I found a test script from 20 years ago here Re: array pre-allocation trick (top of thread array pre-allocation trick) and tweaked it to add a version explicitly assigning to the last item (with "last" in the name). Machine is of course faster so the rate's much different, but the numbers are maybe similar.

#!/usr/bin/perl

use strict;
use warnings;

use Benchmark qw /cmpthese timethese/;

our $size = 100_000;

cmpthese timethese(
    -10 => {
        push       => 'my @arr; push @arr => $_ for 0 .. $::size - 1',
        assign     => 'my @arr; $arr [$_] =  $_ for 0 .. $::size - 1',
        assignpre  => 'my @arr; $#arr = $::size - 1;
                            $arr [$_] =  $_ for 0 .. $::size - 1',
        assignlast => 'my @arr; $arr[ $::size - 1 ] = undef;
                            $arr [$_] =  $_ for 0 .. $::size - 1',
        slice      => 'my @arr; @arr [0 .. $::size - 1] = (0 .. $::siz
+e - +1)',
        slicepre   => 'my @arr; $#arr = $::size - 1;
                            @arr [0 .. $::size - 1] = (0 .. $::size - 
++1)',
        slicelast  => 'my @arr; $arr[ $::size - 1 ] = undef;
                            @arr [0 .. $::size - 1] = (0 .. $::size - 
++1)',
    } => 'none'
);
__END__

            Rate slicepre   slice slicelast assignpre  assign assignla
+st    push
slicepre   124/s       --    -10%      -11%      -21%    -27%       -2
+8%    -35%
slice      138/s      11%      --       -0%      -12%    -19%       -2
+0%    -28%
slicelast  138/s      12%      0%        --      -11%    -18%       -2
+0%    -28%
assignpre  156/s      26%     13%       13%        --     -8%        -
+9%    -19%
assign     169/s      37%     23%       22%        8%      --        -
+2%    -12%
assignlast 172/s      39%     25%       24%       10%      2%         
+--    -10%
push       192/s      55%     39%       39%       23%     13%        1
+1%      --


## Same comparisons node 209382 on my machine (maybe bit more apples t
+o apples)
           Rate  slicepre     slice assignpre    assign      push
slicepre  123/s        --      -10%      -21%      -25%      -36%
slice     137/s       11%        --      -12%      -17%      -29%
assignpre 156/s       27%       14%        --       -5%      -19%
assign    165/s       34%       20%        6%        --      -14%
push      192/s       56%       40%       23%       17%        --
[download]

Thanks, but honestly I don't plan looking into the details.

I second what Schwern said in the SO discussion:

(Only try this if ) You profiled and found a problem.

It would need MANY "40KB arrays" to make this count on a modern machine.

Whatever a "40KB array" means, arrays hold refs to other types, it's totally unclear what is meant here.

This smells like a problem which is best solved with packing into a 40 KB string.

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}