You're doing extremely simple constant arithmetic here. GHC can optimize this type of thing away to nothing as well. Are we talking about contrived examples or real numerical bottlenecks?