Memory Access Profiling: Find and Fix Common Performance Bottlenecks

Does your code deliver desired results yet have performance issues? Is it spending too much time in a hotspot or causing memory bottlenecks? Are your threads jumping around a lot?

On today’s modern processors, high-performing code must be threaded and scalable, vectorized, and tuned. Which is precisely what Intel® VTune™ Amplifier was designed for—to help modernize and optimize your code, whether you’re tuning a simple application or applying advanced optimizations on a threaded MPI application. Intel VTune Amplifier shows you where code performance is lacking and what to do about it, accurately and with very low overhead. Its advanced profiling, data mining, and interpretation capabilities deliver actionable insights—all in a friendly analysis interface.

Settle in for 50 minutes of demonstrations and how-to’s, including:

  • Discovering common performance problems, such as inefficient array access patterns and poor NUMA utilization
  • Answering questions, such as “What are my memory bottlenecks?” and “Which data structures and lines of source code are causing them?”

About the Speaker

Jackson Marusarz, Technical Consulting Engineer, Intel Corporation

Jackson Marusarz is a technical consulting engineer (TCE) in Intel’s Developer Products Division. As the lead TCE for Intel® VTune™ Amplifier, Jackson’s main focus is on software performance analysis and tuning for both serial and multi-threaded applications. His time is split between figuring out how to analyze and tune software, and how to create tools that help others do the same.

Performance varies by use, configuration, and other factors. Learn more at www.Intel.com/PerformanceIndex.