Re: better atomics - v0.6

On Thu, Sep 25, 2014 at 6:03 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> On 2014-09-24 20:27:39 +0200, Andres Freund wrote:
>> On 2014-09-24 21:19:06 +0300, Heikki Linnakangas wrote:
>> I won't repost a version with it removed, as removing a function as the
>> only doesn't seem to warrant reposting it.
>
> I've fixed (thanks Alvaro!) some minor additional issues besides the
> removal and addressing earlier comments from Heikki:
> * removal of two remaining non-ascii copyright signs. The only non ascii
> name character that remains is Alvaro's name in the commit message.
> * additional comments for STATIC_IF_INLINE/STATIC_IF_INLINE_DECLARE
> * there was a leftover HAVE_GCC_INT_ATOMICS reference - it's now split
> into different configure tests and thus has a different name. A later
> patch entirely removes that reference, which is why I'd missed that...
>
> Heikki has marked the patch as 'ready for commiter' in the commitfest
> (conditional on his remarks being addressed) and I agree. There seems to
> be little benefit in waiting further. There *definitely* will be some
> platform dependant issues, but that won't change by waiting longer.
>
> I plan to commit this quite soon unless somebody protests really
> quickly.
>
> Sorting out the issues on platforms I don't have access to based on
> buildfarm feedback will take a while given how infrequent some of the
> respective animals run... But that's just life.

I feel like this could really use a developer README: what primitives
do we have, which ones are likely to be efficient or inefficient on
which platforms, how do atomics interact with barriers, etc.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company