When collecting performance data, it's possible to specify the experiment directory by using the -O option:

gprofng collect app -O experiment-directory-name.er /path/to/application -options

Note that .er is a required suffix for any experiment directory name.

For more information about collecting performance data, see Getting Started With gprofng.

To analyze performance data collected for each function, use the -functions option with the gprofng display text command:

gprofng display text -functions experiment-directory-name.er 

Functions sorted by metric: Exclusive Total CPU Time

Excl. Total   Incl. Total    Name
CPU           CPU
 sec.      %   sec.      %
5.554 100.00  5.554 100.00   <Total>
5.274  94.95  5.274  94.95   mxv_core
0.140   2.52  0.270   4.86   init_data
0.090   1.62  0.110   1.98   erand48_r
0.020   0.36  0.020   0.36   __drand48_iterate
0.020   0.36  0.130   2.34   drand48
0.010   0.18  0.010   0.18   _int_malloc
0.      0.    0.280   5.05   __libc_start_main
0.      0.    0.010   0.18   allocate_data
0.      0.    5.274  94.95   collector_root
0.      0.    5.274  94.95   driver_mxv
0.      0.    0.280   5.05   main
0.      0.    0.010   0.18   malloc
0.      0.    5.274  94.95   start_thread

To limit the number of functions displayed, use the -limit option as follows:

gprofng display text -limit 5 -functions experiment-directory-name.er 

Print limit set to 5
Functions sorted by metric: Exclusive Total CPU Time

Excl. Total   Incl. Total    Name
CPU           CPU
 sec.      %   sec.      %
5.775 100.00  5.775 100.00   <Total>
5.494  95.15  5.494  95.15   mxv_core
0.126   2.18  0.267   4.63   init_data
0.068   1.17  0.104   1.80   erand48_r
0.038   0.66  0.142   2.45   drand48

To list all the metrics that have been collected in an experiment directory, use the -metric_list option without any other options:

gprofng display text -metric_list experiment-directory-name.er

Current metrics: e.%totalcpu:i.%totalcpu:name
Current Sort Metric: Exclusive Total CPU Time ( e.%totalcpu )
Available metrics:
Exclusive Total CPU Time: e.%totalcpu
Inclusive Total CPU Time: i.%totalcpu
                    Size: size
              PC Address: address
                    Name: name

After you have established which metrics have been collected in an experiment directory, select the displayed metrics by using the -metrics options, separating each with a : character:

gprofng display text -metrics name:i.%totalcpu:e.%totalcpu -limit 10 -functions experiment-directory-name.er 

Current metrics: name:i.%totalcpu:e.%totalcpu
Current Sort Metric: Exclusive Total CPU Time ( e.%totalcpu )
Print limit set to 10
Functions sorted by metric: Exclusive Total CPU Time

Name                Incl. Total   Excl. Total
                    CPU           CPU
                     sec.      %   sec.      %
 <Total>            5.775 100.00  5.775 100.00
 mxv_core           5.494  95.15  5.494  95.15
 init_data          0.267   4.63  0.126   2.18
 erand48_r          0.104   1.80  0.068   1.17
 drand48            0.142   2.45  0.038   0.66
 __drand48_iterate  0.036   0.62  0.036   0.62
 _int_malloc        0.013   0.22  0.008   0.14
 sysmalloc          0.005   0.09  0.003   0.05
 brk                0.002   0.03  0.002   0.03
 __default_morecore 0.002   0.03  0.      0.

To sort the performance data according to a specific metric, for example name, add the -sort option:

gprofng display text -metrics name:i.%totalcpu:e.%totalcpu -sort name -limit 10 -functions experiment-directory-name.er 

Current metrics: name:i.%totalcpu:e.%totalcpu
Current Sort Metric: Exclusive Total CPU Time ( e.%totalcpu )
Current Sort Metric: Name ( name )
Print limit set to 10
Functions sorted by metric: Name

Name                Incl. Total   Excl. Total
                    CPU           CPU
                     sec.      %   sec.      %
 <Total>            5.775 100.00  5.775 100.00
 __default_morecore 0.002   0.03  0.      0.
 __drand48_iterate  0.036   0.62  0.036   0.62
 __libc_start_main  0.280   4.85  0.      0.
 _int_malloc        0.013   0.22  0.008   0.14
 allocate_data      0.013   0.22  0.      0.
 brk                0.002   0.03  0.002   0.03
 collector_root     5.494  95.15  0.      0.
 drand48            0.142   2.45  0.038   0.66
 driver_mxv         5.494  95.15  0.      0.

Use the -disasm option and specify the function name to review metrics at an instruction or assembly level:

gprofng display text -metrics e.totalcpu -disasm function-name experiment-directory-name.er

For more information, see https://sourceware.org/binutils/wiki/gprofng and the gprofng(1) manual page.

To review an experiment containing performance data in a graphical user interface, use the gprofng display gui command, which is provided by the gprofng-gui package:

gprofng display gui experiment1.er

If more than one experiment is specified, they're aggregated together.

To compare two experiments containing performance data in a graphical user interface, use the -c option:

gprofng display gui -c experiment1.er experiment2.er

For more information about working with the gprofng-gui tool, see Using the Gprofng GUI.