"Simon Riggs" <simon@2ndquadrant.com> writes:
> Best way is to prove it though. Seems like not too much work to have a
> private ring data structure when the hint is enabled. The extra
> bookeeping is easily going to be outweighed by the reduction in mem->L2
> cache fetches. I'll do it tomorrow, if no other volunteers.
[ shrug... ] No one has yet proven to my satisfaction that L2 cache has
anything to do with this. The notion that you can read a new disk page
into a shared buffer and have that buffer still be live in the processor
cache is so obviously bogus that I think there must be some other effect
at work.
regards, tom lane