Date :
05 December 2016 - 05 February 2017
Time of the event : 09:30 – 13:00
Event type :
Workshops
Description of the activity : OpenACC is a directive-based language, similar to OpenMP, that permits the parallelization of an existing code and its porting to accelerators (GPUs or others). Unlike CUDA, this language allows you to easily write portable code that will run equally well on accelerators and conventional CPUs. In this class, you will learn how to use OpenACC to port an existing code to GPU accelerators.
This workshop incudes a hands-on training sesssion so we recommend to bring a laptop with you to the class.
Prerequisites
This tutorial uses the programming languages C or Fortran, in a Linux environment. Therefore, there are a few prerequisites that you need to have before you consider attending:
-
Being able to login and interact (edit files, navigate directories) with a Linux computer in a command-line environment using SSH
-
A basic understanding of C or Fortran
Knowledge of parallel computing and GPU computing are notprerequisites for this class.
Lesson plan
-
Introduction to accelerator architectures
-
Profiling existing code and gathering compiler information
-
Expressing parallelism with OpenACC directives
-
Expressing data movement
-
Optimizing loops
Event location : Pavilion Roger Gaudry, Université de Montréal , V-221
2900 boul. Edouard-Montpetit
Montreal, QC H3T 1J4