Lecture (Other)

FZJ-2021-04345

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png PRACE Training Course: Directive-based GPU programming with OpenACC

Herten, A. (Corresponding author)FZJ* ; Hater, T.FZJ* ; Haghighi Mood, K.FZJ* ; Hrywniak, M.FZJ* ; Kraus, J.FZJ*

2021

Please use a persistent id in citations: http://hdl.handle.net/2128/29058

Abstract: GPU-accelerated computing drives current scientific research. Writing fast numeric algorithms for GPUs offers high application performance by offloading compute-intensive portions of the code to the GPU. This PRACE Training Course hosted at JSC will cover basic aspects of GPU architectures and programming. Focus is on the usage of the directive-based OpenACC programming model which allows for portable application development. Examples of increasing complexity will be used to demonstrate optimization and tuning of scientific applications.Topics covered will include: Introduction to GPU/Parallel computing; Programming model OpenACC; Interoperability of OpenACC with GPU libraries (like cuBLAS and cuFFT) and CUDA; Multi-GPU Programming with MPI and OpenACC; Tools for debugging and profiling; Performance optimizationThe course consists of lectures and interactive hands-on sessions in C or Fortran.

Note: PRACE Training Course held online.

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 5121 - Supercomputing & Big Data Facilities (POF4-512) (POF4-512)
2. 5112 - Cross-Domain Algorithms, Tools, Methods Labs (ATMLs) and Research Groups (POF4-511) (POF4-511)

Appears in the scientific report 2021

---

Database coverage:

---