In the Call Tree Analysis windows, you can direct Puma to display performance
metrics by procedure (routine). This section describes the naming conventions that Puma
uses when displaying procedure names.
When Puma can determine a procedure's name, it displays the name in case-correct
format. For example:
For procedures in dynamically loaded libraries, Puma cannot identify the procedure
names unless you identify the library before data collection begins (using the
Options menu on the Data Collection window). The default entry in the
Libraries option is *.c. This entry causes Puma to collect data from all
dynamically loaded libraries, but, without the actual library names, it does not
load their symbol tables. Puma cannot, therefore, determine the names of the
procedures defined in the libraries. In this situation, Puma displays the full
pathname of the dynamically loaded library rather than the procedure name. For
example:
There are two conditions under which Puma is unable to determine the name of the
procedure associated with a call frame:
The procedure's name was omitted from the binary, as dictated by
the options used in compilation and linking.
The procedure has a nonstandard call frame format (for example,
for some assembly language procedures).
In this situation, Puma displays the virtual address of the procedure rather than
its name. For example:
where xxxx is the va (virtual address) of the procedure.