Omni Profile Visualization Tool

NAME

tlogview -- Omni tlog visualization tool

SYNOPSIS

       tlogview tlog_file

DESCRIPTION

tlogview is a tool to visualize a log file in "tlog" file format, which is produced by profiling OpenMP programs compiled by the Omni OpenMP Compiler.

The initial view is determined by the longest time stamp. In scale area, the time scale is displayed in second unit. In log display area, each bar indicates the status for each processor. By pointing device, you can zoom up and down to see the log file more precisely.

• Dragging a region in log display area with left button, zoom up the selected region.
• "Zoom up" button zooms up the current log display area.
• "Zoom down" button zooms down back to the previous display area.
• "ShowColorMap" button shows the legend and colors for each event.
• "Exit" terminates the tool.

The color indicate the associated events as follows:

• White -- parallel region (duration)
• Red -- Barrier (duration)
• Yellow -- Critical region (duration)
• Green -- Initialize event of parallel loop. (event)
• Cyan -- Schedule event of the next chunk for parallel loop. (event)
• Magenta -- the beginning of "section" directive. (event)
• Blue -- the beginning of "single" directive. (event)

The duration is displayed by filled rectangle, and the event is displayed by line. Note that nothing appears for a sequential execution out side of parallel region.

Note that when profiling is on, the execution is getting slow down due to the overhead for getting the timing for each event. For more information, see Profile .

ENVIRONMENT VARIABLES

To turn on profiling, set the environment variable OMPC_LOG when executing OpenMP programs. See Omni Environment Variables .

FILES

• bin/tlogivew -- shell command to start tlogview.
• lib/openmp/java/tlogview.jar -- Java jar file of tlogview

The tlogview is implemented using AWT java classes.

Unsupported Features and Known BUGS

The current version does not have vertical scroll bar.

In profiling mode, blocks of tlog records are output in the order of time stamp. A large tlog file is automatically truncated (2000K records).