Perl ABC

1. Perl
›
2. Module
›
3. here

Benchmark - Benchmark (performance comparison)

The Benchmark module allows you to benchmark (performance comparison).

# Loading modules and importing functions
use Benchmark qw/timethese cmpthese/;

You can get the performance information of each function by executing the timethese function. If you specify this as an argument to the compthese function, the standard output will display the comparison result.

# Performance comparison
my $result = timethese ($count, {
  name1 => sub {
    # Process you want to execute 1
  },
  name2 => sub {
    # Process you want to execute 2
  },
  name3 => sub {
    # Process you want to execute 3
  },
});

cmpthese $result;

This is an example for benchmarking. It is necessary to increase the number of executions to some extent to improve accuracy. This example has been run 10000 times.

# Benchmark example
my $result = timethese (10000, {
  loop1 => sub {
    my $i = 0;
    while ($i < 1500) {
      $i++;
    }
  },
  loop2 => sub {
    my $i = 0;
    while ($i < 1000) {
      $i++;
    }
  },
  loop3 => sub {
    my $i = 0;
    while ($i < 500) {
      $i++;
    }
  },
});

cmpthese $result;

The output result is as follows. It can be confirmed that loop3 is nearly 3 times faster than loop1.

# Benchmark output result
Benchmark: timing 10000 iterations of loop1, loop2, loop3 ...
     loop1: 4 wallclock secs (2.98 usr + 0.00 sys = 2.98 CPU) @3351.21/s (n = 10000)
     loop2: 2 wallclock secs (1.75 usr + 0.00 sys = 1.75 CPU) @5714.29/s (n = 10000)
     loop3: 1 wallclock secs (1.03 usr + 0.00 sys = 1.03 CPU) @9699.32/s (n = 10000)
        Rate loop1 loop2 loop3
loop1 3351/s - - 41% -65%
loop2 5714/s 71% - - 41%
loop3 9699/s 189% 70% -

FAQ about Benchmark module

Q. Does the Benchmark module calculate the actual processing time?

A. The time displayed by wallclock is the actual time taken. The value after @is calculated as "number of trials/CPU time".

The value after @is calculated by the formula "number of trials/CPU time". CPU time is the time of the CPU used by the process, so it does not include I/O waits such as database access or time to execute child processes. Also, when sleep function is executed, it is not included in the execution time because it does not use the CPU.

This means that the value after @cannot be used as a reference when benchmarking I/O or calling child processes. The value compared by cmpthese is also "number of trials/CPU time", so the displayed value cannot be used for evaluation as it is.

If you are doing I/O or running a child process, you need to look at the value displayed on the wallclock. However, in most cases, you will use the Benchmark module for the purpose of evaluating the algorithm of the program, so you can use the value compared by cmpthese as a guide.