Application Performance Profiling and Debugging
Tue March 8 2016, 8:00am-5:00pm
University of Oklahoma, Norman campus
Four Partners Place (4PP), 3rd Floor Seminar Room, Suite 3065
301 David L. Boren Blvd. Norman, OK 73019
Contact: Henry Neeman (firstname.lastname@example.org)
We’re now able to offer remote participation via videoconferencing.
Please note that we can only make 98 tickets available for remote participation, because of limitations to the videoconferencing system.
For application developers
- Industry standard best practices to systematically discover performance bottlenecks.
- Measuring and improving application performance.
- Scientific debugging – applying best practices to fix difficult bugs at all scales.
This hands-on course uses HPC debugger and profiler tool suites:
- Sample code will be provided – but having access to participant’s code will be better!
- For MPI, CUDA, OpenMP and Intel Xeon Phi applications.
Performance Optimization : Use HPC code optimization tools to get your application running
faster and scaling higher
- Profile your code and see the bottlenecks and the performance opportunity.
- Explore performance problems including:
- MPI and work imbalance.
- The memory wall and vectorization
- OpenMP synchronization costs.
- Scientific Debugging
Learn the art and the science of high performance debugging:
- Using HPC code development tools on real applications.
- Using a debugger to save time.
- Learn techniques of debugging to shorten the time even further.