VTune is a performance profiler from Intel and is available on the HPC cluster.

General information of Intel : https://software.intel.com/en-us/get-started-with-vtune-linux-os

The VTUne commands have prefix amplxe- *. The most important are:

• amplxe-cl #< launches VTune Command Line.
• amplxe-gui #< launches VTune with a graphical interface.

Profiling steps:

• Data collection (collect)
• Analysis of the results

To collect profiling data, run the executable via amplxe-cl, specifying an action between the following: https://software.intel.com/en-us/vtune-amplifier-help-amplxe-cl-actions

The action necessary to collect data about performance is collect. This action requires an option to specify the type of analysis, including:

• hotspots: Identify your most time-consuming source code.
• concurrency: Analyze how your application is using available logical CPU cores, discover where parallelism is incurring synchronization overhead, and identify potential candidates for parallelization.
• hpc-performance: Identify opportunities to optimize CPU, memory, and FPU utilization for compute-intensive or throughput applications.
• memory-access: Measure a set of metrics to identify memory access related issues (for example, specific for NUMA architectures).

Command example:

amplxe-cl -collect hotspots -result-dir <dirname>  home/test/myApplication

The off-line analysis of the collected data can be done in a graphical way with amplxe-gui:

 amplxe-gui <dirname>

Connect to gui.hpc.unipr.it and perform an interactive job; eg:

 srun -N1 -n68 -p knl --account=<account name>
 
 > module load intel # (include vtune too)
 > amplxe-cl -collect hotspots -result-dir <dirname>  home/test/myApplication
 > exit

It is preferable to launch on the University HPC cluster amplxe-gui from gui.hpc.unipr.it:

  
module load intel 
amplxe-gui  <dirname>

On Marconi you can 'submit an interactive session, in which to perform both the collect and the data analysis (amplxe-gui does not work on the login machine), such as:

srun -N1 -n68 -t 00:30:00  -A <account name> -p knl --pty bash

> module load intel vtune
> amplxe-cl  -collect hpc-performance -result-dir myresults bin/myexe <options>
> amplxe-gui  myresults
> exit

To include debugging information in the results (for example the source code) you need to fill in with the option

   -gdwarf-2