use strict;

use Test::More qw( no_plan );

sub nobull {
    my ($val1,$val2) = @_;
    return $val1 * 10000 + $val2;
}

sub dragonchild {
    join'', map sprintf("%04d", $_), @_
}

my @tests = (
    [ '3214', '0578', '32140578' ],
    [ '3214', '0000', '32140000' ],
    [ '2340', '0230', '23400230' ],
    [ '3214', '6578', '32146578' ],
    [ '3210', '6578', '32106578' ],
    [ '2130', '1000', '21301000' ],
    [ '0004', '6789', '00046789' ],
);

foreach my $test (@tests) {
    is( nobull( $test->[0], $test->[1] ), $test->[2] );
    is( dragonchild( $test->[0], $test->[1] ), $test->[2] );
}
[download]

Update: Expanded the number of tests being handled.

Correctness is in the eye of the beholder. Since the original author said:

then converted back to number by adding 0.

then nobull's implementation is a smidgeon faster than yours since nobull's implementation doesn't need to undergo the extra string to number transformation that yours does. Even if the speed difference is nil, nobull's implementation is also conceptually what zentara wanted, as implied by the "adding 0" bit above.

Yea, the question (and it remains so in all of these) is that the first section is at least 1000.

Otherwise it wouldn't really be 8 bytes digits to begin with (not going to argue semantics here).

I kind of assumed that. Otherwise you'll just have to pad both sides and concat or multiply left by 10000, add to right and pad left.

Update: You could just

sprintf( "%08d", $var1*10000+$var2 )

and call it a day, but that's actually slower than any of the others :(