![]() |
Cube GUI User Guide
(CubeGUI 4.9, revision 6e5e012c)
Introduction in Cube GUI and its usage
|
To start using the measurement plugin, first open CubeGUI and navigate to the "Measurement Plugin" option under the "Open context-free plugin" menu.
If you are unable to locate the plugin in the menu, please ensure that you have followed all the steps listed in the section Cube GUI Plugins.
If you're still having difficulty locating the plugin, one solution is to configure the plugin search path. You can do this by accessing the "Plugins" menu, selecting "Configure Plugin Search Paths," and adding the path to the project's .so file. Once you have done this, restart CubeGUI to activate the plugin.
Alternatively, you can start CubeGUI in verbose mode by using the command line option -verbose, which will list all the plugins that CubeGUI is able to locate. For additional information, please refer to the CubeGUI User Guide.
In this section, we will guide you through an example of how to use the Measurement plugin. Once the Measurement plugin is launched, the user is prompted to either start a new measurement or load a recent one. For this example, we choose to start a new measurement.
After selecting "Start New Measurement", the user will be taken to the Setup tab. Here, the user can specify the MPI and Compiler used to build the program they want to measure. This information is required to select an appropriate Score-P version, since the Score-P version must be installed with the same MPI and compiler.
In our example, the plugin has identified a matching Score-P version in our path. We select this version and proceed to the Instrumentation step.
In the Instrumentation tab, we select the "jacobi" executable using the Browse Executable File Button. Since the executable has not been instrumented yet, we choose to prepare a new instrumentation and specify our preferred build system, which in this case is the makefile.
The plugin provides us with two options: to browse for the makefile or to open a makefile that has been detected by the plugin. Since the plugin has detected our makefile, we choose to open it. This displays a tab where we can edit the makefile.
The plugin opens a window to edit the makefile. At the top of this window, we find instructions on how to apply the correct compiler wrapper to enable instrumentation. In this example, we changed the compiler wrapper from CC=mpicc
to CC=scorep-mpicc
. After making the changes, we save them to return to the Instrumentation tab window.
In the next step, a dialog box appears for specifying the build command of the program. In our case, the plugin automatically suggests the correct command. However, it is possible to customize this command if necessary. After entering the build command, we click the "Build your application" button to execute the command. The build output is displayed in the Virtual Console. If the build is successful, a message is displayed, and the Continue with Analysis Button is enabled.
We click the Continue with Analysis Button to proceed with the Measurement tab.
At this point, we select to perform an initial run, which is the first run that can be performed. We choose to run four processes and two threads. The plugin suggests a directory name for the experiment directory, where the resulting profile.cubex file will be located. Additionally, the plugin suggests the run command. The commands executed in the background and the required variable settings can be viewed in the virtual console.
The plugin offers two options to perform the measurement: a job can be submitted, or the measurement can be executed locally. In this example, we perform the measurement locally.
The measurement was performed successfully, and a corresponding message is displayed. The user now has the option to either take another measurement or to view the results of the measurement. For more information, please refer to Measurement.
As next, see Overview
![]() |
Copyright © 1998 Forschungszentrum Jülich GmbH,
Jülich Supercomputing Centre
Copyright © 2009–2015 German Research School for Simulation Sciences GmbH, Laboratory for Parallel Programming |