08:15 – 09:15 Registration with light breakfast

09:15 – 09:30 Welcome & Introduction

09:30 – 10:15 Parallelism, Performance & Optimization on Intel Architecture

Starting with a brief overview of the latest Intel silicon roadmap we look at how you can use Intel Parallel Studio XE 2018 to get
best performance on both the new Intel® Xeon® Scalable Processors (Purley / Skylake-SP) as well as the Intel® Xeon Phi™
processor family (Knights Landing and Knights Mill). We then discuss three key topics (Vectorization with AVX512, Threading, and
Memory) that you need to address when modernizing code.

Stephen Blair-Chappell

10:15 – 11:15 Practical Session 1: Using Intel Parallel Studio to answer the question “why is my program running so slow”?

In this session, we use three Intel tools, Intel® Trace Analyzer and Collector, Intel® VTune Amplifier XE, and Intel® Vectorization
Advisor to track down the reasons for slow running code in a Lattice Quantum Chromodynamics (LQCD) code. The example is
based on a real problem reported by the HPC community.

Stephen Blair-Chappell

11:15 – 11:45 Coffee break

11:45 – 12:30 Striding Towards Perfection – A step-by-step narrative on optimizing the k-means algorithm

A look at how code modernization techniques are being used in the scientific community to produce code that takes best
advantage of the latest generation of CPU hardware. In this session we improve the performance of the k-mean clustering
algorithm written in C++ by first working on the vectorization followed by improving the threading of the code. The final version
is benchmarked on latest generation of Intel Xeon and Intel Xeon Phi.

12:30 – 13:30 Practical Session 2: Tuning Vectorized Code Using Intel Vector Advisor

In this session, we show how to use Intel Vector Advisor to check how well your code is being vectorized and using the latest architecture available such as AVX512. Additionally, we look at various memory issues, such as non-contiguous memory accesses and unite stride vs. non-unit stride accesses, and how eliminating such issues can lead to significant speed up of vectorized code and improve the quality of code generated automatically by the compiler.

13:30 – 14:30 Lunch break

14:30 – 15:30 Using the Intel Compiler to create portable applications

In this session we take a close look at how you can use the Intel compiler to bring performance and portability to your vectorized
applications. We show how you can take full advantage of the latest instructions sets – such as AVX512 – and yet create
programs that can still safely run on earlier generations of CPU.
Additionally, we describe some of the recent compiler options supported by the latest version the Intel compiler that improve the
reproducibility of floating point results.

15:30 – 15:45 Coffe break

15:45 – 16:30 Optimizing Python Code using the Intel Distribution of Python

It used to be the case that you would never use the words ‘performance’ and ‘python in the same sentence. The Intel distribution
of Python changes all that. In this first of a two-part session we show how you can speed up you Python codes using by
‘Cythonising’ your Python code to achieve native performance.

Stephen Blair-Chappell

16:30 – 17:30 Practical Session 3: Roofline analysis Intel Vector Advisor

Learn how to run a Roofline Analysis using Intel Vector Advisor. The Roofline model combines locality, bandwidth, and different
parallelization paradigms into a single performance figure that shows the performance of the code under test.

17:30 – 19:00 – Networking Cocktail with drinks and finger food