Parallel Computing for Computational Mechanics
Lecture „Parallel Computing for Computational Mechanics“ is an elective course suitable for students in the Master programs „General Mechanical Engineering“ and „Computational Engineering Science“ and is a mandatory course in the second semester of Master study program „Simulation Sciences“
Contact
Calendar
Lecture: Monday 10:30am to 12pm in lecture room GRS 001
Exercise: Tuesday 1:15pm to 2:00pm in lecture room GRS 001
Office hours: to be announced
Instructors
Lecture: Prof. Marek Behr, Ph.D.
Exercises: Max Schuster, M.Sc., Felipe González, M.Sc.
Content
This course deals with the general structure of software in continuum mechanics: finite elements, finite differences or finite volumes in structural mechanics or fluid mechanics. The students become familiar with the latest scalar optimization strategies in cache-based microprocessors, develop a sense for potential bottlenecks which can greatly reduce the performance of a computer and learn how to prevent these problems.
The reasons for parallel computing in engineering sciences will be explained and put in context with the latest developments in software and hardware. In order to get parallel acceleration for a typical computation in continuum mechanics, the students learn about the OpenMP and MPI standards.
The module consists of a lecture (2 SWS) and an exercise (1 SWS), and carries 4 ECTS points as an elective course and 6 ECTS points for students in the Master program „Simulation Sciences“.
Topics
- computational mechanics code
- computer performance
- scalar optimization
- introduction to parallelism
- review of loop-level parallelism
- review of task-level parallelism
- good software engineering practices
- parallel computational mechanics code
- future of supercomputing
Examination
Oral exam will be offered on several alternate dates in August and September.