Diese Website verwendet nur technisch notwendige Cookies. In der Datenschutzerklärung können Sie mehr dazu erfahren.

Zum Hauptinhalt springen
Logo, Startseite der Hochschule der Medien

Modul Advanced Programming of Massively Parallel Processors

(Advanced Programming of Massively Parallel Processors)

Modul:143112 Advanced Programming of Massively Parallel Processors (Wahlpflichtmodul im Grundstudium), Schwerpunkte: Interactive Media, Usability and Games, Software Technology und Engineering
Modulverantwortlicher:Prof. Dr. Stefan Radicke
ECTS-min./max.: 5 / 5
Workload: 5 ECTS x 30 = 150 hours
Modulprüfung: LA
Formale Zulassungsvoraussetzungen: laut SPO
Kompetenzprofil
Lernergebnisse:Students learn how to utilize the Compute Unified Device Architecture (CUDA) programming model in order to leverage the potential of modern massively parallel Graphics Processing Unit (GPU) architectures in advanced software applications and in research projects.

Students deepen their knowledge and understanding by analyzing, presenting and discussing several parallel patterns and algorithms in small groups. These include convolution, stencil, parallel histogram, reduction and minimizing divergence, prefix sum (scan), and merge.

Students further hone their research and implementation skills by realizing one or more advanced parallel patterns in a programming project. These patterns may include, but are not limited to, sorting, sparse matrix computation, graph traversal, deep learning, iterative MRI reconstruction, and electrostatic potential map. Experimental findings about the project, with special emphasis on performance and scalability, form the basis of a concluding research paper.

Learning goals:

  • In-depth understanding of massively parallel hardware and software architectures.
  • Evaluate and apply common parallel algorithms and patterns.
  • Demonstrate mastery of at least one advanced parallel algorithm or pattern.
  • Ability to realize programs based on these algorithms to solve scientific problems.
  • Ability to measure, analyze, and optimize the performance of such programs.
  • Ability to generate and critically interpret these data in order to make scientific contributions in the field.
Weitere Angaben einblenden
Voraussetzung für dieses Modul: -
Dieses Modul ist Voraussetzung für:-
beinhaltet folgende Lehrveranstaltung(-en):
EDV-Nr Lehrveranstaltung Art  SWS  ECTS  Prüfungsform
143112a Advanced Programming of Massively Parallel Processors - 2 5
* kennzeichnet Prüfungsvorleistungen