This video illustrates how Intel Advisor can take python applications, run them through a high grade performance profiler, and get roofline reports, hotpsot analysis, and get recommendations on how to improve your application. The video has a sample using the TensorFlow tutorial, on how to train a model, execute that model, and then analyze its performance.
Intel Advisor can give you insight into the performance of your compiled applications and your Python applications too. It can inspect and do performance profiling for your Python applications, and it's not just native applications.
Through the sample, we test some image identification and recognition using TensorFlow, we look at the source code, look at a complete analysis of our sample and see which loops were the most time consuming.
We also take a look at the roofline report, can see the more compute-bound loops, and can see how it executed, where in the call stack it ran. And look through the recommendations, like using smaller data types (in our case).
Jeff Fritz is a senior program manager in Microsoft’s Developer Division working on the .NET Community Team. As a long time web developer and application architect with experience in large and small applications across a variety of verticals, he knows how to build for performance and practicality. Four days a week, you can catch Jeff hosting a live video stream called 'Fritz and Friends' at
twitch.tv/csharpfritz. You can also learn from Jeff on WintellectNow and Pluralsight, follow him on twitter @csharpfritz, and read his blog at
jeffreyfritz.com