It is helpful to use an actual program to experiment with
the performance analysis tools. If you do not have a program handy,
you can use the example program, vanderbilt, that is shipped with
HP/PAK. This program is used throughout this manual to illustrate
HP/PAK features.
Using vanderbilt |
 |
To use vanderbilt, invoke it in its install location or copy
it to your present working directory.
/opt/langtools/hppak/examples/vanderbilt
|
When invoking vanderbilt, you can control its execution time by supplying
an integer value as an argument. The integer specifies the number of
iterations for the program's execution. For example, to execute
vanderbilt ten times, type
If You Cannot Find vanderbilt |
|
It is possible that HP/PAK's directory of examples was removed to save
space after the tools were installed. If so, you will need to reinstall HP/PAK
to get vanderbilt. See the HP/PAK release document for instructions.
Using Another Example Program |
|
If you want to use a program other than vanderbilt to experiment with Puma,
it is best to compile the program for debugging (the -g compiler option).
The inclusion of debugging information enables you to examine performance data
at the statement level, as well as at the routine and library levels, and to
view the source code that corresponds to a given stack trace.
You may also wish to use a CMA threaded program, since Puma can separate the
performance data collected from multiple program threads. (The operating
system does not support kernel-threaded programs.)
Printable version
