HOW Series “Tools”

The free Hands-On Webinar (HOW) series explores 3 essential software development tools for Intel architecture: Intel VTune Amplifier, Intel Math Kernel Library (MKL) and Intel Advisor. Each 2-hour webinar comprises a theoretical part and a hands-on component, which is live demonstration of programming and optimization techniques on modern Intel Xeon processors and Intel Xeon Phi coprocessors.

The Colfax HOW Tools training series is an integral part of the Intel Modern Code Developer program which supports developers in leveraging application performance in code through a systematic optimization methodology.

For similar upcoming trainings, see our Training Calendar.

Webinar 1 – Hotspot-guided optimization with Intel VTune Amplifier XE

This session features select topics of performance diagnostics using Intel VTune Amplifier XE.

• Discussion of general performance optimization challenges.
• Demonstration of the usage of VTune to identify top hotspots on traditional CPUs and Intel Xeon Phi coprocessors.
• Illustration of the optimization report functionality of Intel compilers. Hints on using the reports to find scalar, vector and memory optimization opportunities.

The demonstrated exercises are based on a case study referred to below, the astrophysical library HEATCODE.

What you will learn: how to detect performance issues with a range of techniques, from simple scalability tests and compiler output to the usage of specialized tools.

Slides:  Colfax-HOW-Tools-VTune.pdf (18 MB)
HEATCODE slides (presentation at LBNL, 2014):  HEATCODE-LBNL.pdf (9 MB)
HEATCODE paper: http://arxiv.org/abs/1311.4627

 

Webinar 2 – Practical usage of Intel Math Kernel Library: performance tuning tips and usage with coprocessors

This session features select topics of achieving high performance using the Intel Math Kernel Library (MKL).

• Overview of available functionality in MKL.
• Tuning tips: threads count and affinity, choice of problem size and type.
• Automatic offload: supported function, what it takes to achieve scaling.
• Using automatic offload to accelerate certain R, Matlab calculations.

What you will learn: how to get good performance in standard tasks without “reinventing the wheel”.

Slides:  Colfax-HOW-Tools-MKL.pdf (5 MB)
Code: Colfax-HOT-Series-MKL-webinar-code.tgz (10 KB)

 

Webinar 3 – Guided Code Vectorization with Intel Advisor XE

This session features the usage of an optimization tool called Intel Advisor. The discussion is illustrated with an example workload that computes the electric potential in a set of points in 3-D space produced by a group of charged particles. The example workload runs on a multi-core Intel Xeon processor with Intel AVX2 instructions. The hands-on walk-through demonstrates how to diagnose and fix common issues in vector calculations:

• Vector dependence
• Unexpected type conversion
• Inefficient memory access patterns

What you will learn: how to achieve significant performance boosts using guidance from a dedicated software tool.

Slides:  Colfax_HOW_Tools_Advisor_r1.0.pdf (5 MB)
Code: Colfax_Advisor_Vectorization_Code.zip (50 KB)

Getting Your Questions Answered

Feel free to post any questions about these webinars in Colfax Research forums.