course: Computational Methods on GPU

number:
141375
teaching methods:
lecture with tutorials
responsible person:
Prof. Dr. Ralf Peter Brinkmann
lecturer:
Dr. Denis Eremin (ETIT)
language:
english
HWS:
3
CP:
4
offered in:

Exam

All statements pertaining to examination modalities (for the summer/winter term of 2020) are given with reservations. Changes due to new requirements from the university will be announced as soon as possible.

Date according to prior agreement with lecturer.

Form of exam:oral
Registration for exam:FlexNow
Duration:30min

goals

The students know selected computational methods, their massively parallel analogs and basic programming techniques on graphics processing units (GPU).

content

  • GPU as a modern means for general-purpose massively parallel computations

  • General GPU architecture and CUDA operational model

  • Basic CUDA syntax

  • Optimization strategies in GPU programming

  • General guidlines for the algorithm parallelization

  • Prefix Scan

  • Massively parallel

    matrix algebra algorithms, sorting algorithms, random number generation algorithms, Monte-Carlo algorithms, Fast Fourier Transform algorithms, etc...

requirements

none

recommended knowledge

Basic programming skills with C/C++/Fortran programming languages