just use s///, it will be just as fast

Not even close:

$from = 'abcd'; $to = '1234';;
@t = map 'axbxcxdx' x 10, 1 .. 1000;;

cmpthese -1,
{
    tr => q[
        my $tr = eval "sub{ \$_[0] =~ tr[$from][$to] }"; 
        $tr->( $_ ) for @{[ @t ]}; 
    ],
    s  => q[ 
        my %tr; @tr{ split'', $from } = split'', $to; 
        s[([$from])]{ $tr{$1} }eg for @{[@t]} 
    ],
};;

     Rate     s    tr
s  32.0/s    --  -97%
tr 1054/s 3189%    --


@t = map 'axbxcxdx' x 1000, 1 .. 1000;;

cmpthese -1,
{
    tr => q[
        my $tr = eval "sub{ \$_[0] =~ tr[$from][$to] }"; 
        $tr->( $_ ) for @{[ @t ]}; 
    ],
    s  => q[ 
        my %tr; @tr{ split'', $from } = split'', $to; 
        s[([$from])]{ $tr{$1} }eg for @{[@t]} 
    ],
};;
            (warning: too few iterations for a reliable count)

      s/iter      s     tr
s       2.65     --   -99%
tr 1.98e-002 13317%     --
[download]