Re: Weird performance issue with Strawberries and Inline::C

[Anonymous Monk]

How about this:

use strict;
use warnings;
use feature 'say';

use Benchmark 'cmpthese';

use constant L => 1e5;

my @cases;

my $str = '1' x L;
push @cases, $str;

substr $str, rand L, 1 , '0' for 1 .. L;
push @cases, $str;
push @cases, '01' x ( L / 2 );

say $^V;

for my $s ( @cases ) {
    cmpthese -2, {
        count => sub { $s =~ tr/1// },
    }
}

__END__

v5.32.1
         Rate count
count 21711/s    --
         Rate count
count 21607/s    --
         Rate count
count 21348/s    --


v5.42.0
         Rate count
count 16271/s    --
        Rate count
count 2422/s    --
        Rate count
count 8136/s    --

String composition affects as trivial operation as counting. Looks like someone not heeded the advice to "write like C3, not C4" and relied on compiler voodoo, when writing Perl itself, not silly solution for PWC. Though of course it may be something else rather than innocent "if" with its jumps. Maybe situation can't be shrugged off after all.

[ikegami]

What's with the rand? Your results are meaningless. The results from the different versions can't be compared to each other, and the results from one version can't be compared to each other. Finally, none of your test cases uses OOK scalars, and the whole question is about OOK scalars.