Seminar: Intel® Composer

Wednesday Apr 22, 2015, 1:30 PM → 5:00 PM Europe/Berlin

Raum 217 (KIT, Gebäude 20.21, Campus Süd)

Das Steinbuch Centre for Computing lädt Sie herzlich am 22. April 2015 zum Seminar "Intel® Composer" ein (13:30 - 17:00 Uhr).
Das Seminar wird Live-Vorführungen beinhalten. Praktische Übungen am eigenen
Notebook sind nicht vorgesehen. Bitte beachten Sie, dass die Teilnehmerzahl (50) des Seminars begrenzt ist. Es fallen keine Teilnahmegebühren an.

---

The Steinbuch Centre for Computing is hosting on April 22 2015 the seminar: "Intel® Composer (1:30 - 5:00pm). The seminar includes live demonstrations, but no hands-on.
Please note that the number of participants is limited to 50. No course fees apply.

Slides

• 00_-_Agenda.pdf
• 01_-_Intel®_Parallel_Studio_XE_2015.pdf
• 02_-_Intel®_Architecture_for_Software_Developers.pdf
• 03_-_Intel®_C++_and_Fortran_Compiler.pdf
• 04_-_Vectorization_for_Intel®_C++_and_Fortran_Compiler.pdf
• 05_-_Numerical_Stability_with_Intel®_C++_and_Fortran_Compiler.pdf
• 06_-_Intel®_Math_Kernel_Library.pdf

Welcome & Overview about Intel® Software Development Tools

Provides an overview about Intel® Software Development tools from Intel® Parallel Studio XE 2015 Cluster Edition. For developers not familiar with those suites, all components are briefly explained. For developers already working with some of the tools we also list the latest key features.

Intel® Architecture & Vectorization

Vectorization is an important aspect of Intel C++ and Fortran Compilers. Understanding the technical possibilities, limits and roadblocks of vectorization is crucial for compiling efficient and performance oriented applications. This module gives an introduction to the Intel architecture (IA32, Intel®64 & Intel® MIC) in terms of their vectorization capabilities (also known as SIMD features).

Intel® C++ & Fortran Compiler - Advanced Topics

Those training modules provide an introduction of the feature set of the Intel® Parallel Studio XE 2015 Cluster Edition tool suite which consists of the C++ and Fortran Compilers, domain specific performance libraries and multithreading libraries. The focus of the first module is describing advanced optimization options, standards and GCC compatibility, code dispatching for heterogeneous systems, and a summary of the latest features for both Linux and Windows hosted compilers. A brief overview of OpenMP 4.0 (and an outlook to 4.1) complements the session. The focus of OpenMP 4.0 is offloading to Intel® Xeon Phi™ coprocessors. It also includes Intel extensions with respect to the upcoming OpenMP* 4.1.

Break

Numerical Stability with Intel® C++ & Fortran Compiler

This topic raises awareness about the effects of numerical stability for floating point computations and describes ways to influence them with the Intel® C++ & Fortran Compiler. It also includes guidelines how to achieve numerical reproducibility for multiple (heterogeneous) systems for the different programming models supported by Intel® Parallel Studio XE 2015 Cluster Edition.

Intel® Math Kernel Library

For the second module and as part of the performance libraries, Intel® Math Kernel Library is explained. Basic functionality and recently added features are addressed.

Q&A