Devel::Profile - tell me why my perl program runs so slowly


    perl -d:Profile
    less prof.out


The Devel::Profile package is a Perl code profiler. This will collect information on the execution time of a Perl script and of the subs in that script. This information can be used to determine which subroutines are using the most time and which subroutines are being called most often.

To profile a Perl script, run the perl interpreter with the -d debugging switch. The profiler uses the debugging hooks. So to profile script the following command should be used:

        perl -d:Profile  

When the script terminates (or periodicly while running, see ENVIRONMENT) the profiler will dump the profile information to a file called prof.out. This file is human-readable, no additional tool is required to read it.

Note: Statistics are kept per sub, not per line.



How often to save profile data while running, in seconds, 0 to save only at exit. The default is every 2 minutes.


Filename to save profile data to, default is prof.out


Time spent running code not in 'subs' (such as naked code in main) won\'t get accounted for in the normal manner. By default, we account for this time in the sub '<other>'. With this variable set, we leave it as 'missing' time. This reduces the effective runtime of the program, and the calculated percentages.


example ouput:

    time elapsed (wall):   86.8212
    time running program:  65.7657  (75.75%)
    time profiling (est.): 21.0556  (24.25%)
    number of calls:       647248
    %Time    Sec.     #calls   sec/call  F  name
    31.74   20.8770     2306   0.009053     Configable::init_from_config
    20.09   13.2116   144638   0.000091     Configable::init_field_from_config
    17.49   11.5043   297997   0.000039     Configable::has_attr
     8.22    5.4028      312   0.017317     MonEl::recycle
     7.54    4.9570    64239   0.000077     Configable::inherit
     5.02    3.3042   101289   0.000033     MonEl::unique

This is a small summary, followed by one line per sub.

time elapsed (wall)

This is the total time elapsed.

time running program

This is the amount of time spent running your program.

time profiling

This is the amount of time wasted due to profiler overhead.

number of calls

This is the total number of subroutine calls your program made.

Followed by one line per subroutine.


The name of the subroutine.


The percentage of the total program runtime used by this subroutine.


The total number of seconds used by this subroutine.


The number of times this subroutine was called.


The average number of seconds this subroutines takes each time it is called.




pseudo-function to account for otherwise unacounted for time.


At least one call of this subroutine did not return (typically because of an exit, or die). The statistics for it may be slightly off.


At least one call of this subroutine trapped an exception. The statistics for it may be slightly off.


This module was written so that the author could profile a large long-running (daemon) program. Since normally, this program never exited, saving profiling data only at program exit was not an interesting option. This module will save profiling data periodically based on $PERL_PROFILE_SAVETIME, or the program being profiled can call DB::save() at any time. This allows you to watch your profiling data while the program is running.

The above program also had a very large startup phase (reading config files, building data structures, etc), the author wanted to see profiling data for the startup phase, and for the running phase seperately. The running program can call DB::reset() to save the profiling data and reset the statistics. Once reset, only "stuff" that happens from that point on will be reflected in the profile data file.

By default, reset is attached to the signal handler for SIGUSR2. Using a perl built with "safe signal handling" (5.8.0 and higher), you may safely send this signal to control profiling.


There are no known bugs in the module.


    Yellowstone National Park.


Jeff Weisberg -