Table of Contents
Advertisement
Making Software Performance Measurements
To set up the trace command for duration measurements
When measuring a recursive function, module duration will be measured
between the last recursive call and the true end of the recursive execution.
This will affect the accuracy of the measurement.
If a module is entered at the normal point, and then exited by a point other
than the defined exit point, the entry point will be ignored. It will be judged
the same as any other unused prefetch, and no time-duration measurement
will be made. Its time will be included in the measure of time spent outside
the procedure or function.
If a module is exited from the normal point, and then re-entered from some
other point, the exit will also be assumed to be an unused prefetch of the exit
state.
Note that if you are making duration measurements on a function that is
recursive, or one that has multiple entry and/or exit points, you may wind up
with invalid information.
This section describes how to:
•
Set up the trace command for duration measurements.
•
Initialize duration performance measurements.
•
Interpret duration measurement reports.
To set up the trace command for duration
measurements
Specify a trace display depth of 512.
1
Trace after and store only function start and end addresses.
2
For duration measurements, the trace command must be set up to store only
the entry and exit points of the module of interest. Since the trigger state is
always stored, you should trigger on the entry or exit points. For example:
trace after symbol_entry or symbol_exit only
symbol_entry or symbol_exit counting time
The previous command depends on the generation of correct exit address
C A U T I O N
symbols by the software development tools.
276
Advertisement
Table of Contents
