Intel® VTune™ Amplifier offers several tools you can use as you learn, including:
About Tutorials
VTune Amplifier XE provides several tutorials that use sample code to explain how to:
Identify where your application is spending time, detect the most time-consuming program units and how they were called.
Identify locks and waits preventing parallelization.
Identify hardware-related issues in your application such as data sharing, cache misses, branch misprediction, and others.
Analyze I/O bound applications and minimize I/O Wait time.
Identify false sharing issues in your code and understand the cause.
Identify issues in a hybrid OpenMP* and MPI application using Intel MPI Performance Snapshot, Intel Trace Analyzer and Collector, and Intel VTune Amplifier.
VTune Amplifier for Systems provides tutorials that provide hands-on instructions how to:
Configure and run the Advanced Hotspots analysis on a remote embedded Linux system and view the collected data on the Linux* host.
Configure and run the Basic Hotspots analysis on a remote Android* system and view the collected data on the Linux or Windows* host.
Configure and run the Energy Analysis on a target Android system using the command line Intel SoC Watch collector and import and view the collected data with VTune Amplifier GUI on the Linux or Windows host.
You can access the tutorials from:
Tutorials Index at <install_dir>/documentation/<locale>/tutorials/index.htm in the Intel VTune Amplifier installation folder.
Note
The default installation path for the VTune Amplifier XE is /opt/intel/vtune_amplifier_xe_version. For the VTune Amplifier for Systems, the default <install_dir> is:
- For super-users: /opt/intel/system_studio_version/vtune_amplifier_for_systems
- For ordinary users: $HOME/intel/system_studio_version/vtune_amplifier_for_systems
Getting Started page > Training and Documentation section, available from:
Help menu - Choose Help > Getting Started with Intel VTune Amplifier 2017 for Systems
Help toolbar button
Intel Developer Zone at https://software.intel.com/en-us/articles/intel-vtune-amplifier-tutorials
About Sample Applications
VTune Amplifier ships with the following code samples:
Sample Application |
Summary |
|---|---|
tachyon_find_hotspots at <install_dir>/samples/tachyon_vtune_amp_xe.tar.gz |
Displays a rendering of a graphical image via 2D ray tracing. Demonstrates: Identifying hotspot functions and code lines in a C++ application. Performance issues: wrong algorithm choice, ineffective parallelization. Companion tutorials: Finding Hotspots (VTune Amplifier XE, VTune Amplifier for Systems), Enabling Performance Collection on an Embedded Linux System (VTune Amplifier for Systems) |
tachyon_analyze_locks at <install_dir>/samples/tachyon_vtune_amp_xe.tar.gz |
Displays a rendering of a graphical image via 2D ray tracing. Demonstrates: Identifying locks preventing efficient parallelism in a C++ application. Performance issues: locks and waits. Companion tutorial: Analyzing Locks and Waits(VTune Amplifier XE, VTune Amplifier for Systems) |
tachyon_vtune_amp at <install_dir>/samples/tachyon_vtune_amp.tar.gz |
Displays a rendering of a graphical image via 2D ray tracing. Demonstrates: Identifying hotspot functions and code lines in a C++ application for an Android target. Performance issues: ineffective parallelization. Companion tutorial: Finding Hotspots on an Android Platform(VTune Amplifier for Systems) |
matrix at <install_dir>/samples/matrix_vtune_amp_xe.tar.gz |
Calculates matrix transformations. Demonstrates: Identifying general hardware issues in a C++ application on the host system and on the Intel® Xeon Phi™ coprocessor (code name: Knights Corner). Performance issues: poor CPI rate, cache misses, retire stalls, execution stalls, and others. Companion tutorials: Identifying Hardware Issues and Finding Hotspots on the Intel Xeon Phi Coprocessor (VTune Amplifier XE) |
nqueens_parallel at <install_dir>/samples/nqueens_parallel.tar.gz |
Solves the nqueens problem for various board sizes. Demonstrates: Identifying general hardware issues in a Fortran application. Performance issues: thread contention, ineffective parallelization. Companion tutorial: Finding Hotspots (VTune Amplifier XE) |
diskio at <install_dir>/samples/diskio_vtune_amp_xe.tar.gz |
Works with the I/O device using the system file cache and user buffer. Demonstrates: Identifying ineffective usage of the I/O resources in a C++ application. Performance issues: CPU usage overhead, high I/O Wait Time, big I/O Queue Depth. Companion tutorial: Analyzing Disk Input/Output Waits |
linear_regression at <install_dir>/samples/linear_regression.tar.gz |
Generates the summary statistics of points to give the linear approximation. Demonstrates: Identifying false sharing and its cause in a C application. Performance issues: false sharing. Companion tutorial: Identifying False Sharing |
jitprofiling <install_dir>/samples/jitprofiling.tar.gz |
Demonstrates: Using the JIT profiling API. |
serial_nqueens_csharp parallel_nqueens_csharp at <install_dir>/samples/jitprofiling.tar.gz |
Computes the number of solutions to the nQueens problem for a given board size. Demonstrates: Identifying hotspots in a Microsoft Visual C#* application. |
Simulates electrophysiological heart activity. Demonstrates: Identifying performance issues in hybrid OpenMP* and MPI code. Performance issues: MPI-bound, communication issues, vector instruction sets, serial and parallel code inefficiencies. Companion tutorial: Analyzing an OpenMP and MPI Application (VTune Amplifier XE, Intel Trace Analyzer and Collector, MPI Performance Snapshot) |
To install and set up the VTune Amplifier sample code:
Copy the archive file from the installation directory to a writable directory or share on your system.
Extract the sample from the archive.
Note
VTune Amplifier sample code and corresponding tutorials are also available at https://software.intel.com/en-us/product-code-samples.
Samples are non-deterministic. Your screens may vary from the screen shots shown throughout these tutorials.
Samples are designed only to illustrate the VTune Amplifier features and do not represent best practices for tuning any particular code. Results may vary depending on the nature of the analysis and the code to which it is applied.