Intel VTune Amplifier

Version	System A	System B	System C
VTune Amplifier XE 2022 (default)	―	+	+

Key: + = Available for use, ― = Not available for use.

Version	Modulefile Name
2022	iutils/xe2022

In the Intel compiler environment, execute the module command as shown below.

$ module load iutils

For details on the module command, see Modules.

Commands	Purpose
vtune-gui	Activates VTune Amplifier GUI.
vtune	Activates VTune Amplifier command-line version.

Commands	Purpose
amplxe-gui	Activates VTune Amplifier GUI.
ampxle-cl	Activates VTune Amplifier command-line version.

Options	Purpose
-collect=string	Specifies the analysis type.
-app-working-dir=string	Specifies the working directory.
-r, -result-dir=string	Specifies the directory in which to save the result.

Types	Purpose
concurrency	Displays a Concurrency of multithread.
hotspots	Makes a Hotspots analysis.
lightweight-hotspots	Makes a Hotspots analysis of high speed.
locksandwaits	Makes a Locks and Waits analysis.
snb_memory-access	Makes a memory access analysis with Sandy Bridge.

When using Intel VTune Amplifier, compile with the debugging option -g -O2. For details on compile options, please also refer to the webpage here.

$ icc -g -O2 test.c

In the Exceed onDemand , Intel VTune Amplifier is activated when executing the amplxe-gui command.For details of xrun , see Interactive Processing.

$ xrun amplxe-gui

Select File > New > Project, the window below appears. Enter the appropriate Project name, then click Create Project.

The Project Properties window appears, specify the target program in the Application text box, then click OK. If an argument is required, specify it here.

The window below is back after the setup of project properties has completed, click the New Analysis icon(triangle symbol) in the toolbar.

The Choose Analysis Type window appears, specify the target of an analysis, then click Start.

Some time later, the analysis results is displayed. This example below indicates that the grid_intersect function takes a lot of CPU time.

The GUI version of VTune can handle OpenMP program by specifying the number of threads with -A option when running xrun.

Example: analyze with 8 threads

xrun -A t=8:c=8 amplxe-gui

After launching VTune, select "Advanced Hotspots", select "Analyze OpenMP Regions", and push "Start", then 8-thread analisys starts.

MPI program can be analyzed with CUI version only.

When using Intel Inspector, compile with the debugging option -g -O2. For details on compile options, please also refer to the webpage here.

$ icc -g -O2 test.c

To use Intel VTune Amplifier CUI, use the amplxe-cl command. In this example, a directory for outputting the results are specified before starting to make a Hotspots analysis. For details of tssrun , see Interactive Processing.

$ tssrun amplxe-cl -collect hotspots -r=./result ./a.out
Parser returned a non-zero error code reading dat/balls.dat
Aborting Render...
Using result path `/home/w/w55060/intel/amplxe/samples/tachyon/result'
Executing actions 75 % Generating a report                                     
Summary
-------
　
Elapsed Time:  0.101
CPU Time:      0
Executing actions 100 % done

To view the analysis results with Intel VTune Amplifier GUI, execute the amplxe-gui command in the X-Windows(GUI) environment(e.g., Exceed onDemand). For details of xrun , see Interactive Processing.

$ xrun amplxe-gui ./result/result.amplxe

The CUI version of VTune can handle MPI and OpenMP parallel program. -A option is needed to specify the number of processes or threads, and mpiexec.hydra is also needed.

Example: 4 MPI parallel

$ tssrun -A p=4 mpiexec.hydra amplxe-cl -collect hotspots  -r=./result ./a.out
qsub: waiting for job 1258261.jb to start
qsub: job 1258261.jb ready

amplxe: Analyzing data in the node-wide mode. The hostname (nb-0041) will be added to the result path/name.
amplxe: Collection started.
Hello world from processor nb-0041, rank 2 out of 4 processors
Hello world from processor nb-0041, rank 3 out of 4 processors
Hello world from processor nb-0041, rank 0 out of 4 processors
Hello world from processor nb-0041, rank 1 out of 4 processors
amplxe: Collection stopped.
amplxe: Using result path `/home0/w/w00001/mpitest/b/result.nb-0041'
amplxe: Executing actions 19 % Resolving information for `libtmip_psm2.so'     
amplxe: Warning: Cannot locate debugging symbols for file `/opt/app/intel/impi/2017.2.174/intel64/lib/libtmip_psm2.so'.
amplxe: Executing actions 75 % Generating a report                             
amplxe: Executing actions 100 % done                                           

Collection start time     07:23:49 19/12/2017 UTC
Collection stop time      07:23:50 19/12/2017 UTC

Summary
-------
Elapsed Time:       0.553
Paused Time:        0.0  
CPU Time:           0.310
Average CPU Usage:  0.435

qsub: job 1258261.jb completed