Architecture: A Quantitive Approach, Morgan Kaufmann Publishers Inc, Elsevier, 6. Auflage, 2018.
• S. Pasricha, N. Dutt: On-Chip Communication Architectures; Morgan Kaufmann Publishers Inc., 2008.