On 12/13/2012 12:19 AM, Peter Geoghegan wrote:
> On 12 December 2012 22:11, Mikko Tiihonen
> <mikko.tiihonen@nitorcreations.com> wrote:
>> noticed a "XXX: It might be worth considering using an atomic fetch-and-add
>> instruction here, on architectures where that is supported." in lock.c
>>
>> Here is my first try at using it.
>
> That's interesting, but I have to wonder if there is any evidence that
> this *is* actually helpful to performance.
One of my open questions listed in the original email was request for help on
creating a test case that exercise the code path enough so that it any
improvements can be measured.
But apart from performance I think there are two other aspects to consider:
1) Code clarity: I think the lock.c code is easier to understand after the patch
2) Future possibilities: having the atomic_inc/dec generally available allows other performance critical parts of
postgrestake advantage of them in the future
-Mikko