[Boost-users] BGL performance: is pool_allocator possible and does it help iteration performance?