Workshop on programming of parallel architectures

(german version)

On February, 25th at 4pm (Mainbuilding, F435) we will have a workshop about programming on parallel computer architectures.

In this workshop, common concepts of modern parallel computer architectures will be discussed and accompanied by concrete examples. By using recent technologies, such as SIMD, Multi-Threading and OpenMP, programs can be sped up considerably, e.g., allowing for interactive frame rates in virtual reality applications or physical simulations.

Focus of the workshop will be application development on the Welfenlab's own IBM cell cluster, which has a theoretical maximum computing power of 5 TFlops. This enormous computation power will permit the real time computation of physical models and thus enable realistic interaction in virtual reality systems.

The workshop is open for everyone interested and addresses especially students who plan to write their thesis in the field of virtual reality / haptic interaction / ... at our institute.

Contents

• Concepts of parallel architectures
• x86 architectures
  • SIMD, SSE3 intrinsics
  • OpenMP and Intel TBB
• Cell architecture:
  • controlling the SPUs
  • using SIMD
  • memory access and data transfer via DMA
• MPI for distributed architectures
  • Applications
  • fundamental operations: Scatter, Gather, Allreduce

	Attachment	Size	Date Added

Slides (PDF)