| |||
| Home > Tracing Tutorial > Capturing profiling information > Summing profiling data over multiple runs | |||
To sum profiling data over multiple runs, do the following:
If you have not yet done so, perform the procedure described in Capturing the initial profiling information.
Select Sum Profiling Data from the Profiling Data menu of the Analysis window.
Select Set PC to Entry Point from the Debug menu of the Code window.
If you reload the image, or load a new image, the trace details are cleared and Sum Profiling Data is disabled.
Click Run on
the Debug toolbar of the Code window to start execution.
When prompted, enter 5000 for the number of runs through the benchmark. The profiling data is updated. An example is shown in Figure 10.92.
If you no longer want to sum profiling data, select Sum Profiling Data from the Profiling Data menu of the Analysis window.
Compare Figure 10.92 with Figure 10.89. The summing of profiling data reveals the following information:
The Count column reveals the following:
After the first run:
functions Proc_7 and Func_1 were
called the most times
all other functions were called a similar number of times.
The second run shows the same trend as the first run.
The B=>E% column reveals the following:
The total execution time after two runs is equal to the total execution time of the first run, plus the total execution time of the second run.
The time taken to execute a particular function after two runs is equal to the time taken to execute it after the first run.
The percentage of the total execution time that is spent in the code of a particular function is the same as that after the first run. However, this might not always be the case.
The Exec% column reveals the following:
The total execution time after two runs is equal to the total execution time of the first run, plus the total execution time of the second run.
The time spent from entry to exit of a particular function after two runs is equal to the time spent from entry to exit of that function after the first run.
The percentage of the total execution time that is spent in calls to a particular function is the same as that after the first run. However, this might not always be the case.