A few pretty much orthogonal bits.

• Order the list of surnames by frequency, commonest first.
• Think about index. The surnames are fixed strings.
• @foo = substr( $_, $tart, $offset) =~ /$re/; has a lot to recommend it.
• The m//sg form suggests you are slurping the file. Unless the records are messy, it's often better to process each record as it's read.
• Benchmark. Regex and search code performance is subject to subtlety, and small things can make a big difference.

After Compline,
Zaxo