Trainingskurs "Vectorisation and portable programming using OpenCL"
(Kurs-Nr. 88/2016 im Trainingsprogramm des Forschungszentrums)
Zielgruppe: | Scientists who want to use GPU systems, Xeon Phi co-processors, or many-core CPUs with OpenCL |
Inhalt: | |
Teilnahmevoraussetzungen: | Kenntnisse in Linux, z.B. make, command line editor, Linux shell, Erfahrung mit C |
Sprache: | Der Kurs wird auf Englisch gehalten. |
Dauer: | 2 Tage |
Termin: | 17. - 18. März 2016, 9.00 - 16.30 Uhr |
Ort: | Jülich Supercomputing Centre, Ausbildungsraum 2, Geb. 16.3, R. 211 |
Teilnehmerzahl: | mindestens 5, maximal 24 |
Referenten: | Andreas Beckmann, Willi Homberg, Ilya Zhukov, JSC; Prof. Dr. Wolfram Schenck, FH Bielefeld |
Ansprechpartner: | Willi Homberg |
OpenCL provides an open, portable C-based programming model for highly parallel processors. In contrast to NVIDIA's proprietary programming API CUDA, a primary goal of OpenCL is portability across a diverse set of computing devices including CPUs, GPUs, and other accelerators.
Topics covered will include:
- Introduction to vectorisation
- Programming model of OpenCL
- Datatypes and OpenCL vectorisation features
- Tuning for architectures like CPUs, accelerators (GPUs), and co-processors (Xeon Phi)
- Heterogeneous multi-device programming