STAT - Introduction man page

intro_stat, intro_STAT — introduces the Stack Trace Analysis Tool (STAT) package

IMPLEMENTATION

Cray Linux Environment (CLE)

DESCRIPTION

The Stack Trace Analysis Tool (STAT) is a scalable, lightweight debugger for parallel applications. STAT works by gathering stack traces from all of a parallel application’s processes and merging them into a compact form. The resulting output indicates the location in the code that each application process is executing, which can help locate a bug. The Stack Trace Analysis Tool (STAT) package includes three commands to invoke and control STAT as well as analyze its output..

Note: The capitalized STAT commands have been deprecated in favor of all lowercase commands and will be removed in a future release. The STAT command is now stat-cl, STATGUI is now stat-gui, and STATview is now stat-view.

stat-cl

Invokes STAT to periodically gather stack traces from a parallel application’s processes and assemble a profile of the application’s behavior over its process space and time. Results are written to a .dot file that can be read using either stat-view or stat-gui. For more information, see the stat-cl(1) man page.

stat-view

Invokes a graphical user interface (GUI) for viewing the .dot output files generated by STAT or ATP (Abnormal Termination Processing). stat-view provides navigation of the call prefix tree and also allows manipulation of the call tree to help focus on areas of interest. For more information, see the stat-view(1) and intro_atp(1) man pages.

stat-gui

Invokes a graphical user interface (GUI) to run STAT and view the generated call prefix trees. This GUI provides a variety of operations to help focus on particular call paths and tasks of interest. It also includes an interface to attach a heavyweight debugger, such as Totalview, to a subset of the application’s processes.

In addition to the operations provided by stat-view, stat-gui provides a toolbar to control STAT’s operation. For more information, see the stat-gui(1) man page.

To use the STAT package or view any of its man pages, load the STAT module.

% module load stat

ENVIRONMENT VARIABLES

The STAT package leverages MRNet for scalable communication.

STAT_MRNET_STARTUP_TIMEOUT

Defines the number of seconds for compute node daemons to wait for connection into the MRNet communication tree. The default value, 31 seconds, has been set to address medium size runs. Large runs may need this value increased.

A side effect of this timer value is that it defines the minimum time between stat-cl/stat-gui invocations. Therefore, setting a large value will mean that successive runs will need to be separated by that large amount of time.

STAT - Introduction man page

IMPLEMENTATION

DESCRIPTION

ENVIRONMENT VARIABLES

SEE ALSO