Quick & dirty bcrypt benchmark

bcryptbench.pl

#!/usr/bin/perl

use 5.12.0;
use strict;
use warnings;

use Benchmark ':hireswallclock';

my $time;
my $rounds = 0;
do {
        $rounds++;
        $time = Benchmark::timeit(1, sub {
                crypt('foobar', "\$2b\$$rounds\$01234567890123456789012")
        });
        say "For $rounds it took ".timestr($time);
} while ($time->real < 30);

As an example, here is the result for a i7-7820HQ (a CPU from 2017):

$ ./bcryptbench.pl
For 1 it took 6.98566e-05 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 2 it took 9.53674e-07 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 3 it took 2.28882e-05 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 4 it took 1.19209e-05 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 5 it took -4.17233e-05 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 6 it took 1.12057e-05 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 7 it took -1.50204e-05 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 8 it took 7.86781e-06 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 9 it took 3.09944e-06 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
For 10 it took 0.056061 wallclock secs ( 0.05 usr + 0.00 sys = 0.05 CPU) @ 20.00/s (n=1)
For 11 it took 0.108737 wallclock secs ( 0.10 usr + 0.00 sys = 0.10 CPU) @ 10.00/s (n=1)
For 12 it took 0.223427 wallclock secs ( 0.22 usr + 0.00 sys = 0.22 CPU) @ 4.55/s (n=1)
For 13 it took 0.435473 wallclock secs ( 0.43 usr + 0.00 sys = 0.43 CPU) @ 2.33/s (n=1)
For 14 it took 0.884831 wallclock secs ( 0.88 usr + 0.00 sys = 0.88 CPU) @ 1.14/s (n=1)
For 15 it took 1.75997 wallclock secs ( 1.76 usr + 0.00 sys = 1.76 CPU) @ 0.57/s (n=1)
For 16 it took 3.50631 wallclock secs ( 3.50 usr + 0.00 sys = 3.50 CPU) @ 0.29/s (n=1)
For 17 it took 7.03564 wallclock secs ( 7.03 usr + 0.00 sys = 7.03 CPU) @ 0.14/s (n=1)
For 18 it took 13.9885 wallclock secs (13.98 usr + 0.00 sys = 13.98 CPU) @ 0.07/s (n=1)
For 19 it took 28.1749 wallclock secs (28.17 usr + 0.00 sys = 28.17 CPU) @ 0.04/s (n=1)
For 20 it took 55.8184 wallclock secs (55.81 usr + 0.00 sys = 55.81 CPU) @ 0.02/s (n=1)

On an old X230 (Core i5-3320M from 2012):

$ curl -Lfs https://gist.github.com/TobiX/d152d818ab14713e181f1fcfc39bf653/raw/b26396d865b42ae9b478c18703e8a96079dd78b9/bcryptbench.pl | perl
For 1 it took 4.29153e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 2 it took 2.31266e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 3 it took 5.00679e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 4 it took 2.02656e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 5 it took 1.81198e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 6 it took 7.86781e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 7 it took 1.71661e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 8 it took 2.81334e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 9 it took 1.71661e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 10 it took 0.0676622 wallclock secs ( 0.06 usr +  0.00 sys =  0.06 CPU) @ 16.67/s (n=1)
For 11 it took 0.132322 wallclock secs ( 0.13 usr +  0.00 sys =  0.13 CPU) @  7.69/s (n=1)
For 12 it took 0.263124 wallclock secs ( 0.26 usr +  0.00 sys =  0.26 CPU) @  3.85/s (n=1)
For 13 it took 0.537851 wallclock secs ( 0.53 usr +  0.00 sys =  0.53 CPU) @  1.89/s (n=1)
For 14 it took 1.0945 wallclock secs ( 1.09 usr +  0.00 sys =  1.09 CPU) @  0.92/s (n=1)
For 15 it took 2.16098 wallclock secs ( 2.15 usr +  0.00 sys =  2.15 CPU) @  0.47/s (n=1)
For 16 it took 4.27836 wallclock secs ( 4.27 usr +  0.00 sys =  4.27 CPU) @  0.23/s (n=1)
For 17 it took 8.74663 wallclock secs ( 8.72 usr +  0.01 sys =  8.73 CPU) @  0.11/s (n=1)
For 18 it took 17.4105 wallclock secs (17.38 usr +  0.01 sys = 17.39 CPU) @  0.06/s (n=1)
For 19 it took 35.8238 wallclock secs (35.59 usr +  0.02 sys = 35.61 CPU) @  0.03/s (n=1)

Another datapoint, Ryzen 7 3700X (2019):

For 1 it took 2.90871e-05 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 2 it took 4.76837e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 3 it took 5.24521e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 4 it took 1.19209e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 5 it took 5.24521e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 6 it took 8.10623e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 7 it took 3.09944e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 8 it took 4.29153e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 9 it took 3.8147e-06 wallclock secs ( 0.00 usr +  0.00 sys =  0.00 CPU)
For 10 it took 0.044385 wallclock secs ( 0.04 usr +  0.00 sys =  0.04 CPU) @ 25.00/s (n=1)
For 11 it took 0.088104 wallclock secs ( 0.08 usr +  0.00 sys =  0.08 CPU) @ 12.50/s (n=1)
For 12 it took 0.175497 wallclock secs ( 0.17 usr +  0.00 sys =  0.17 CPU) @  5.88/s (n=1)
For 13 it took 0.352071 wallclock secs ( 0.35 usr +  0.00 sys =  0.35 CPU) @  2.86/s (n=1)
For 14 it took 0.704556 wallclock secs ( 0.70 usr +  0.00 sys =  0.70 CPU) @  1.43/s (n=1)
For 15 it took 1.40713 wallclock secs ( 1.40 usr +  0.00 sys =  1.40 CPU) @  0.71/s (n=1)
For 16 it took 2.8093 wallclock secs ( 2.80 usr +  0.00 sys =  2.80 CPU) @  0.36/s (n=1)
For 17 it took 5.60176 wallclock secs ( 5.60 usr +  0.00 sys =  5.60 CPU) @  0.18/s (n=1)
For 18 it took 11.1452 wallclock secs (11.14 usr +  0.00 sys = 11.14 CPU) @  0.09/s (n=1)
For 19 it took 22.4514 wallclock secs (22.44 usr +  0.00 sys = 22.44 CPU) @  0.04/s (n=1)
For 20 it took 45.6537 wallclock secs (45.64 usr +  0.00 sys = 45.64 CPU) @  0.02/s (n=1)