<
< * Consider padding the spinlock structure to have spinlock registers
<   on different CPU cache line
<
<   http://archives.postgresql.org/pgsql-hackers/2005-10/msg00498.php