You could try this

#!/usr/bin/perl --

use strict;
use warnings;

use Benchmark 'cmpthese';

{
    my $small = join'', map { ' ' x  (int rand 2) . ','.  (int rand 2)
+.$_ } 0 .. 1_000;
    cmpthese(-3,{
        'small1' => sub { my $f = $small; $f =~ s/\s+,\s+|\s+,|,\s+/,/
+g; },
        'small2' => sub { my $f = $small; $f =~ s/\s*,\s*/,/g; },
    });
    undef $small;
}
print "\n";
{
# 1_000_000 was too much for my machine
    my $bigg = join'', map { ' ' x  (int rand 2) . ','.  (int rand 2).
+$_ } 0 .. 200_000;

    cmpthese(-3,{
        'bigg1' => sub { my $f = $bigg; $f =~ s/\s+,\s+|\s+,|,\s+/,/g;
+ },
        'bigg2' => sub { my $f = $bigg; $f =~ s/\s*,\s*/,/g; },
    });
    undef $bigg;

}
__END__
perl 5.10.1

         Rate small1 small2
small1  365/s     --   -69%
small2 1171/s   221%     --

        Rate bigg1 bigg2
bigg1 1.27/s    --  -77%
bigg2 5.41/s  327%    --


perl 5.8.8

         Rate small1 small2
small1  785/s     --   -51%
small2 1587/s   102%     --

        Rate bigg1 bigg2
bigg1 2.59/s    --  -62%
bigg2 6.86/s  165%    --
[download]