In the video Jeff looks at identifying the critical path through an algorithm and narrow down where we can tune to improve performance. He also shows how Flow Graph Analyzer works, to both model some applications that you might want to work with or build, using things like Intel's Threading Building Blocks (TBB), or analyzing an existing application and test and see where your performance bottlenecks are and identify and be able to tune those.
Jeff learns about a computer vision experiment that Intel Advisor’s Flow Graph Analyzer is used to analyze and improve the performance on.
In this video Jeff models and builds out how we want a multi-threaded application to work. And looks at how the modeling works inside Flow Graph Analyzer.
We go throguh a model in Flow Graph Analyzer, going through things like: starting a source node (which can take in any kind of data source and apply a lambda function to it to output some data we could continue processing further down the line), setting up branches three different branches of functions that are going to process the output of that source node, specifying the execution weight for each node, and how to change the scalability to change from one element to ten elements.
Jeff also takes a look at how to do performance analysis on an existing application.
The Computer Vision Article referenced from Intel Parallel Universe is at: http://bit.ly/ARVision
Download a copy of Intel Advisor at: https://software.intel.com/en-us/advisor