Node-level performance optimization @ CSC

22 May 2019, 09:00 → 24 May 2019, 16:00 Europe/Helsinki

Training room Dogmi, 1st floor (CSC - IT Center for Science)

Description

This course includes topics on code optimization for x86 platforms and efficient code parallelization on node-level using OpenMP threading. Advanced aspects of threading and optimization, such as selected topics of new features in OpenMP 5.0 standard, will be covered during the course.

Learning outcome

• Awareness of modern features in x86 CPUs
• Ability to vectorize computations
• Ability to use advanced features of OpenMP
•  Ability to increase code performance using threading and x86 optimization

Prerequisites

• Good knowledge of C/C++ or Fortran
• Good knowledge of threading using OpenMP
• Basic knowledge of modern CPU architectures

Agenda (preliminary)

Day 1: Wednesday, May 22

• Course introduction and intro to CSC's new systems
•  Introduction to the latest Intel(r) Xeon(r) Scalable Processor
•  Performance analysis methods and tools
•  Vectorization using SIMD

Day 2: Thursday, May 23

• OpenMP SIMD vectorization
• Optimizing memory accesses

Day 3: Friday, May 24

• Advanced OpenMP features
• OpenMP performance considerations
• Future outlook

Lecturers: 

Sebastian von Alfthan (CSC), Mikko Byckling (Intel), Fredrick Robertsén (CSC)

Language:  English
Price:          Free of charge

Node-level po 22.-24.5.19_hand out.pdf