Rewritten in Cerl (C++ transliterated into something that a Perl interpreter in a good mood may care to execute):

my $myword;
my $wlist  = "dict.txt";
my $strLen = length $myword;

open my $cfile, '<', $wlist or die "Can't open $wlist: $!\n";

while (defined (my $line = <$cfile>)) {
    next if $strLen != length $line;
    my @match;
    my @used;
    my $matchcount = 0;

    for (my $i = 0; $i < $strLen; $i++) {
        my $ch = substr $myword, $i, 1;
        for (my $y = 0; $y < $strLen; ++$y) {
            next if $ch ne substr $line, $y, 1;

            if ($used[$y] eq $ch) {
                $match[$matchcount] = 0;
            } else {
                $used[$y]           = $ch;
                $match[$matchcount] = 1;
                $matchcount++;
                last;
            }
        }
    }

    my $c = 0;

    for (my $t = 0; $t < $strLen; $t++) {
        ++$c if ($match[$t]);
    }

    return $line if ($c == $strLen);
}

close $cfile;
[download]

This is of course completely untested and not what I would consider to be Perlish, although it is strict compliant. Note that the minor structure changes I made could actually be pushed back to the original C++ code to clean up that code somewhat.