Walkthrough: Using Profiler APIs

  • Article

The walkthrough uses a C# application to demonstrate how to use the Visual Studio Team System Developer Edition Profiling Tools APIs. You will use the profiler APIs to limit the amount of data that is collected during instrumentation profiling.

The steps in this walkthrough generally apply to a C/C++ application. For each language, you will have to configure your build environment appropriately.

Typically, you will start to analyze application performance by using sample profiling. If sample profiling does not provide information that pinpoints a bottleneck, instrumentation profiling can provide a greater level of detail. Instrumentation profiling is very useful for investigating thread interaction. 

However, a greater level of detail means that more data is collected. You might find that instrumentation profiling creates large data files. Also, instrumentation is more likely to impact the performance of the application. For more information, see Understanding Instrumentation Data Values in Profiling Tools and Understanding Sampling Data Values in Profiling Tools

The Visual Studio profiler allows you to limit the collection of data. This walkthrough offers an example of how to limit the collection of data by using the profiler APIs. The Visual Studio profiler provides an API for controlling data collection from within an application.

For native code, the Visual Studio profiler APIs are in VSPerf.dll. The header file, VSPerf.h, and the import library, VSPerf.lib, are located in the Microsoft Visual Studio 9\Team Tools\Performance Tools directory.

For managed code, the profiler APIs are in the Microsoft.VisualStudio.Profiler.dll. This DLL is found in the Microsoft Visual Studio 9\Team Tools\Performance Tools directory. For more information, see Microsoft.VisualStudio.Profiler.

Prerequisites

This walkthrough assumes your choice of development environment is configured to support debugging and sampling. The following topics provide an overview of these prerequisites:

How to: Choose Collection Methods

How to: Reference Windows Symbol Information

By default, when the profiler is started, the profiler collects data at the global level. The following code at the start of the program turns global profiling off.

        DataCollection.StopProfile(
        ProfileLevel.Global,
        DataCollection.CurrentId);

You can turn off data collection at the command line without the using an API call. The following steps assume your command line build environment is configured to run the profiling tools and as your development tools. This includes the settings necessary for VSInstr and VSPerfCmd. See Command-Line Profiling Tools.

Limiting Data Collection Using Profiler APIs

To create the code to profile

  1. Create a new C# project in Visual Studio, or use a command line build, depending on your preference.

    Note

    Your build must reference the Microsoft.VisualStudio.Profiler.dll library, located in the Microsoft Visual Studio 9\Team Tools\Performance Tools directory.

  2. Copy and paste the following code into your project:

    using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.VisualStudio.Profiler;

namespace ConsoleApplication2
{
    class Program
    {
        public class A
        {
         private int _x;

         public A(int x)
         {
          _x = x;
         }

         public int DoNotProfileThis()
         {
          return _x * _x;
         }

         public int OnlyProfileThis()
         {
          return _x + _x;
         }

         public static void Main()
         {
        DataCollection.StopProfile(
        ProfileLevel.Global,
        DataCollection.CurrentId);
          A a;
          a = new A(2);
          int x;    
          Console.WriteLine("2 square is {0}", a.DoNotProfileThis());
          DataCollection.StartProfile(
              ProfileLevel.Global,
              DataCollection.CurrentId);
          x = a.OnlyProfileThis();
          DataCollection.StopProfile(
              ProfileLevel.Global, 
              DataCollection.CurrentId);
          Console.WriteLine("2 doubled is {0}", x);
         }
        }

    }
}

To collect and view data in the Visual Studio IDE

  1. Open the Visual Studio IDE. On to the Analyze menu, point to Profiler, and then select New Performance Session.

  2. Add your compiled binary to the Targets list in the Performance Explorer window. Right-click Targets, and then select Add Target Binary. Locate the binary in the Add Target Binary dialog box, and then click Open.

  3. Select Instrumentation from the Method list on the Performance Explorer toolbar.

  4. Click Launch with Profiling.

    The profiler will instrument and execute the binary and create a performance report file. The performance report file will appear in the Reports node of the Performance Explorer.

  5. Open the resulting performance report file.

By default, when the profiler is started, the profiler will collect data at the global level. The following code at the start of the program turns global profiling off.

        DataCollection.StopProfile(
        ProfileLevel.Global,
        DataCollection.CurrentId);

To collect and view data at the command line

  1. Compile a debug version of the sample code you created in the "Creating Code to Profile" procedure, earlier in this walkthrough.

  2. To profile a managed application, type the following command to set the appropriate environment variables:

    VsPefCLREnv /traceon

  3. Type the following command:VSInstr <filename>.exe

  4. Type the following command:VSPerfCmd /start:trace /output:<filename>.vsp

  5. Type the following command:VSPerfCmd /globaloff

  6. Execute your program.

  7. Type the following command:VSPerfCmd /shutdown

  8. Type the following command:VSPerfReport /calltrace:<filename>.vsp

    A .csv file is created in the current directory with the resulting performance data.

See Also

Reference

Microsoft.VisualStudio.Profiler

Other Resources

Visual Studio Profiler API Reference (Native)

Getting Started with Profiling Tools

Using the Profiling Tools From the Command-Line