Getting Started with GistView


This tutorial will help you get an overview of the capabilities of GistView, the data viewer portion of Gist by walking you through the procedures you might use in a real session to view and analyze your data. The tutorial is designed to be read in order, but feel free to jump around to the sections you're most interested in.

The following topics will be covered in this tutorial. For a more in-depth treatment on specific areas, please consult the section in the user manual which deals with the topic you are interested in.

Running GistView

GistView requires a system running a Java Virtual Machine supporting Java 2.0 or higher, and a graphics terminal in order to display the interface.

In order to run GistView, you may type "java GistView" on systems which support a command line interface. There are no command line options to specify.

Once GistView starts, you should see something similar to this:

The exact display will vary depending on the system you use.

Loading a log file

Before you begin analyzing your data, you must load the log file generated by the Gist instrumentation library. A sample log file has been provided with the Gist distribution for the purpose of this tutorial. The sample data file is located under the Test directory in the Gist distribution and is called main3.glg. Select "Open File" under the File menu, and choose Test/main3.glg in the file browser.

The demo file should appear like the image below once the file has opened. You can click on the image to see the full size version.

This is GistView's "Home Screen", the place where you will typically begin, moving to other sections through this interface. The large window towards the left side of the screen is called the Event Window. This is where the timeline display of the data collected will be shown. This screen shows each of the recorded event sequences along its own timeline. As you can see in the example above, there are four event sequences in this case, numbered 1, 2, 3, and 4.

As you can see above, the window shows all the events in the log file, in this case is somewhat crowded. You can zoom in to view a smaller portion of the file, or out to get an overview of what happened in the long term during your program's execution. Choose "Zoom In" or "Zoom Out" from the Display menu to do this. Choosing zoom out repeatedly will double the size of the time window which is displayed until the entire file is contained in the window. With a real data file this might be too much data to be useful. Zooming in or out changes the size of the time window displayed by a factor of two by default, although this can be changed if desired. If the time window would extend beyond the limits of the file when zooming out, the time window is scaled such that the window does not extend beyond the limits of the file.

Scrolling through the data

The Event Window is scrollable, both horizontally and vertically in order to allow viewing of the full data set without compressing everything into one window. Use the scroll bars located below and to the right of the Event Window to move the display to the portion of the record you want to view. GistView attempts to scale the event sequence displays in order to fit as many on the screen as possible, but with data files containing many event sequences this isn't always possible.

Indicator

Notice the vertical black line in the EventWindow and the Slider at the bottom of the Event Window. They together called a Indicator. You can move the line by draging the Slider.

Profiling your data

The upper right panel in the GistView window contains a profile graph. When program start up, it's empty. You can activate the graph by moving the Indicator to a certain time point, clicking ProfileGraph Button in the Tool Bar. Doing so will bring up a graph showing the profile of each PID at the point the Indicator specified. The profile graph displays data about each of the profiled event sequences, one per column in the graph.

Viewing Events/States

GistView uses a event based view as its default. As you've seen, the actions recorded by the instrumented program are shown on screen as a series of time point, represented by colored lines. each of these lines is an event which was logged by the libgist library. The panel on the center of the right side of the GistView window has two buttons which let you select/deselect the State View and the Event View. Pressing the "States" button will give you a view of the events recorded in the program. Pressing it in this case should give you the following display:

Exiting GistView

Exiting the program is as simple as choosing Quit from the file menu. Saving your data file is not necessary as GistView doesn't modify the data.

GIST Parallel Performance Debugging Tool