Some time ago I wrote about profiling methods. Now I’m going to continue with application performance profiling in Visual Studio and describe main steps in using Performance Wizard.

First of all you should choose some project (application) to analyze.  It’s better to have source code of the application, but you can use executable files as well.

Step 1.  Configure and run a performance session

So let’s start. Make sure that you run your Visual Studio as administrator. After opening project in VS choose Analyze -> Launch Performance Wizard menu item. In Performance Wizard dialog specify profiling method to be used (read my previous post) and choose the application to profile.

After closing the wizard your application will be started. Then do whatever you want with your application (call functions, press buttons or menus, etc) to load it or just wait for some time.

Step 2.  Analyze application performance report

And after you close your application or after you press Stop Profiling link in Visual Studio, profiling report will be generated and the Summary view of the profiling report will appear in the main window in Visual Studio.  On the picture bellow you can see example of Summary profiling report (CPU Sampling was used as profiling method).

Let’s have a closer look at this report. At the top of Summary view you can see a graph describing changes of  CPU usage relative to time. This graph helps to determine at what moments of time your application intensively loads CPU. You can also zoom this graph or select some time interval and filter by it.

The next very interesting part of Summary performance report is called Hot Path. It helps to determine the most expensive call path based on sample counts for chosen interval of time (the whole interval of profiling by default). If you click a function name you can see details of cost distribution for the function (the Function Details view of the profiling data will be shown).

The Function Details view presents a graphical view of the profiling data for the selected function, showing all the functions that called that function and all the functions that were called by the selected function. This view also gives you idea of what lines of code need to be refactored to decrease CPU usage.

Except described above views of performance report you can also use other views for deeper analysis:

·         Call tree view

·         Modules view

·         Caller / Callee view

·         Functions view

·         Lines view

·         Marks view

·         Processes view

·         IPs view.

And of cause you can save report or export report data, or compare reports.

Step 3. Revise code of your application and rerun performance session

Analyzing different views of performance report will help you to find places in your code to optimize. And after making changes in one or more functions, you can repeat the profiling run and compare the data to see the difference that your changes have made to the performance of your application.