Re: testing HS/SR - 1 vs 2 performance

Simon Riggs wrote:
> On Sun, 2010-04-18 at 08:24 +0100, Simon Riggs wrote:
>> On Sat, 2010-04-17 at 18:52 -0400, Tom Lane wrote:
>>> Simon Riggs <simon(at)2ndQuadrant(dot)com> writes:
>>>> What I'm not clear on is why you've used a spinlock everywhere when only
>>>> weak-memory thang CPUs are a problem. Why not have a weak-memory-protect
>>>> macro that does does nada when the hardware already protects us? (i.e. a
>>>> spinlock only for the hardware that needs it).
>>> Well, we could certainly consider that, if we had enough places where
>>> there was a demonstrable benefit from it. I couldn't measure any real
>>> slowdown from adding a spinlock in that sinval code, so I didn't propose
>>> doing so at the time --- and I'm pretty dubious that this code is
>>> sufficiently performance-critical to justify the work, either.
>> OK, I'll put a spinlock around access to the head of the array.
>
> v2 patch attached

Given the discussion about the cyclic nature of XIDs, it would be good
to add an assertion that when a new XID is added to the array, it is

a) larger than the biggest value already in the array
(TransactionIdFollows(new, head)), and
b) not too far from the smallest value in the array to confuse binary
search (TransactionIdFollows(new, tail)).

--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com