Intel® HPC Developer Conference 2017

Hands-on Tutorials


Julia*: Petascale Made Simple
(1 hr 25 min)

This session introduces Julia* and its tools and libraries for making parallel, high-performance computing easy on a massive scale.

Keno Fischer, Viral Shah, and Ranjan Anantharaman, Julia Computing Inc.

A High-Performance Machine Learning Framework for HPC Cloud
Part 1 & Part 2
(55 min)

The two major trends in computing systems are the growth in high-performance computing (HPC) and the machine learning for big data phenomenon with an accompanying cloud infrastructure. This tutorial weaves these trends together using some key technologies of HPC and big data convergence in a scalable machine learning platform.

Langshi Chen, Mihai Avram, Supun Kamburugamuve, and Judy Qiu, Indiana University

Presentation (PDF)

Parallelize Python Applications with PyCOMPSs

Parallelize Python* Applications with PyCOMPSs

PyCOMPSs is an approach to support a Python task-based parallel programming model where the tasks’ data-dependences are inferred at runtime. The session presents a tutorial on how to parallelize Python applications with PyCOMPSs, including results on hybrid SSF clusters with KNL nodes.

Rosa M. Badia, Barcelona Supercomputing Center (BSC)

Presentation (PDF)

Tuning for the Intel® Xeon® Scalable Processor (Part 1 & Part 2)
(45 min)

Understand tuning opportunities using Intel® Parallel Studio XE 2018 and how it relates to code optimization. Focus on new build flags in Intel® compilers, optimizations for Intel® Performance Libraries with hands-on exercises exploiting latest features in Intel® VTune™ Amplifier and Intel® Advisor.

Carlos Rosales and Dmitry Prohorov, Intel
Zakhar Matveev, Aleksandar Ilic, INESC-ID/IST, University of Lisbon

Presentation (PDF)

Parallel Performance Evaluation Using the TAU Performance System
(1hr 26 min)

The TAU Performance System is a mature, portable, performance evaluation tool available on HPC platforms. It supports profiling as well as tracing for programs written in C++, C, Fortran, Python*, and Java* using MPI, OpenMP, Apache Spark*, pthread, and OpenCL™ code. This tutorial introduces TAU and related tools.

Sameer Shende, University of Oregon

Presentation (PDF)

GANS in Neon

Deep Learning for Fast Particle Simulation Using the neon™ Framework

This session discusses the application of convolutional generative adversarial networks to simulate particle energy showers in electromagnetic calorimeters. We present the development of our model in the neon™ framework, with detailed explanation of implementation and optimization steps to reproduce three-dimensional images of energy showers.

Sofia Vallecorsa, CERN

Presentation (PDF)

Imperial College

Optimized Symbolic Finite Difference Computation with Devito

This tutorial demonstrates Devito, a finite difference framework that allows users to solve partial differential equations from only a few lines of Python. Explore how to use automated code generation to execute highly optimized stencil code for a range of scientific problems.

Michael Lange, Navjot Kukreja, and Fabio Luporini, Imperial College London

Accelerate Algorithms on FPGAs Using the OpenCL™ Platform

Accelerate Algorithms on FPGAs Using the OpenCL™ Platform

FPGA technologies can be leveraged as an ideal custom coprocessor to boost the performance of your algorithms. In this session, explore how to accelerate algorithms on FPGAs using the OpenCL™ platform.

Karl Qi, Intel Corporation

Presentation (PDF)

Singularity: Containers for Science

Singularity: Containers for Science

Singularity containers have been gaining widespread adoption in both enterprise and scientific computing due to their ability to facilitate extreme portability and reproducible software stacks. Learn about the new features that make it even more applicable for science.

Gregory Kurtzer and David Godlove, SyLabs Inc.

Presentation (PDF)