High Performance Computing

Prof. Dr. Gundolf Haase

Abstract

This course introduces into the field of High Performance Computing (HPC) in mathematical areas and engineering sciences. The students will acquire specialties of recent and future hardware concepts as well as on supported software and compilers. The course work will be organized such that all course topics will be implemented on the appropriate hardware ranging from a single CPU via multiple CPUs to clusters of CPUs and GPUs/Xeon Phi. The students will be able to adapt research specific code such that they can take advantage of available computer resources. The four main goals of the course consist of

• Knowledge of the students on algorithms and data structures for HPC and active use of this knowledge.
• The students get in touch with HPC related concepts and architectures, and the students are able to adopt new developments in this area onto the problem under consideration.
• Standard compiler and software support for parallel computer architectures is known and used by the students for solving mathematical problems by means of HPC hardware.
• The students are able to write/adapt parallel programs on various parallel platforms.

Schedule

• 17.07.2014
  • 09:00 - 10:30: Intro into vectorization and parallelization
  • 10:45 - 12:15: Shared Memory Parallelization
  • 13:00 - 14:30: First steps with vectorization
  • 14:45 - 16:15: First steps with OpenMP
• 18.07.2014
  • 09:00 - 10:30: Accelerator programming, some hardware, some CUDA
  • 10:45 - 12:15: Pragma driven parallelization with OpenMP4.0/OpenACC
  • 13:00 - 14:30: First steps in Xeon Phi programming
  • 14:45 - 16:15: First steps in GPU programming
• 19.07.2014 optional
• 21.07.2014
  • 09:00 - 10:30: Distributed memory computing: MPI
  • 10:45 - 12:15: First steps in MPI
  • 13:00 - 14:30: Mixing distributed and shared memory computing
  • 14:45 - 16:15: First steps in hybrid parallel computing
• 22.07.2014
  • 09:00 - 10:30: How to accelerate my own code?
  • 10:45 - 12:15: Work on production/example code
  • 12:30 - 14:00: Work on production/example code

Place

TUM, Zentrum Mathematik, Room MI 03.08.020

Registration

Registration is required due to the limited number of participants.
For binding registration please send an email to puchert@ma.tum.de