Hey Ampere Developers,
How do Performance Engineers characterize and fix performance issues?
Have you ever wondered how application performance engineers go about finding that needle in a haystack when a customer comes to them with a performance issue? How do you go from “My application is running slower than I expect” to understanding exactly what the issue is, and fixing or working around it?
In this video, the first of a series on the Ampere performance toolkit, Bhakti Hinduja breaks down exactly how top-down performance analysis works, starting with reproducing a performance issue with a synthetic benchmark, understanding which subsystem is constraining the performance in the “bad” case, and finally drilling down to identify the root cause.
https://www.youtube.com/watch?v=ecJDNhcPfbY
In part two of the series, Bhakti explores the individual tools in the Ampere Performance Toolkit, and digs a little deeper into the types of problems solved by each of them:
https://www.youtube.com/watch?v=4PXcbRtD6Ys
The Ampere Performance Toolkit makes it easy to perform each of these steps, accelerating your ability to understand your own system performance, including:
- Capturing system configuration and benchmark results with the Ampere Perfkit Benchmarker: GitHub - AmpereComputing/ampere-perfkit-benchmarker · GitHub
- Analyzing system performance to identify bottlenecks with the Ampere System Profiler: GitHub - AmpereComputing/ampere-system-profiler · GitHub
- Drilling down to CPU measurements to understand how CPU cores are spending their time with the Ampere PMU Profiler: GitHub - AmpereComputing/ampere-pmu-profiler: Performance characterization and analysis tool for Ampere CPUs · GitHub
Stay tuned as we publish much more content and use-cases of people seeing benefit from using these tools!
Are you already on the Ampere Developer Community!
Do you have questions about the Ampere Performance Toolkit, or feedback about your own performance debugging journeys that you would like to share? Our Ampere Developer Community Forum is the place to be!
Join the conversation: https://community.amperecomputing.com
Happy coding and see you in the next edition!
The Ampere Developer Relations Team