xlcb Graphical Browser:
The graphical version of LCB, xlcb, is invoked with a single argument,
naming the lightweight corefile
to be used as input:
xlcb <lightweight-corefile-name> &
xlcb vector_manipulation.lcb &
The initial window that comes up is the Overview Graph. It presents the
calling structure of the program as a tree, or call graph. Each node in
the graph represents a routine in the call stack of one or more processes,
with each node colored
according to the number of active processes associated with it at the
time of program termination.
The node colored in red indicates the location of the culprit process.
To get more information on a routine, select its corresponding node by clicking
on it. When selected, the node is highlighted (it shows up as white in the
pictures below) and the message area
displays the number of processes active in the corresponding routine.
A Color Key window
also appears within the Overview Graph view when xlcb starts up.
It provides the legend for the node colors, and its message field
displays the executable name, culprit
process identifier, the reason for failure and the location of
failure. You can show or hide this window at will via the Options menu.
Click here to see the names of the different areas
in the Overview Graph view.
By clicking the Call Graph button at the bottom of the Overview Graph
you can change to a more detailed representation of the program's calling
structure. This view
gives a zoomed-in version of the call graph, where each node displays the
The Call Graph view shown below includes a
Routine Detail window, which supplies process id number and last recorded source
location for each process that was executing in the selected routine.
This window appears when a node is selected for the first time in the Call
Graph and Overview Graph views.
You can hide or show this window at will via the Options menu.
By clicking the Culprit Only button at the bottom of the Overview Graph
view, you see the display of
the call stack of just the culprit process.
By clicking the Find Routine... button at the bottom of the Overview Graph
view, you get an alphabetic list of all routines represented in the current
view. You can then select any routine in the list, and the corresponding
node in the current view will be highlighted.
- To see an overall picture of the calling structure of a parallel
program when it failed, select the Overview Graph view by clicking its
button in the controls area.
- To change the layout of the call graph, choose the Horizontal
Vertical Layout entry in the Options menu.
- To see a zoomed-in version of the call graph showing the
routine names, select the Call Graph view by clicking on that button in the
- To see the call graph of only the culprit process, select the
Culprit Only view by clicking on that button in the controls area.
- To find all occurrences of a particular routine in the current graph,
use the Find Routine dialog.
If the dialog is not present, select the Find
Routine... button in the controls area. Scroll the
list if necessary to find the desired routine and select it. The
corresponding node will be highlighted; the display will scroll to reveal the
node, if it is not visible already. To find other occurrences of the same
routine, keep "selecting" it by clicking on the routine name.
- To find the number of processes that were active in a given routine,
click on a node corresponding to the routine in either the Overview or Call
Graph view. The message area will display the required information.
- To find the caller-callee information about a given routine, click on
the corresponding node in the Culprit Only view. The message
area will display the required information.
- To hide the Color Key window, choose the "Hide Color Key" entry in
the Options menu.
- To find information about each process present in a specific routine,
use the Routine Detail window. If the window is not present, select the
Detail Window entry in the Options menu. This window
displays each process identifier and its source location in the routine.
(This window is available only in the Overview Graph and Call Graph views.)
- To trace the calling sequence that led to the culprit's failure, follow
the path from the root of the graph (corresponding to the main program
module) to the culprit process in the any of the three views. This gives the
sequence of calls leading to the culprit routine. In Culprit Only view,
selecting each node in turn will also provide specifics on the calling
location within the routine.
- To view the executable name or the reason for failure, use the Color
Key window. If the window is not currently visible, select the Show Color
Key Window entry in the Options menu. The text in the window provides the
desired information. It also displays the culprit process identifier and the
source location of the failure.
Parallel Tools at OSU home page
This document was last updated 2 Aug. '95.
For further information, contact